The discourse around the bidding war to acquire Warner Bros. Discovery is wrong. Having worked at WBD, I think many underestimate Netflix’s technological advantage and misunderstand who Netflix is actually competing against. This is all part of a broader shift in video being disrupted by technology.

Ben Thompson has some great analysis in his Dec 8 piece that I will be referencing heavily. He describes the changes that have happened in text and audio media that have already passed: an existing business model pre-tech, a period of transition as tech was able to disrupt, and a new equilibrium. To understand where video is going, we need to see where text and audio already went.

Comparing the changes to text, audio, and video

Text

Pre-1990s and early 2000s, the status quo was both aggregated and distributed regionally: local newspapers (maybe multiple) monopolized coverage on local sports teams and affairs, while national newspapers covered major stories.

Then, as technology became able to disrupt this business model, an awkward transition period appeared (2005-2015s). The NYTimes was struggling from being attacked on both ends, social media became prevalent, and new business models online like direct-to-consumer and ads became possible.

Now, a new equilibrium is present. Large aggregators like the NYTimes are a default source and subscription (for those who distrust social media for reporting in particular). Other niche publications like the Wall Street Journal and Financial Times have also rebounded. Ironically, traditional media companies became disruptors themselves, powered by cheaper distribution through technology, against smaller players.

On the long-tail side of market hyper-specific and more individualized sources have exploded to fill the void. Stratechery itself is a great example: Ben Thompson provides top shelf industry level analysis for a reasonable price. This is way easier to access as an individual instead of businesses compared to the past. Billy’s BBQ is another gem: Billy writes about Arsenal (London soccer team) with extreme detail, referencing annotated images and GIFs from matches, his own custom stats models during transfer windows, and player profiles. The quality of this writing probably far exceeds historical London Times sports columnists, and Billy’s work is free. I likely wouldn’t have even had access to anything like this 10 years ago, let alone 25. This exists for many teams and sports across the world. Here’s the counterintuitive part: the NYTimes is as responsible for killing local newspapers and smaller writers as platforms like X or Substack. The big aggregators didn’t just survive disruption - they became disruptors themselves, using cheaper digital distribution to squeeze out the middle.1

Text is the least information heavy medium to transmit on the internet. It makes sense that disruption happened to text ahead of audio and video. However, audio was not far behind...

Audio & Music

Music did not end up in a barbelled state like text media. Previously, radio acted as a main aggregator, TV channels as secondary aggregators, with physical distribution also acting as a growth channel.

Now, a few large aggregators like Spotify offer similar access via subscriptions. Direct to consumer platforms like Bandcamp haven’t gained mainstream adoption, as consumers prefer the newfound convenience of the aggregators. The hyper-specific end of the text barbell hasn’t formed for audio yet.

A way to understand this is through ‘consumption units.’ A news article is a discrete object that you somehow find, read, and finish. I don’t know many people who binge read articles 😅 (if you do you’re probably already a Stratechery sub). A song is a few minutes, but listening is not discrete. One doesn’t consume one song and often stop. People consume in continuous flows: playlists, albums, radio-style streams, background audio while working. The unit of consumption is the session, not track. This makes the aggregator necessary infrastructure for people’s listening habits, and is a big reason why every aggregator has roughly the same catalog due to licensing requirements. The moat are your playlists, your listening history, and algorithmic profile.

This is why music didn’t barbell the way text did: the consumption pattern itself--continuous, passive, session based--makes aggregation necessary. Reading requires active selection; you find an article, read it, and finish it. Listening is often the opposite. Most people prefer decisions about what to play next to be made for them. The direct creator-to-listener model works fine for merch, live performances, and some Patreon memberships, but not for the primary listening experience.

Once networks covered enough of the world and bandwidth was high enough for satisfactory listening, streaming services began to take off. This is a similar story for film and TV.

Film & TV

The Film and TV industry is currently in a huge state of flux. Netflix attempting to buy WBD is a big sign of the times.

Netflix has reaped the rewards of a massive technological advantage compared to most of the Hollywood incumbents over time, and hardly competes with the rest of the industry (HBOMax, Paramount+, Peacock, etc.). When I worked at WBD, we felt that Netflix was so far ahead they were bored.2 Traditional studios spent years reaching basic streaming competence. Table-stakes features like global availability, basic personalization, performant streaming were priorities before they could even begin competing on Netflix’s terms.

Owning Content vs. Leasing Content

But Netflix has a problem. Netflix is able to build audiences for content it does not own. Ben Thompson highlighted three examples:

• Drive to Survive: Netflix made Formula 1 popular in America, but does not own live race rights, where most of the money is. It built demand for a product someone else sells.

• Suits: licensed from NBCUniversal for cheap and became more popular on Netflix than during its original run. However, for any continued deals, Comcast will price higher and still owns the IP for any new spinoffs.

• K-Pop Demon Hunters: Sony licensed the movie to Netflix, who used their platform to create exponential growth. Sony still owns the IP and will profit off of merchandise and surely charge more for sequels.

Netflix and WBD

What Netflix gains from WBD is coverage across viewing modes: HBO’s prestige catalog for engaged, lean-forward watching (‘tent-pole’ content spikes attention share briefly); HGTV and Food Network for lean-back background viewing that can help prevent churn; Harry Potter and Cartoon Network for kids (an underrated revenue driver, especially for ad-supported tiers); and DC’s game studio for attention-share beyond video, which will pair nicely with their cloud gaming investments.

Netflix is not competing with Hollywood studios. It is competing with YouTube and TikTok. That is to say: the relevant scarce resource is not content. It is attention. Netflix is paying approximately $75 billion for WBD for roughly 1% of total viewing hours... to some, a massive overpay. This is especially expensive when recognizing that YouTube and TikTok own their content through creator relationships. Their economics are fundamentally different. Viewed through this lens, especially since Netflix does not have the distribution channels of a historic Hollywood studio, this should not cause anti-trust concerns or call for the end of Hollywood. I think changes will be slow, taking years to influence content strategy.

Predicted New Equilibrium

On the aggregator end, multiple platforms with different content mixes will survive. Unlike Spotify—where catalog parity forces competition on features and price—video exclusives create genuine differentiation. Netflix, Disney, Peacock, will coexist with different content portfolios. Consumers will choose preferred platforms or stack subscriptions.

A mix of tent-pole content (lean-forward weekly drops of trending IP) will continue to be paired with auxiliary content (lean-back old IP that people re-watch often). I hypothesize that this lean-back style of content (Suits, Friends, etc.) will continue to thrive for the next decade, but begin to decline. This will be because of generational viewing preferences: millennials were conditioned to like older IP and re-watch it, while Gen-Z and younger haven’t been. We’ve been 1) conditioned to not re-watch as often because of slot-machine-like dynamics around content and 2) grew up with passive viewing on YouTube from the start.

Subscription prices will approach cable bundle costs. Content creation is expensive... Hollywood has not become cheaper. It is much more expensive than distribution - whether it be cable or AWS servers. It makes sense that prices will rise to match actual production costs, since that is the primary driver of cost.

On the creator end, premium content distributed on social platforms. YouTube, TikTok, Instagram as venues for creator-made films and premium content, content that increasingly encroaches on ‘lean forward’ viewing habits while still maintaining the ‘lean back’ base. Pricing may take some time to get right... perhaps a few dollars to lease a film, probably with ads. The same content at different price points with varying quality and ad load could also be available. The burgeoning creator end is already happening.

• Nail House: RocketJump made low-budget series on YouTube as an early pioneer. Now producing a feature film, Nail House, that will be distributed directly or in theaters. Production was funded in hours through a mega crowdfund that promised behind-the-scenes access and merchandise as value adds. Others use Patreon for similar purposes.

• Two Sleepy People: A project from Camp Studios, a Gen-Z creator collective. They secured a theatrical limited release (opening weekend January 22) without entering the festival circuit and without marketing spend. They used their networks and strong production chops to generate enough interest. Straight from production to theaters through direct distribution.

Subscribe now

I don’t think these are anomalies - I think they are early signals of a potential video future. Text barbelled because distribution got cheap and consumption stayed discrete. Music didn’t barbell because consumption is continuous and passive. Video is somewhere in between and has space for both: lean-back content will stay aggregated, but lean-forward viewing might fragment the same way text did. They may be the hyper-specific end of the barbell finally growing.

The question I am unsure of is how much of ‘prestige TV’ is actually a consumption pattern that creators can serve directly. I’d bet that in five years, a creator-funded film will outperform a mid-budget studio release-not necessarily because it’s better, but because the creator already owns the audience. When the distribution gets cheap, the advantage shifts to whoever has the direct relationship.

Hollywood thinks it’s competing for content. It’s competing for attention, and it is losing. I'm curious whether others are seeing creator-to-audience models working in other industries… leave a comment!

Leave a comment

I’ve gathered my favorite media and memories of the year. My absolute highlights are marked in bold and with a star.

YouTube

• 🌟 Chris Young - a former 3 star Michelin chef who now makes a smart thermometer, and my favorite food/cooking YouTube videos to advertise it. The production value is insane and the topics lean into chemistry and physics

• Sharks Aren’t Fish - by Hank Green

• Cleo Abrams - an awesome science channel with a ton of access to ongoing research

• Howtown - two video journalists who used to work at Vox. I particularly like their ‘research showdowns’ (aka debates) which are structured well - I learn a lot from them!

• Extranet Shaquille - is the alt channel of a cooking YouTuber where he posts monologues on all sorts of topics. I love his pros and narration!

Podcasts

• 🌟 Gear, Articles of Interest - the most interesting podcast series I’ve listened to in a looong time, talks about military clothing, outdoor clothing, how they are linked, and how clothing represents both from 1900-today. Grapples with the role of the US military and how it is portrayed (with clothing) to the US public.

• Sarah Paine - full series on Dwarkesh Patel’s podcast. I didn’t know much about ‘grand strategy’ or how the military education system worked, but her talks about it are interesting. She’s a great storyteller!

• Maintaining Human Intelligence in the AI Era | David Krakauer (President of the Santa Fe Institute) on The Generalist.

• Dialectic - overall my favorite podcast right now, constantly getting my reading cinematic universe to talk for 2-3 hours! Ordered favorites from this year alone:

  • Tamara Winter - Tacit Trust & Caring Curiosity

  • Linus Lee - Engineering for Aliveness

  • Henrik Karlsson - Cultivating a Life that Fits

  • Nadia Asparouhova - Ideas That Infect

  • Geoffrey Litt - Software You Can Shape

Books

• 🌟 The Years of Lyndon Johnson - Robert Caro

• Flowers for Algernon - Daniel Keyes

• City of Fortune: How Venice Won and Lost a Naval Empire - Roger Crowley

• The Dispossessed - Ursula K. Le Guin

Leave a comment

Film & TV

• 🌟 The Rehearsal S2 - one of my favorite bits of TV ever... Nathan is crazy I love him

• One Battle After Another

• Haven’t seen yet, but excited about Marty Supreme & Zootopia 2

Music

• 🌟 Choke Enough - Oklou (I failed and was only top 2% globally on Spotify 🥲)

• Fear of Symmetry - Oneohtrix Point Never (insane visualizer by Uncanny)

• LUX - Rosalia

• Cycles - DJ_Dave

• Star - 2hollis

Food (NYC)

• 🌟 Naks (Filipino) - their Ube Pie dessert is delicious

• Danubius (Romanian)

• Ceres (Pizza)

• Sante Fe BK (New Mexican)

• Studio Bumi (popup communal dining)

• Thoughtful Coffee (popup cafe, may be opening a physical location next year!)

Share

Things to do in and around town (NYC)

• 🌟 Jack Whitten’s life-spanning exhibition at the MoMA

• 🌟 The Glass House and Grace Farms in New Canaan, CT. Do the extended tour at the Glass House!

• 1014 events - went to crazy dinner at the German Consulate General New York

• Will’s *Who Made This?!* art-making fundraiser for local charities

• A Walk Through the Universe at the Hayden Planetarium

• Carnegie Hall Symphonies

• The re-opened Breuer - although it was pretty disappointing imo... they kinda wasted the potential with the way they staged things

• Fort Greene Orchestra

Hi everyone,

Hope you are all doing well! I have been well: I have more time for running again, the weather is beautiful, and I have been reading plenty. Here’s what I’ve been up to:

I’ve also had time to continue learning through fun projects, and I have one to share with you all today! I made my very own bot… meet bot-drei!

Why?

I wanted to learn about model fine-tuning and RAG by doing. Building an AI version of myself seemed like a great place to start: personal enough to notice what worked and what didn’t, and I figured I had enough training data. I wanted to make bot-drei!1

Going in, I assumed that fine-tuning would be the hard part, and require loads of data. RAG across my blog posts and website? Should be simple enough - that sounds like a product problem!

Turns out the hard part wasn’t where I expected.

The question that emerged: how do you balance personality with accuracy when your knowledge base is actually pretty small? Give the model too much freedom and it hallucinates.2 Lock it down too hard and it becomes a glorified search engine with no personality. I wanted to find a middle path.

Fine-Tuning (making it sound like me)

This turned out to be straightforward. 10 minutes to train a small open-source model on 100-200 examples of my writing using LoRA on my Windows desktop with an old Nvidia GPU.3 Training, merging weights, changing model formats, all done flawlessly thanks to Unsloth.

LoRA (Low-Rank Adaptation) works by freezing almost all of a model’s existing weights and training only a few small “adapter” matrices that sit inside its attention layers. Instead of updating billions of parameters, it learns a low-rank approximation, basically a compact patch that nudges how the model represents relationships between words. That’s why you can fine-tune a 4B-parameter model4 so quickly: you’re only training a few million parameters while keeping everything else intact. When you merge the LoRA into the base model later, it behaves as if it had been fine-tuned. Kind of like tailoring a piece of clothing.

The model picked up my patterns, email and text style, and how I structure ideas. That was the easy part.

Leave a comment

The Routing Problem

Once I had a model that sounded like me, I needed it to answer questions about my actual work without making things up. The system routes queries based on how confident it is about retrieved context:

• Pre-filtering catches greetings and short queries - skip RAG, respond conversationally with personality.

• Full pipeline handles substantive questions through query expansion, intent detection, hybrid retrieval, and confidence-based response generation.

Three confidence bands determine response style:

• High (≥0.6): authoritative synthesis

• Medium (0.4-0.6): quote and contextualize with explicit limitations

• Low (<0.4): conversational fallback, honest about what the system can’t answer

Building the RAG System

This took many iterations across a few days. Getting the balance right was difficult, especially since measurement isn’t straightforward.

I also learned that my knowledge base is small: 10 blog posts, resume, LinkedIn profile, website content. This constraint created a core driving tension behind this system: let the fine-tuned model’s personality run free and it fills gaps with hallucination, or constrain it too hard to retrieval and you get a search engine that responds to your name…

How queries get processed

Query expansion adds synonyms and related terms. Someone asks about “projects” and the system searches for “work,” “portfolio,” “experience” too.

Intent detection routes questions to the right content types, like resume entries, blog posts, project descriptions. Metadata filtering (tags) prevents weird cross-contamination (so resume queries don’t surface personal blog posts).

Hybrid retrieval (the interesting part)

I use two complementary methods because each catches what the other misses:

• Dense embeddings (70% weight) handle semantic similarity. When someone asks “what has Andrei built?”, embeddings find content about projects and experience even if those exact words don’t appear. It understands meaning through vector distance (cosine similarity).

• Sparse BM25 (30% weight) catches exact keyword matches. If someone asks about “Sfânt” or “fine-tuning”, BM25 ensures content with those specific terms ranks highly. Pure semantic search sometimes misses exact technical terms or proper nouns.

The system combines both result sets with weighted scores, applies time decay5 to favor recent content, then reranks based on intent match. Retrieves up to 50 candidates, narrows to the 3 most relevant.

Subscribe now

Confidence

The confidence score drives how the model responds:

• High confidence (≥0.6): Temperature drops to 0.3, responses up to 1500 tokens. Authoritative synthesis in my voice, direct information delivery.

• Medium confidence (0.4-0.6): Same parameters but explicitly acknowledge limitations. Quote relevant passages and contextualize them.

• Low confidence (<0.4): Temperature back to 0.6, 500 tokens max. Skip RAG context injection entirely, respond conversationally, be honest about limitations. This happens a lot when people ask about things completely outside my content.

Chunking Strategy (surprisingly important)

Chunking determines how much context the model sees for each retrieval. Too small and you lose the thread of an argument. Too large and the embedding becomes generic, retrieval accuracy drops, and you waste tokens on irrelevant text.

Different content needs different treatment:

• Resume entries: no chunking. Job descriptions are already bulleted and summarized, they need full context.

• Projects and experience: 1000-token chunks. Structured portfolio data loses meaning when fragmented.

• Blog posts: 500 tokens with 100-token overlap, preserving heading boundaries so I don’t split mid-thought.

Deployment (pretty annoying)

I wanted to host the GGUF model somewhere with OpenAI-compatible API endpoints without spending hundreds of dollars. I tested locally with LM Studio and liked that pattern - no need to convert model formats.

I tried RunPod first. Unfortunately, their sign-up system bugged out processing my $150 minimum (!) deposit multiple times. Even when I got past that, I couldn’t actually deploy: they had no inventory of any GPUs under a few dollars per hour, and even those wouldn’t spin up correctly. After a day of troubleshooting, I gave up.

Hugging Face Inference Endpoints turned out to be significantly easier. It is more expensive per hour than RunPod’s cheapest plan would have been, but the simplicity trade-off was worth it. After uploading my model, I can start/stop the endpoint on command through their UI, which keeps costs manageable.

Leave a comment

What I learned

Fine-tuning is straightforward now. Even with minimal background in the underlying math, the tooling works and guides are plentiful. With a decent GPU (or Google Colab credits), you can adapt a model to your voice in minutes. It took longer to format my training data than actually train.

Creating a useful RAG pipeline is harder than I thought! You’re constantly fighting the model’s prior - it wants to complete patterns from training on billions of tokens. Getting it to stick to a few pages of writing and a resume without losing the model’s voice entirely is tough.

I think the answer isn’t choosing personality or accuracy. It’s smart, contextual routing: personality for greetings, precision for facts, honesty when confidence is low.

And finally, the result: personality when wanted, precision (with clickable sources!) when required, honesty when neither fits. The model sounds like me and (mostly) tells the truth about me!

Try it on my site at andrei.bio/chat.

“Inventions have to make money.” James Dyson

James Dyson and his eponymous company are a rare successful version of an engineer’s company. The perspective of seeing the world as engineering challenges is often an impetus for cofounders to start a company, but that DNA tends to fade as visions expand and headcount grows.

Seen in this light, Dyson seems to have turned his Buckminster Fuller tinted engineering lens on the world into a successful 21st century corporation. He has continued innovating beyond just products for decades, even after retiring as CEO.1 The company also invests in young talent through its own Dyson Institute, a move that runs counter to industry trends drifting toward seniority and software.

Rather than mere cleverness in one-off invention, the modern sense of engineering 2 emphasizes disciplined problem solving: applying analysis, iteration, and design at scale. It is about structuring decisions so that problems become reliable and productive systems. This is the mindset Dyson has institutionalized, turning an engineer’s lens into a company-wide operating system.

Depth vs. Breadth

Dyson goes deep and broad at the same time. On the depth side, their obsession with miniaturization and airflow powers almost everything they make: vacuums, hair dryers, fans, air purifiers. Iteration on these platforms drives down noise, improves efficiency, and keeps performance moving forward. In their offices sit vintage jet engines that Dyson has refurbished himself. Propulsion and airflow are central to the company’s identity, and that tinkering with the past is encouraged.

On the breadth side, Dyson treats the world as a set of adjacent engineering problems. Airflow research spills into hair styling tools. Farming experiments driven by Dyson’s personal interests end up as hair oil companion products. Battery work from a canceled car program feeds back into home appliances. Its a an engineer’s systems view applied to a 21st-century consumer brand.

Leave a comment

Structure

“I suppose I see it a bit like an engineer… try several things out in an organized manner.” James Dyson

Dyson has kept the company organized around this view as well. Rather than category roadmaps, projects often begin as engineering challenges around tangential applications of core knowledge and technology. It still feels like James Dyson can point in a new direction and mobilize a team to explore it.

The 2019 shift of Dyson’s global HQ to Singapore reinforced this approach. Singapore puts them close to advanced supply chains and strong mechanical engineering talent, making it a great base for both R&D and production. Here’s a walkthrough of the space, it feels just as much a lab as a corporate HQ.

Further, Dyson is an outlier in hiring younger talent with an eye for engineering and training them in their own technical processes. This has even grown to include their own university of sorts. This is smart counter-positioning in a time of ballooning salaries for senior technical talent.

Subscribe now

Closing

At their recent Berlin event, Dyson showed off new forays for the company: robotics and even further advances in miniaturization. A £2.75B investment plan and the Singapore expansion position Dyson less as a British appliance company and more as a global, systems-driven manufacturer. Farming operations still sit largely in the UK, but the corporate center of gravity has shifted.

Dyson is unusual: an engineer’s company that scaled. Its products are expensive, sometimes eccentric (the air-purifying headphones were a definite miss) but the company has carved a profitable path by treating the world as a series of engineering problems to solve.

Initial Inspiration

I wanted to explore different ways people could interact with AI. I’ve been bored of the back and forth (often in chat) interaction patterns that most AI products have. There is often a lot of latency in this back and forth too, which takes me out of the experience - they feel bad to use! This has only gotten worse with better quality - reasoning can take a long time. Instead of directly prompting (which is hard to do well, especially when unexperienced!) - why can’t the AI be more environmental - emerging when I need it? Aren’t LLMs supposed to be smart with context? I wanted to think of AI as something that sits in the margins, attentive but unobtrusive.

Since I started journalling recently - I thought it would be a good place to explore new interaction patterns. I wanted to stay in the flow of writing, but get suggestions on how I could go deeper on the fly. When done, I wanted to better reflect on what I wrote.

Prototypes

Below are screenshots of in progress prototypes. I would demo the prototype to a few friends and bootcamp mates as I got closer to the feeling I envisioned, collect feedback, and iterate from there.

To start, I quickly settled on this double spaced layout for writing. There’s space for inline ‘nudges’ from my system to display cleanly next to the part of the text they relate to, almost like annotations.

My first idea to display ‘themes’ when done writing was a spacial bubble layout. While fun to play with/animate, it wasn’t efficient at displaying what parts of the text related to the grouped themes. It also did not work well on mobile.

I pivoted from the bubble layout. Eventually, I settled on keeping the text the same, and instead just highlighting the parts of the text that were part of themes.

The final layout → no more ‘write’ and ‘analyze’ tabs. Instead, the analysis happens in the background and becomes available to display when done.

How It Works

Nudges (questions inline)

Engine that balances these considerations:

• Characters written and time since last API call

• Punctuation (commas are a treated as a light form of punctuation)

• Topic shifts (light NLP to make sure nudge isn’t stale when response is received)

• Timeouts and debouncer to avoid spamming API calls while maintaining a good nudge cadence

Put simply: nudges work best when they match the rhythm of the writer’s thinking. Just the right speed, not stale, and thoughtful.

When an API call is sent, Refract generates an object with a nudge. Refract asks an LLM to think of a few and pick the best. My initial system was split across two calls and generated great nudges but that way too slow. Combining them into one prompt was the best of both worlds - the tradeoff between speed and quality was a recurring motif. The returned object also includes a confidence score - demo versions have a lower acceptable confidence score for more frequent nudges.

Themes (highlights after writing)

1. Run the entire text 20s before the goal time ends through an embedding model

2. Cluster those embeddings into groups by similarity

3. Send those groups back to an AI API to label them and generate an appropriate highlight color (we know embeddings with similar values are similar, but not *how*)

Themes are less about classification and more about reflection. At their best, nudges sometimes dramatically change the themes written about - or continue to go deeper. Pinning the nudge and seeing the theme colors change or get more intense in saturation helps writers notice that.

Leave a comment

Layout

The hardest part wasn’t getting the AI to generate thoughtful nudges or recognize themes, it was positioning elements in the browser. Text areas weren’t designed for this, so I had to trick the DOM into letting me layer intelligence on top of plain text

To my knowledge, text areas are treated as a single DOM element and cannot be drilled into. Therefore, I made a mirror of what is inside the text area to help me position contextual elements. By tracking the text-area state through a map of sentence IDs, we can make sure the UI element displaying the nudge, highlight, etc. aligns with the text that generated it. Copying whats in text area split by punctuation, and using spans for each sentence with proper dimensions to replicate the layout in a hidden layer, we can find positions using browser APIs. This approach allows Refract to ‘layered on’ UI elements with correct positioning, with a request for animation approach to keep scrolling in as much sync as possible. Some browsers unfortunately (like mobile) do worse with this system and there is a bit of lag sometimes present.

Conclusion

Everyone that has written on Refract has written for longer than their intended goal, often not realizing that they’ve gone long beyond their initial goal! People are generally satisfied and sometimes impressed with the quality of the nudges, and like seeing how their themes changed as they wrote. while balancing latency and quality continues to be a challenge, i’m excited about new ways to bring LLMs - and AI in general - into a more collaborative relationship with people. In undergrad, I read about a concept called ‘soft tech’ where architects were proposing a future for technology in the context of the home, embedded into the architecture itself. Refract is a small attempt at ‘soft AI’: not something that demands attention and surveillance, but something that lives quietly in the background that amplifies human agency and thought. I remain curious and optimistic about a future where AI is built in the context of human creativity, connection, and reflection. Try it yourself at refract.site.

Subscribe now

Selected Links

- New Tiny Awards! Which was your favorite?

- Full Neon Genesis Evangelion - one of my favorite shows - cinematic breakdown and analysis series on YT!

- Small $39k off-grid passive solar home - I like the kitchen :)

- Cartoon inspired Parisian apartment - not my exact style preference, but I respect it, it is very well executed

Leave a comment

GPT5 is pretty good

I have seen a lot of mixed opinions on the quality of GPT-5.

The model family is large. While the available variants vary, Cursor offers a grand total of 8 models to choose from.1

The way ChatGPT behaves with GPT-5 under the hood will likely improve how most people interact with it for most use cases.

Prompt Adherence

It feels much more deterministic in its response. I don’t know if that’s due to temperature nuance, improved prompt adherence, or something else in the architecture, but the model feels like it follows directions for specific tasks much more tightly than before.

For me, this is a huge benefit. While some use cases (like creative writing) will suffer, many others benefit. For example, GPT-5 is now a much better coding agent and thought partner. I trust the model to do what I ask, making it feel like a collaborative partner. This is how I want to use AI, even if it comes at the cost of creativity or perceived style.

User Experience

In the GPT-5 keynote, OpenAI had a whole segment discussing how good the new models are at helping people with their health questions. Sam backed this up in a follow-up podcast with Cleo Abrams, saying that a big chunk of consumer ChatGPT questions are health related. It makes sense that insane pattern matching technologies can help people spot connections they might have missed.

A peer of mine at Fractal has been exploring unexpected connections between people models can find in various ways.2 Part of this exploration involved building a full-blown product, but also simple experiments like matching people based on interests at a small group discussion he hosted.

GPT-5’s interface changes also bring smart contextual provenance and emergent complexity to the chat experience.3 To me, chat-based LLMs are best used as next level search technologies. Having an efficient router model (even if imperfect) to swap between the family of variants is critically important for the experience the majority of users will encounter. This has improved in the new web interface (from imo poor to… let’s say mediocre :] ).

Subscribe now

Road Ahead

Problems remain. A subtle addition is the ability to get answers faster - nice! Unfortunately, ChatGPT simply stops its response when I hit the ‘get answer faster’ button and returns a ‘model not found’ error. It is still complicated to know when to use deep research, turn on web search from the bottom chat bar, give control to app windows on the MacOS app, etc.

I am frankly not even sure how to turn on ‘study mode’ - which I believe is one of the best ways to use LLM chats today. This is especially true when paired with Canvas, which is still criminally under-explored!

People who dunk on GPT-5 for being worse at 5+1 are missing the point. ChatGPT was never a calculator or pure Google replacement. Instead, it is an evolution: a new tool with different, incredible strengths and surprising weaknesses that we must learn to use effectively.

Like the foundational technology that is Web2 social media, LLMs will produce both positive and negative externalities. My goal (and I hope our goal) is to amplify their best aspects and side-effects, not cram them into ill-fitting roles. GPT-5 and its collection of small improvements feels like a step in that direction.

Reggie James’s recent FWBFest presentation makes this distinction clear, framing it through McLuhan with a focus on hardware. Its well worth a quick read, especially if you are unfamiliar with McLuhan’s frameworks on technology.

Leave a comment

Lately I’ve been thinking: do people actually like slop?

If so, why?

I don’t mean “AI slop” specifically. The suspicion that technology flattens narrative weight isn’t new. Socrates distrusted the written word; he thought it would erode memory (sounds familiar). The rhetorically tricky Sophists troubled the Greeks long before ‘rage bait.’ The camera was engineered and painting changed. Theater gave way to film, and now film faces diffusion models and LLM screenplays. Drew Austin wrote about how even Google Maps flattens the world.

However, I don’t buy the narrative that the tools are inherently bad and to blame. The idea that history is just a long slide into dumber and dumber forms doesn’t hold up. Someone reminded me recently: artists are responsible for the art they create. If humans are a species that makes tools, of course we’ll use the tools we create. Everyone wants to press the red button, to pick up the stick and walk it home. The difference is always in the quality of the application.1 This has always been true.

So why do people reach for slop?

Leave a comment

I don’t think slop is inherently bad either. Maybe it’s about numbness, boredom, or just the desire to pass time without feeling. Slop isn’t being force-fed… people willingly engage with it. Meta found that people prefer branded content to posts from their friends and family. McLuhan argued that every new technology amputates some part of us, for better or for worse, and maybe the long march of narrative accessibility has made us less critical toward what we consume. Each new tool makes production more accessible. And with that comes more slop.

But what even is slop anyway? Is it the early work of a new medium? Low effort creation with little intention? Perhaps how we define it says something deeper about us? Judging creation does seem like is a cultural act. What we choose to consume reflects the kind of culture we have, or want to have.

Slop often lacks craft, or taste (perhaps this is another reason this word has counteractively become a term of art in tech, especially for designers.) The Bauhaus rejected the divide between artist and artisan: art and toolmaking are the same discipline. A chair, a typeface, a building, a city street… all should be built with thought and care. Maybe slop is what happens when technology outpaces craft - or the human capacity to thoughtfully apply it to our needs.

When there’s so much narrative (content, stories, noise) out there, it becomes overwhelming. The simple, cognitively easy narrative becomes more appealing by comparison. Byung-Chul Han (BCH) points to this in The Crisis of Narration. Just as Walter Benjamin noticed a shift in how stories were told in his time, BCH observes something similar for ours. For BCH, stories specifically require distance and contemplation… time to process. But what exactly is a story?

BCH distinguishes between information (news, short form social media, clickbait) and stories (especially as it relates to you as a individual). Information and fast narratives often mimic stories, but are really just “soap operas.” In The Crisis of Narration, BCH argues that as we transitioned from an oppressive society to one focused on achievement and self-optimization, individuals began to exploit themselves under the illusion of freedom. There’s a lot more to this, but a crucial piece is the erosion of collective storytelling in a hyper-individualized society. It follows that the sort of ailed people BCH describes in Burnout Society end up living out a kind of internal, perverted story. This is heightened with panopticons when social media is at its worst.

Theories matter too. Theories are like stories, but with some crucial distinctions. They select which parts of the world matter and rearrange them into something meaningful. A good theory can invite someone to see themselves inside a system, or to navigate a complicated world more clearly. It is how internal stories can collectivize into communities.

But when misused, or applied too quickly and without context, information can also feel like a theory. We see this today in reactionary politics: views that seem inflammatory at first often become much more nuanced when thought through carefully. And yet, there are still massive internet mobs emotionally invested in simplified, fast-moving narratives. describes another sort of perverted story that happens today in his recent piece Ideas Are Driving Me Mad.

Leave a comment

Some may ask: in the age of AI, who even needs reflection (or stories)? The AI will reflect for us! BCH would argue this is impossible because of the very nature of AI. He says AI is pure intelligenz (intelligence), but lacks geist. A German term that roughly translates to “spirit,” geist is closer to a conscious mind with a soul, creativity, and moral capacity. But intelligence without geist is just computation.

Today, AI has become a symbol of pure intelligence. So it’s a mistake to outsource reflection. AI can help us spot patterns, really well! It can even reduce the time and distance we need for pondering for our own stories. But it can’t reflect in our place. That kind of meaning-making is still on us.

I believe technology can be utilized, like the tool it is, to help us be better. AI can do great things to help us tell better stories and create healthy theories to help us all live better lives. However, this isn’t easy work, and not the prevailing consensus in technology. Just reference a16z and Cluely for that…

I look to the Bauhaus again for inspiration.2 They believed that bringing art, architecture, and technology together in service of a more livable society. That spirit still matters, and it seems that the sort of ‘geist’ behind this isn’t part of AI today. We then need to hold onto it ourselves in the application of this new tool. Perhaps we can hope to mend previous wrongs as the change is so dramatic.

I want to help build a version of technology that helps us avoid the shortcomings of our current society that thinkers like Byung-Chul Han - and Walter Benjamin before him) have helped us define. Benjamin once said that storytelling is dying because people no longer have experiences worth retelling. I think people still have stories worth telling, oftentimes people just not given the time or space to process them. Tools can’t do that work for us. But they can create space for it if we build them with that intention. We have to ask what kind of lives we actually want and build in that direction.

And if we don’t, then maybe the slop is just us. Or at least, the version of us most of us have chosen to become.

What do you all think? Let me know if you have any sources of optimistic inspiration, especially as it relates to new technology. I am still slowly working towards being able to contribute to this… the bootcamp ends in one month! I hope you are all well - I’m doing well, just a little tired : )

I recently concluded the mid-bootcamp break and am back at it. We are now focused on expanding skills related to our interests, and getting ready for a new career. I am excited to dive deeper into some newfound interests… what I want to do has changed a lot as I’ve learned.

Part of prepping for interviews has meant the construction of a new personal site. It seems to be tradition for me to rebuild it every two years, so I’ve decided to start memorializing my old versions. You can view it here, still a work in progress.

I also have a suggested blog to read, the latest Are.na Editorial piece on footnotes! I love footnotes 🤓.

Week 4: Framer Motion & HTML Canvas

I am glad to have found a balance between velocity and fundamental understanding this week. I am proud of the artifacts I was able to produce by Friday evening while not taking the 'easy way out'. One of said artifacts is this brief algorithm visualization website.

Creating animations I was more complex than I had imagined. I initially thought it would be simple using Framer Motion (having seen examples and showreels online) to slide elements across the DOM at will. Unfortunately, I was unable to setup flowing animations for sorting algorithms all that well at all! I dug into how Framer Motion works under the hood - tl;dr, it seems to be built for animating React components themselves rather than flowing visuals across a web page.

So, I turned to an old friend: HTML Canvas. Having used trusty canvas for a course in undergrad, I was familiar with its verbose yet clear logic to 'draw on the web'. I was able to create 'animation' in a stop motion way, which paired well with 'snapshots' of algorithms as they went through their logic. I was able to create a simple canvas element that scales well to different aspect ratios and resolutions.

However, HTML Canvas doesn't truly allow for moving animation. If I had more time, I would have likely combined HTML canvas with SVG, or perhaps only used SVG. SVG is dynamically scalable in straightforward ways like HTML Canvas, but is also animatable. Many animations on the web that appear to 'draw in' use SVG underneath. I look forward to when I can combine HTML Canvas and SVG in the future for even more interactivity with React!

Week 5: Barbells Revisited

Why everything feels extreme but isn’t

Week 5 became an uncomfortable mirror. I’m quick to tell others to “trust the process”, but the moment my own work plateaus, I start to beat myself up. I’m hardly alone. Dennis (founder of Amie) poured himself into software he loved; when the launch metrics sagged, he re-skinned it as a generic productivity SaaS in search of faster growth. How do you learn at your own pace and still build things of Quality? True exploration is slow, deliberate. It requires what Rich Hickey calls “hammock time” – space for the mind to wander, ruminate, and recombine.

The trap of binary thinking

Peter Norvig’s classic Teach Yourself Programming in Ten Years reminds us mastery is measured in decades. Yet entrepreneurship keeps chanting winner-takes-all mantras: “30-day bootcamps,” “no-code millionaires,” “Build your MVP with AI!” We fetishize lifelong craft while secretly pining for a month-long hack to success.

Revisiting barbell theory

It therefore follows that tech is increasingly reliant on barbell theory as the stakes seemingly get higher. It is easy to adopt this framework wholeheartedly when AI is the driving force behind much of innovation. The country that gets AGI first dominates the next century, maybe the rest of humanity. Tina He’s recent exploration, The Illusions of Barbell Theory, along with my old piece, Barbells Everywhere, question the validity of this claim. Tina does a great job of illuminating the distinctions between perception and reality of barbell theory, driving home the idea that genuine work demands time... time that, in a world hungry for quick results, is often sacrificed. Many ambitious young people fall into this trap: being too ambitious is a clever form of self-sabotage indeed. With today’s technology and access to information, it takes surprisingly little time to find the edges of known knowledge. From there, the hard work of step-by-step exploration begins, like tiny slime molds branching out on a map of knowledge still shrouded in a fog of war. The messy truth is that most meaningful work occurs in the neglected middle distance between these extremes, probing the edges.

A prime example of dedication to the craft is Robert Caro, who has spent 45+ years writing about two individuals. His research is meticulous: he would travel and interview staffers, their spouses, friends, skim microfilm newspaper archives in small county offices in rural Texas... all for one small quote or headline to drive home a character trait he's trying to make. His approach underscores the value of deep and sustained effort. Meaningful work requires time and dedication, even if it is not immediately apparent.

There are, of course, alternatives

Take Bryce Roberts and Indie.vc. By treating profitability as the starting constraint, not an eventual pivot, Bryce shows that startups can grow on their own. No “go-big-or-bust” barbell, just steady compounding: keep more of the cap table, choose when—or if—to raise again, even share cash dividends along the way. Same upside, less whiplash, even for the investors.

Week 6: Tensions

Reflection

Our project this week was to demo a game by Saturday morning. Working in groups for the first time, we needed to decide together a) what we wanted to build and b) how to go about doing that. While we continued refining our knowledge of previously learned technologies, we’d also needed to learn how to communicate as a group.

I was surprised to learn how unsolved of a problem collaborative development is. We quickly tried to establish our own patterns, which did not work very well at first! However, each day got better and we started to roll. By the end of the week, I began to understand the shortcomings and limits of existing practices. So much is just convention! It is surprising so much DevEx (at least, what we experienced in this week) felt like walking through honey. Rao’s concept of ‘tensions’ came to mind. A quick definition c/o o3:

The basic units of sense-making aren’t “problems” or “trade-offs,” but live, opposing forces—tensions. Think of a tension as two incompatible, yet mutually sustaining poles (eg. freedom ↔ security, speed ↔ quality).

Anyways, I also recently watched the second season of Nathan Fielder’s The Rehearsal. It deeply investigates the relationship between first officers and pilots of commercial airplanes. He posits that one of the leading unsolved reasons for crashes and accidents is poor communication practices in the cockpit. In many ways, this is also true for software development - with much lower stakes of course. I recommend a watch for anyone working with others in positions of power in the workplace!

I started the bootcamp I discussed in my last post earlier this month. Every Saturday, we have time to sit and reflect on the week. I’ve been posting the writing on a website that will eventually turn into my portfolio. I wanted to share the technologies I’ve learned so far, as well as general reflections from each week.

Technologies Learned:

• React single page app: a basic game (I picked Tic Tac Toe)

• Styling: TailwindCSS, making pixel-perfect components from Figma designs

• Local server: made my game run locally using vite-express

• Database: app references a Postgresql database of games hosted on Supabase

• Websockets: auto updating game state for live online multiplayer games

• Minimax algorithm: Best next move (minimax algo) autocomplete button built into game

• Deployment: backend via Dockerfile to Render, frontend via Vercel.

• NextJS (with App Router)

• NextJS AI SDK (tool calling, text generation and streaming, object generation, generative UI, etc.)

• ElysiaJS

• OpenAI & Google Gemini APIs

• Better Auth

• Sentry

• Vercel Analytics

• Framer Motion

Reflections

Decisions - Week 1

One is constantly making decisions when programming. From the outside working with engineers, this seemed obvious: of course, from my perspective as a Product Manager, an engineer would need to pick the right tools for the job! But it turns out that's just the starting point. At every step after that, a programmer is constantly making decisions. How should my app be structured? How should this component manage state? Oh, this function needs to do more logic work... maybe it should be split into its own function. Since I'm writing in Typescript, decisions about types are constantly swirling around my mind! Referencing Grant Slatton's 'Algorithms we develop software by' it turns out there are many turns on the path to the solution.

At the end of this first week, we were tasked with creating a personal blog NextJS website to reflect on for the rest of the program. I asked why we should maintain our infrastructure instead of just using Substack. Andrew (the lead instructor and co-founder of the bootcamp and space it runs out of) had a simple answer I have heard before: When you build software you start to see other programmers' decisions everywhere. Why is this button all the way over here? Why can I not change this thing? Sometimes it is better to just make your own decisions. Turns out we are also going to be using NextJS next week, so it is a nice easy way to get familiar with the technology 🤓 Here is the link, I'll work on styling it nicer, and will eventually be synced with this blog and part of my personal site.

I have previously learned to think this way before: in high school, when I was learning programming and robotics. Sadly, I have since forgotten this lesson and largely abandoned this very sequential way of thinking. I think re-learning this perspective has helped me better understand the engineers in my life. When every small action is a decision with side effects, it follows that those decisions would be reasoned and deliberate. This perspective can be applied to much outside of programming.

Leave a comment

English as Abstraction - Week 2

Someone asked about the differences between systems software engineers and architecture software engineers, since the names could imply (referencing the physical world) that the roles have similar scopes. Eugene, one of our co-instructors, answered with: "They have no overlap." He went on to explain that a lot of software engineering job terms - like 'Staff' - but also in the languages themselves - like 'interface' in TypeScript - are generally arbitrarily chosen. This is evident everywhere in languages I have come across. Abstract concepts are attempted to be represented by physical terms in as familiar a way as possible, but so are moves in the code. For example, a function in TypeScript can be defined as functionName = () => logic. The => indicates the function starts its work after the arrow.

It is therefore no surprise to me that software engineers have a bend towards literature and philosophy (especially metaphysics). I have read about how programming languages are written in English for humans to better understand how the electrons are moving on the hardware, and Andrew (our main instructor) repeated this often in our first week as well. However, coming across so many different web frameworks and technologies these last two weeks, I have a new broader view at just how wide this horizon is.

Product Like an AI Engineer - Week 3

Product Work (as a software engineer)

Working AI products cleanly into existing product flows was a challenge technologically - at least with my last-gen minded design brain. For example, I wanted someone to define a focus for their practice session before I brought them to the actual LLM chat interface. Ideally, this focus would be pre-populated with the coach's practice plan as a response, and formatted to look like a special UI element, since it is a practice plan and not a normal AI message! Getting this to work with my still growing understand of the Vercel AI SDK was a challenge.

First, I learned I needed to reload the page to get the tool call responsible for generating the practice session to recognize a message had been inserted with the user's intended focus. To get this reload to happen without other side effects in a NextJS React app was difficult -- just reloading simply would cause user messages to send twice, or appear twice in the database, or cause the AI to re-send the practice plan inside of a normal message.

Still a beginner, I was forced to dive deeper into the inner workings of React's render logic and came out a better engineer for it. The assigned readings of the week also paired nicely with the 'on-the-job' lessons. Vercel's own Common Mistakes with the NextJS App Router helped dispel misunderstandings, as did Dan Abramov's One Roundtrip Per Navigation. Another theme of the week was to learn more about motion in web apps. Nanda Syahrasyad's posts about how Framer Motion and SVG work in this regard were excellently laid out! Eventually, I'd like to dive deep into how shadows help create depth on screen interfaces - this is something I do not understand very well yet!

Career (or what I could see myself spending 20 years pursuing)

I took a walk with Andrew to discuss my future career, where he essentially challenged me to think bigger about the scope of what work I'm contributing to.1 After a few days of thinking, I've come up with three themes I could see myself working around for 20 years, which also align to my skills:

1. Old Steve Jobs computers as 'bicycle for the mind' type shit

   • Generally, I think that the best technology either helps amputate generally tedious activities (dishwashers, laundry machines, etc.) or extends our reach (access to information like search engines or online media, better payments, etc.).2 I'm drawn to where computation, interface, and new intelligences collide: AR/VR as new HCI, AI as interface architect, etc.

   • Some companies/prototypes in this domain that I find interesting:

     1. Exasearch built ground up for AI

     2. Flora fresh interface for creatives using new AI capabilities

     3. From DeepMind - Gemini Flash 2.5 Lite creating UIs on the fly that are context aware

     4. Notes app that 'builds itself'

     5. Mainframe (Jordan Singer's latest company, built Orchestra and Cobot)

     6. Sublime new take on categorizing the internet

     7. Daylight Computer hyper-specific purpose computing device

     8. Bevel quantified self, built on Apple Watch (essentially Whoop competitor)

2. Living in harmony with nature

   • This includes access to abundant clean energy, ways to access real wilderness, architecting the built environment to be in harmony with nature. There are many ways technology can help humanity and nature coexist better.

3. Engaging and shaping your community

   • I think technology can help engage people to their surroundings more, and by extension get them to be more engaged in shaping it. Why can you not interact with your federal government, city, or neighborhood, or street better through technology? I think there should be software that empowers people to shape their communities. For example, some social media (Substack) can help proliferate ideas about what your community should be like. Companies like Kaizen Labs is helping build simple websites for municipalities. There are many ways technology can help in this goal.

When I made my website a few years ago, I wrote: "I'm interested in how technology can help people directly, especially as it relates to expression and the physical world." Ironically, I copy and pasted that paragraph directly when starting this blog; it still feels true. I am excited to take a concrete first step toward these larger ambitions.

Leave a comment

Growing up, I made things. I loved (and still love) to play, cook, build Legos, make computers, paper planes, help grow clubs at school… all sorts of things. Making something new with friends are many of my happiest memories. I have always been inspired by people who have agency to grow and/or create their own. When I was younger, I was surrounded with such people: my high school was filled with very gifted students who all seemed to excel at their own thing, which was an inspiring environment for me to grow in.2 However, not having the same resources as many students at the school, I was insecure about my own talents and skills. I did not have access to the same level of tutoring, test prep, parental homework help, or general professional connections that they did – I felt this acutely in technical classes like math. I thought I was smart, but nowhere near as bright as my peers. As college admissions demanded I chart my future, the diverging trajectories were obvious in my mental graph of ‘progress.’3 My parents’ immigrant scarcity mindset was a constant shadow in my childhood, quietly boxing out my own worldview. My lack of self confidence began to grow into anxieties: I had to make sure I could stand on my own two feet after graduation. The thought of affording New York City felt intimidating.

Still, chasing my curiosities meant I knew what I wanted to study in undergrad. I planned to build my own major around making things: equal parts humanities, engineering, history, and finance. However, at NYU I was exposed to classmates with a completely new level of elite wealth and connections. I shied away from the actual making of things and shifted my attention towards the scaffolding around making as a refuge. Software engineering had always been a means to an end, a tool, right? Understanding the why and how things came to be became my new obsession. Why did some venture capital funds have such strong world views? Why did certain ideas survive over the decades, when others boomed and then fizzled out? What policies and geopolitical currents helped create the tech industry of today? I read histories of Silicon Valley, gorged on all the takes on current tech news, debated concepts like effective altruism, e/acc, and abundance. The more I learned, the more I could begin to see the layers beneath every new trend and ideology. These themes had been present in my youth (both at school and outside the classroom)4, but I learned of their deep histories and broadening implications in college.

Working throughout college to make money, I had a buffet of amuse-bouche bites across tech: VC, startups, S&P 500 firms. After graduating, I ended up in a seemingly ‘safe’ job at a legacy media company, working on streaming technology products. After a year of shifting priorities every quarter, management changes, and team layoffs, I felt like I had learned little and had no path towards growth. I started to look elsewhere. Having a hard time with so little experience in a bad job market, I ended up in a similar legacy media company, building internal tools instead. I hoped a change of scenery would help me continue learning and rekindle my excitement about my chosen career. Unfortunately, in many ways it ended up being worse… I quickly realized a big driver of my professional ennui was a lack of agency.5 Spending my time on what I think is important (learning new skills, directly helping people with their lives, etc.) was replaced by status games and constant start/stops from shifting leadership priorities. While I attempted to suggest improvements for our work structure and procedures, there wasn’t enough will to change anything.6 Outside of work, I was unable to spend my time as flexibly as I would like. Working towards embellished job titles, pay raises solely based on seniority, and finding comfort in job security because I am the only one who knows how something is (inefficiently) done is not the sort of career I am interested in building for the rest of my life.

So, I gave two weeks notice.7 I want to return to making things that help people with those that inspire me and have more skin in the game. My professional sehnsucht is one driver, but there are many more contributing to this decision. For one, I have read a lot about the tech industry and its history over the last six years. I have very much enjoyed learning the history deeply related to ongoing events, but after a certain point simply reading about things was not enough.

Tech has always had many different political influences and aspirational futures to chase. Camps grow and shrink, utopian ideals come and go, but the built technology remains. Seeing tech go through a dramatic shift from when I was in high school, I am feeling impelled to build towards the futures I want – to build opinionated products. General tech 'accelerationism' is an overdue correction to years of growing doom and gloom in Silicon Valley and the West (the piece itself itself surveys a host of canon startup culture pieces by thinkers like Samo Burja, Tyler Cowen, Patrick Collison, and Marc Andreessen) even though so much software has been made.8 Nonetheless, I believe the world still needs so much more software, the possibilities seem endless. AI is only going to supercharge these tailwinds. While much of the tech whiplash of the last two years has been disappointing to see, I’ve also found inspiration around me in NYC and online. Here are a few highlights:

• Venkatesh Rao’s writing around Protocols

• Demo night culture

  • South Park Commons - ‘Request for curiosity’

  • Abundance Politics & NYC - Fireside Chat & Demo Night

• Quotes of advice to young people

  • quotes to 20 y.o. Self

  • Letters to a Young Technologist

• Tech community builders/creatives

  • Reggie James is a writer and NYC tech community builder, has brought together different communities with tech too, like fashion, etc.

  • Conor Davidson helps host creative code events at Index

  • Socratica

  • Esmaralda

  • RadialXChange (their work with Audrey Tang specifically)

• People who have taken their own first steps

  • Weber Wong started a startup incubated at a creative code graduate program after being a banker and VC

  • Jasmine Sun left a growing startup to be an independent writer (journalist?) and is producing inspiring content left and right. Ironically just released a podcast on agency.

  • Dorothy Ren is now a creative technologist @ Every.to after graduating a bootcamp, was previously a PM and VC

  • Sean X has a high degree of sovereignty over what he works on, typically involving frontier tech and hardware - I’d like to return to working on some hardware projects!

  • Anjan Katta and his search for a different paradigm of technology with Daylight Computer

Taking all of this into account - the changing winds of mainstream tech culture, my ennui with corporate America, my desire to work with people I’m inspired by - have pushed me into a different direction. I quit my job to partake in an intense, three month long bootcamp to help round out some skills I’m missing to hopefully be a ‘product engineer’ or ‘design engineer.’ The pull to start a coffee shop or become a geologist park ranger has certainly flickered across my mind - but, if I’m being honest with myself, those won’t solve my frustrations. Applying the ikigai framework to myself, it makes sense to move the focus of my work while still utilizing the skills I’ve spent the last decade building. I believe this new direction can use my previous experience as a strength, and still leaves the opportunity for more flexibility (freedom, agency) and choice in the future!9 So, after two weeks of travel to London and Copenhagen, I will start learning six days a week at bootcamp as a first step in a new direction :). Cheers to first steps.

Leave a comment

Hey everyone, short updates before the main story for this post:

• Happy new year! My word for the year (instead of resolutions) is "agency.”

• Next Sfânt popup will be February 8th, mark your calendars.

• I am looking to play tennis in NYC during the winter months, lmk if you are willing to split court rentals.

• Quick takes on odds and ends at the end of the post!

• With that…

Barbells

Generally, I’m a pretty big hater of dichotomies, opposing poles, sweeping this or that statements, etc. These arguments are almost always reductionist, and I believe gradients as an alternative is a much healthier way to view things. However, I have increasingly noticed opposing extremes as a phenomenon in all sorts of niches, which I will call barbells (two weights on opposing ends).1 I believe far apart nodes–barbells–are increasingly becoming a fixture of modern society across culture (especially online) and in finance (VC, business, etc.).

Similar theories in the wild (Twitter, IG):

Media

• Digital Media

  • YouTube micro influencers ←→ Mega influencers (Mr. Beast, Like Nastya, Mark Rober)

  • Self publishing2 ←→ traditional publishing

• Traditional Media3

  • Streaming: Britbox/Crunchyroll ←→ Netflix (or other Hollywood accountant fueled streamer like Disney+)

  • Blockbuster actors ←→ generic cast members

• Social and trad media will only be amplified in this regard with AI content, on both ends of the barbell.

• Some more thoughts on this one: Purely economically, it makes more sense for many brands to distribute their advertising money (which ultimately drives these digital algos, shoutout Filterworld by Chayka) over higher engagement micro influencers for smaller CPMs than a huge influencer account with seemingly more reach. However, it is inevitable that there will be huge accounts that continue to draw revenue, this is especially true for other forms of media like streaming.

Sports

• Running loads of zone 2 ←→ peaking on race day

• Pro athlete salaries, especially true in uncapped sports (tennis, baseball), but true across the board (soccer, NBA)

• Biohackers/life-extenders/training over-optimizers ←→ live in the moment, ability and mobility focused fitness

Health

• Telemedicine ←→ concierge IRL medicine

• Preventive naturopathic care ←→ cutting-edge tech driven treatments

• General practitioners ←→ specialist consultants

• Biohackers ←→ no fluoride non-GMO

VC

• Niche funds (hustle fund, themed funds like Indie) ←→ PE adjacent mega funds (a16z, Tiger Global, NEA)

• 10x engineer for startups4 ←→ ‘normal’ employees of a company (esp. when taking equity into account)

• Unicorns/fund returners ←→ no exit

Loosely, ‘culture’ (bit of a catch all)

• Casual interests ←→ lifelong pursuits

• Massive Open Online Courses (MOOCs) and self-learning ←→ elite universities/exclusive courses with ridiculously high price tags

• Travel to urban centers ←→ rural retreats (staycations as well?)

• Solo travel vs. group travel (either friends or unknown)

• Budget airlines ←→ luxury travel

• Fast fashion ←→ luxury brands

• E-commerce giants ←→ niche boutiques (connections to social media)

• Zero-waste lifestyles ←→ high-consumption (maxx) culture

• Simple 1.5x credit cards ←→ status/exclusive events/airport lounge credit cards

Conclusion on Opposites

Robert Cialdini’s concept of sold opposites: ‘everyone has one, so must I’ or ‘not many people have this, it is desirable.’ Another example: Imagine a completely automated, speedy check-in process entirely on your phone. Now imagine a personal check-in process with many staff touchpoints, porters carrying your bags. There are many great ways to check into a hotel, and they can be total opposites.

Many of these relate to compensation - perhaps this could be explained by applying the pareto principle to returns (most of the returns/views/productivity etc. from a small handful of the provided resources)

Leave a comment

Is this a human trait…?

Re: Romania rn

Images above from Beauty and Biodiversity in Romania’s Hayfields - Atmos, which discusses the ecological benefits of age-old forms of haymaking.

In my opinion, there is a big culture war going on in Romania re: industrialization and rural flight to the city over the last 20 years. Much of Romanian culture stems from rural/distributed regional traditions that are very pagan/land linked in origin… the loss of this (which was in part emphasized secularly during the Communist regime) has led many older voters to support nationalist candidates… but the times have changed. It is interesting this piece gets published now, from a very different political direction.

There is no return to this form of society while maintaining the same pace of economic growth the country has experienced over the last 15 years. The most recent presidential election (and its accompanying drama, worth reading about if you are unaware) is just a flash point of these building tensions. There is a definite segment of the nation nostalgic for bygone times and nostalgia, times that are incompatible with today’s economic and political landscape, and even with the freedoms they now enjoy… unfortunately, I see this in my family as well. This has been written about by some of the academic diaspora as well: shoutout the ‘Future of: Nostalgia’ edition of Kajet.

While I naturally an inclined to support biodiversity supporting systems of farming, to maintain these farms en-masse is aspirational given Romania’s status and relationship to the EU. Even though the land holds so much culture, and these practices are ecologically very beneficial, it is the unfortunate reality that their preservation has largely been co-opted by corrupt populists with other motives. The economic growth that has helped lift the country from decades of decline (90s-mid 10s) is, in no small part, fueled by foreign investment transforming agriculture practices (enriching both the foreign investors and local elites). While purchase power, productivity, etc. have grown nationally at record amounts, the government remains corrupt in opaque ways, leading to undirected anger in those who do not see their conditions improving. This is a large reason there is growing populist fervor for candidates who would continue to harm the population that hasn’t gained from Romania’s economic growth… a familiar story…

I am not sure how to change this system to redistribute the wealth better, especially considering the tradition of communal land working and sustenance…

Subscribe now

• The Sound of Love - tbh, I often get emotional reading YouTube comments under songs, this is a cute way to further that experience

• An interactive essay: Worlding - Future Observatory Journal

• Interesting seeing Nabeel’s writing be proliferated now, I liked his piece on Palantir: Principles - Nabeel S. Qureshi

• Re: internet filtering, here’s Sean’s take on how AI will change human memory: Unbundling memory - by Sean Thielen-Esparza - Language

I hope you are all well and able to spend the winter holidays with loved ones. I was able to make it home for just over a week after thinking I wouldn’t be able to - the last few days have been busy with baking and cooking!

Context

Now that the food has been made, I have some time to sit and write about ‘filtering tools,’ which was the winning topic in my last post’s poll. Being a tool nerd, very much online, and sort of in the HCI field, I have long looked for specific tools that help make the internet a more manageable place. I am always curious to see new ways we can interact with all the digital information the internet provides. This is especially important when I find something I want to keep.

One of the reasons I write these posts is to help clarify my writing - this is not a novel concept1. The Greeks and Romans had a culture of note taking and commentary (rough translations) - hypomnemata - which I find inspiring even though I haven’t been able to successfully implement it (and think digital tools to do so may be a sham…). Knowledge management, internet filtering, and productivity software has become increasingly large trend over the last 5-7 years, peaking during the pandemic. In my opinion, the amount of content, lockdown leading to more creative production, tools reaching a level of feature maturity, and influencers promoting them all resulted in a big circular bubble.

Further, people looking for (somewhat niche) tools to categorize the internet will probably have attachments to the things they are collecting, and the formats they are collected in. It can be a very personal process, especially when it takes so long to set up some of these tools! Because of this, and related popular themes of productivity, knowledge, and process automation, there can also be a lot of virtue signaling.

Ads for ‘my mind’ and Perplexity (click in to see the full images):

For example, these themes are referenced in ‘multi layered calendars.’ Amie recently attempted this, waxing poetic at the founder’s process in making the software which generated substantial funding and Twitter hype. It ended up taking years to be ready for release, and was executed poorly (half-baked features, poor performance, etc.). The next one bubbling up seems to be Joi. I believe this is pretty common in software dev, but especially so for more personal software like internet filtering tools.

Unfortunately, I think much of the bubble was grown by people chasing aspirational goals that won’t be found in digital tools, or at least not yet. With the technology of today, it is often more work to organize these typically cumbersome tools to yield results that will benefit the creator.2 There is a high barrier of entry for most tools, which I was never committed to scale (and I believe many believe they scaled it, happy to do the busy work of building part of the system, without actually completing it enough to make it useful). This may change with AI making tagging - a typically crucial part of building these systems - automatic, but that is yet to be seen. Some of the newer entrants are starting ‘from the ground up’ with AI technologies built into these sorts of tools, and I am optimistic they will be more than a marginal improvement on the previous batch of products.

With that, here’s everything I’ve used that I think is noteworthy. A crown attached to the icon indicates that I still use it (sometimes begrudgingly).

Reading

Readwise is one of my favorite pieces of software out there. It works excellently, has unique features, and aggregates everything I can throw at it. I have been following Readwise’s development since the team started working on it and can vouch that they are extremely attentive and active.

Writing

Google Docs is the simplest cross platform piece of software I have found. I would prefer to use others with built in assistants and less features to streamline the process, but they either cost too much or are too complicated to use on multiple devices. Further, GDocs seems to have the best copy and paste experience into Substack :).

General Notes/Knowledge

I think these tools aren’t worth the effort necessary to upkeep the system or get started. They are also sometimes oriented to a more daily task management replacement workflow that I don’t need. There are so many sorts of these tools and thousands of ‘getting started’ or ‘how I use __’ or ‘__ hacks and systems’ videos for these them.

Web Annotation

Also shouting out Greenwich, which shows a little tunnel marker next to another user’s annotation - leads to fun little internet finds! Curious seems to be popular among Waterloo kids, and I believe my high school helped develop Hypothesi.is???

Browsers

By far my favorite category! Browsers dictate much of how we experience and navigate the internet and are made up of so many subtle yet crucial interactions. Urbit, Beaker (rip) and Mainframe are more nontraditional, especially Urbit. Keep an eye out for Mainframe as Jordan Singer (one of my favorite design engineers) and co. release more information about it in 2025.

What I’ve settled on

For now, I’m using Arc, Readwise, and Sublime. I like how Readwise aggregates all sorts of sources and lets me highlight quickly without limits (I highlight as I read just to make it more active), and the team is quick to add new features. I use Sublime in large part because I like the team and ethos behind it - it is still a young product, but I find its related card feature useful because of the sort of people that use Sublime :).3 The team behind Arc tends to lean aspirational these days, but I think their initial product is solid after a few years of use - I can’t really go back to more normal browsers.

If you’re curious, check out my public collections here, and check out Sari’s - Sublime’s founder - writing and other work!4

I don’t have any strong opinions on what will come, but I am very eager to see what comes next.

Quick interesting tidbits

1. Every is experiment with extendable articles - pretty cool, it can help you navigate the piece, reference the sources, etc.

2. Finally got around to the 11 Labs app, tried adding an article into their ‘make it a podcast’ feature and the whole UX was unexpectedly pleasant! I particularly liked the music/experience when waiting for the podcast to be generated.

3. I didn’t include journaling in the roundup above as I am new to these tools but wanted to shoutout Stoic (really excellent vision and pursuit of it), Gravity (Malhar’s early WIP but really sleek to use), and How We Feel (nonprofit dev’d, simple, ez to share with friends)!!!

That’s all everyone - enjoy the rest of 2024! Almost a quarter way through the century…

- Andrei

Hi everyone, hope you are all doing well and found ways to reflect with the falling leaves. I read this commencement address and wanted to share it:

“I hope you live without the need to dominate, and without the need to be dominated. I hope you are never victims, but I hope you have no power over other people. And when you fail, and are defeated, and in pain, and in the dark, then I hope you will remember that darkness is your country, where you live, where no wars are fought and no wars are won, but where the future is. Our roots are in the dark; the earth is our country. Why did we look up for blessing — instead of around, and down? What hope we have lies there. Not in the sky full of orbiting spy-eyes and weaponry, but in the earth we have looked down upon. Not from above, but from below. Not in the light that blinds, but in the dark that nourishes, where human beings grow human souls.”

Ursula K. Le Guin

excerpt from A Left-Handed Commencement Address, Mills College, 1983

I’ve been doing alright – I still feel unchallenged and unfulfilled at work but have been sincerely trying to at least snag one interesting side project (about 10% of my time…). Unfortunately, the two best descriptors I have for my new work environment are still ‘hostility’ and ‘bureaucracy’… not all is doom and gloom, I have been enjoying playing tennis and running this fall! I made my tennis ladder’s playoffs and ran the Bronx 10 Miler in 1:18!

In a recent trip to visit my friend Connor, I found a book by Le Corbusier I’d never seen before, and I was struck by its ending. It’s a little long to paste here as a quote and includes an accompanying image - find it in this footnote if you’d like to read it.1 It reminded me of the ‘Classic’ and ‘Romantic’ divide (or lack thereof) that has long been a theme in my life. If this concept is new to you, I’ve written about it previously here:

This contradiction has come up again as I revisit what I think I should spend my time doing. Looking back at my childhood with this lens, I have always been drawn to both sorts of applications (Romantic photography and Classical mousetrap vehicles for competition), but I’ve realized this year that I had many other Romantic (design focused?) inclinations that were honestly stunted or ignored. My mom tells me stories of walks she’d take me on when I was a toddler - I would name out every car brand because I noticed their design differences. I also liked going to the auto show with my dad to see new model designs in middle school, even though I’m not much of a car person. I just thought it was interesting to experience. I didn’t go to many summer camps when I was little, but creative cooking summer camps were always the ones I’d pick. Today, I love experimenting with cooking as a creative escape from my day to day, even exploring coffee tasting menus with Sfânt. There are many such instances from my childhood that I frustratingly did not realize the significance of until recently.

The host and guest on this podcast (guest is Matt from Union Square Ventures, their new ‘AI Tools Tinkerer’) discussed the ‘kid who takes things apart’ archetype. While I don’t think I literally took thing apart like many famous people (builders) seemingly did, but I was always very interested to understand how things worked. I never took shallow explanations as satisfactory, I really wanted to get into the weeds about why things are they way they are - in engineering and many other contexts.

Ironically, these realizations come at a time when I am considering changing to a more technical career - and I’ve frankly never had the confidence to pursue my Classical inclinations fully. One where I could have the potential agency to include my Romantic expression in time, but that would be its own journey. What sort of work can support my ambitions while also allowing me to feel fulfilled with how I spend most of my time? I’m not sure how to resolve this uncertainty without just trying something else, and soon.

If you have any advice, you know how to reach me :). With that, I’ll continue to the rest of this slightly longer post than usual. First off, a quick poll on which of these ideas to flesh out first:

Reflections from the Rocky Mountains

I’ve been yearning to travel after having less days off this year and managed to book a long-weekend trip to visit my college friend, Connor, in Boulder, Colorado (around an hour towards the Rockies from Denver). Here are some notes–both surprising differences, interesting similarities–on how I think Boulder compares to the environments I have grown up in, and how it relates to trends on my mind.

The focus on sustainability, and a near-spirituality linked to the natural beauty of the outdoors, reminded me of the PNW. Notably, there was a distinct New Mexican influence compared to the PNW, but it was more similar to what I’m used to than not. This is something (while occasionally too much for me) I miss about the PNW while living in the NE. It is a curious counterculture to the history of atomic, mining, and oil research the region is known for–UC Boulder has a list of physics Nobel laureates, and a ginormous refinery is the only real feature of the drive from Denver International Airport to Boulder. I learned that one of the world’s atomic clocks is in Boulder, maintained by the Department of Commerce! 

Everyone seems to have some sort of connection to the outdoors. There were many hikers, bikers of all sorts, runners, and people just getting outside with their families. I can see why so many professional endurance athletes are based out of Boulder: its peaceful, the mountains are meditative and ever-present, the altitude helps, and a host of supporting talent: equipment shops, performance science, nutrition, and physical therapists. I can’t imagine what it’s like in peak summer or winter sports seasons.

It was sad hearing a Lyft driver say it had rained twice since March… and hearing stories of wildfires entering residential neighborhoods from Connor as we drove around one day - Oh yeah, this neighborhood is all new high-end developments where older family homes used to be, before they got burned down and a pittance in insurance money. Those families were displaced. This wasn’t surprising given the region’s housing policies; Boulder city, until earlier this year, was legally only able to increase its housing supply by 1% annually. Our Lyft driver lived about a 25-minute drive away, in another municipality entirely.

Besides the stunning peaks rising from the plains, I was struck by the sense of ‘small town America’ that I so often hear about but still don’t truly understand. The University of Colorado’s main campus is in Boulder, and it was parent weekend when we visited. The cool parents joined their kids at frat parties and bar outings before Coach Prime commanded his Buffs to an easy victory. This big state school energy is very different from my preppy yet hippy pre-college experience. It is also remarkably unlike the careerist culture I found myself embedded in at university (the generally wealthy students floating without purpose aside).

• Old Granola vs. New Hippie 

  • Rayon Vert ← cool outdoors/biking/fashion brand with a special founding story, this feels very old Boulder to me. 

  • I thought this Freitag bag that’s fully recyclable is also cool, although the price and branding make it feel newer hippie to me.

  • The amount of camo skins entering the cultural zeitgeist is insane right now (esp. in places like Brooklyn and I guess by extension Boulder). This feels very ‘New Hippie.’ Last weekend, Lily and I went to study a few coffee shops in East Williamsburg and saw six men wearing camo hats with orange embroidery/text in the span of two hours. I think the trend is so omnipresent because of everything that’s going on in the US right now… even brands like Topo Design and Klean Kanteen are coming out with camo merchandise/accessories. I saw this Horses2 (love the channel) merch item today, and the Brat mixed with camo trend is … its just gone too far, ok? xD

In short, old Boulder (and in many ways the surrounding towns of Lafayette, Louisville, etc. since there has been an economic outward push from Boulder) felt like a place where people care about the place they live in…

However, there was a slightly melancholic air to our time there. Speaking with Connor’s family about how the area’s communities had changed over the decades had an air of nostalgia. For whatever reason, the feeling reminded me of the ending tone in Hunter Thompson’s Fear and Loathing in Las Vegas. Not directly, but the sense that the promise of previous waves had broken and washed back to sea…

Speaking of that, I’ll end with this talk Cabel (Panic Inc. CEO) gave. The talk centers around a piece of McDonalds wall art, I was really moved by the end:

The artist involved really seemed to care about his work 🙂. The best parts of Boulder and its surrounding culture had this energy.

Aside on tech

• Mark Zuckerberg’s rebrand is seemingly complete - he has stopped appearing in the media for now after 6 months of nonstop involvement. He even appeared in the flesh at the Acquired live event, was exclusively profiled by Bloomberg, and also helped create a 6.5-hour Meta history podcast. Some think he’s right on smart glasses/open-source AI frameworks and wants the people to know I guess, but I think this speaks to a more triumphant return of ‘Great Men’ in Western culture…

• The ‘Bitter Lesson’ - Moore’s Law just makes brute forcing technologies is sometimes the best approach (as we are seeing with LLMs)  

• Importance of timing 

• Apple released some of their old WWDC talks. They shed a light on some of their design philosophies more clearly. I was surprised by how contemporary they still feel.  WWDC18: The Qualities of Great Design | Apple

• Every.to, an online tech media company, is bundling AI writing and content production tools into its subscription. (For example, Make It Sparkle) 

Fun internet and art

• Just thought this guy building a fully DIY digital camera from scratch was pretty cool.

• Sound art :D 

• Minitap, a sequencer browser extension 

• So many creative web conferences are either advertising on traditional digital-media outlets like IG, etc. or on secluded blogs, hard to find are.na channels, etc. Further barbell-ing of the internet (dark forest in some ways), which I have many thoughts about. 

• Low shutter speed photos have become rather trendy since about June, especially in sports photography. I took a photo class in college, and my final project was about movement in NYC (and used long shutter speeds to play with that theme ;) )

I’ll leave you all with that!

Take care everyone - Andrei

2

Highlights

☕ I read the most in the morning

💌 I subscribed to 18 new Substacks

❤️ I liked 75 posts

📜 I scrolled 2 meters in Notes

Top Substacks

Lenny's Newsletter by Lenny Rachitsky

A weekly advice column about building product, driving growth, and accelerating your career.

Top post this summer: Time management techniques that actually work

Kneeling Bus by Drew Austin

Notes from the collision of meatspace and the internet.

Top post this summer: Microtrends at the End of the World

Blackbird Spyplane by Blackbird Spyplane

Your No. 1 source for “unbeatable recon” into style, travel & culture

Top post this summer: No more small plates!!

Share your own Summer Recap

You can see your own summer recap in the Substack app. I’d love to see what you’ve been reading.

Get my Recap

Quickly, on sports branding re: the Olympics 

• On signed Zendaya, the two co-founders were on the podcast ‘How I Built This,’ released a new technology for racing shoe uppers called LightSpray (the release aesthetic matched that of Nike’s Alphafly 3 release), and added a new ‘Dream Together’ page about their ambassadors before the Olympics, and are releasing mini-docs on YouTube about their biggest athletes (one for Helen and Sintayehu). They also had a fashion/tennis crossover club vibed event in a Navy Yard warehouse during the US Open, pictured above.

• Bandit supplied kits for Saint Vincent and the Grenadines for the Olympics!  

• Cool video on sports tech video (with some ethics discussion) ft. Nike’s test lab at HQ… Nike finally had a campaign that at least got the people talking.

and quickly, re: this blog 

I am returning to writing as a method for making sense of my thoughts this fall. I was preoccupied with life during the summer, but am eager to think through new concepts with you all for the rest of the year :). I have the next few editions scoped out - let me know if you have any requested topics for me to cover!1 With that, a song for the rest:

Inspirations

Intro

Since June, I have been feeling lost about where to best point my professional and creative energy. I have frankly felt unsatisfied with my work - I haven’t felt challenged in any meaningful way since I have started. I feel eager, especially after a calmer senior year of university, to grow by applying my talents to the problems I have so long trained to tackle! I have practiced my role in 3 or 6 month stints many times as a student, but have not found any continuation on those learnings so far full time. 

More broadly, I am also saddened by the general cultures I have found myself in. Ironically, the largest challenge I have faced has been navigating older corporate cultures! I have felt myself yearn for the work cultures described at Stripe, Faire, even smaller consumer startups.2 I am trying, slowly, to bring change to the archaic organization I find myself in, but being a junior and new member makes that… challenging.3 I have also been inspired by people (many mentioned below) that have undertaken a more purposeful path in life - they are ‘live players’ in their lives (agentic people?) and point their attention with purpose, typically disregarding expected norms around progression.4 I aspire to learn from their examples. 

Subscribe now

To better understand where my frustrations grow from and to be able to point towards better examples, I started listing people and institutions who I have long found inspirational and aspirational. I then began articulating why that was the case. Posting it below, perhaps you will find someone interesting5 🙂

Professionally 

Individuals

• Matt Klein

  • Why: Calls himself a “Cultural Theorist + Digital Anthropologist + Strategist + Writer explaining media and social shifts to advise partners on the future.” I think the way he expands on an interest of his (anthropology) beyond his work (which also sounds like a unique, fitted role) is cool. He uses his podcast/blog to speak to some of his academic heroes, which is also really cool!

• Spencer Chang

  • Why: Is now independent after pursuing playful design at various institutions. The way Spencer has been able to support the work he wants to do is aspirational- he wrote a blog recently describing the type of ‘professional freedom’ I also think I am looking for. 

• Reggie James (especially his post on making noise)

  • Why: Strong tech ideology/opinions, multidisciplinary creative inspirations/inputs (many that also resonate with me), etc. Makes a lot of the intersections/near futures he wants happen himself (he organized the talk I went to with Freddy Anzures).

• Ryan Mather

  • Why: Is able to prototype with physical and software in really unique and forward-looking ways. Aspirational ITP-esque work that I think would be cool to work on.

• Jasmine Sun

  • Why: Has been able to connect her love of writing to much of what she spends her time doing – inspiring to see her continue to evolve the way she does this over the years. 

• Elliott & Kristoffer

  • Why: Tiny Awards, poetic computing, Extra Practice as a shared studio…! So much inspirational work that I wish to apply to my own efforts eventually. Kristoffer also operates Cloudlord Management.

• Henrik Karlsson

  • Why: I think the way he approaches designing his life is something I want to begin trying. He approaches his life and raising his kids with a lot of agency to result in outcomes he has thought about. 

• Adam Ho

  • Why: The way he continues to evolve his aesthetic, while also working on new ‘frontier’ technologies, is interesting to me.

• Kelin Zhang

  • Why: Strong personal perspective, multidisciplinary designer, understands personal impact on professional career. Also has conviction to follow her wishes → basically imported the viral French Olympic muffins into NYC through her own willpower (audience on social too).

• Jordan Singer

  • Why: Cool ideas around new frontier technologies and design. He is then able to execute (build) his ideas really well (and seemingly taught himself many of the skills necessary to build well).

• Matthew Siu

  • Why: His creative explorations of interfaces are really fun!

• Kevin Cannon

  • Why: Fun mix of tech and design, across agencies to AI startup to university (teaching on the side).

• Laurel Schwulst

  • Why: I think the way they approach technology (especially ‘soft tech’) with an artist’s perspective is really cool. They are seemingly finding their own professional path with this perspective.

• Willem Verbeeck

  • Why: Excellent independent documentary photographer, but has also carved niches in automotive (both photo and video), photography bags and accessories (and even built out a physical popup for his bag brand by himself) - I admire how he has a distinct voice across all of his content, and how he has projects where he visits locations many times over multiple years to make a project. 

• Gracias Ventus

  • Why: Created her own clothing brand centered on luxurious materials. I admire the way they brand their releases, and how she discusses running a business from SE Asia. 

• Austin Robbs

  • Why: Patagonia gear designer - I admire how much he tinkers with his own gear to make it fit himself/keep on going.

• Ben Thompson (Stratechery author)

  • Why: Clearly articulates his perspectives, has been independent and monetized in a way to keep it that way before it was the fashion.

• James Polshek 

  • Why: Believed that architecture should be a catalyst for change and serve the public good. More: In Memoriam James Stewart Polshek - Columbia GSAPP 

Groups & Institutions

• Basically anyone in the Baukunst Collective

• Los Feliz Engineering (Jake Dow-Smith made their site)

• Deep Science Ventures

• NaN type (makers of this game to advertise one of their new fonts)

• Buildspace (a16z + Y-Comb funded maker group - unfortunately closing up shop because they did not think they could scale much further)

• Sanctuary Computer (Software studio part of a co-op with other studios. I like how they publish yearly reports on finances/sustainability and experiment with org structures.)

Generally

• Food! Dali food Substacks, OTR in-depth food history, Tasting History w/ Max Miller general video history, Kenji Lopez Alt…!

• Generally, specialty coffee as experience and how to convey its value (some examples).

• Makers of free tools like this dithering tool that I love, or this Spacetype Generator (thank you Connor for sending this to me!) 

• Amateur/not fully pro athletes that still perform close to elite levels (Bandit media members/Tracksmith semi-pros, etc.)6

  • KARWENDEL | The North Face 

  • Floberg Runs - YouTube

  • Karue Sell (going pro in tennis at 30)

• This Latoya Ruby Frazier’s Moments of Solidarity exhibition at MOMA was one of the most moving things I’ve experienced at a museum in a long time, surprisingly in a way that only an in-person art exhibition could be even though most of the medium was photography!

• Repurposing old server racks as storage systems for homes. 

In Summary

After beginning a new job roughly two months ago, I am still unsatisfied in many ways.7 I remain optimistic and patient that I will find a role that is right for me eventually. In the meantime, I will continue to focus my energy on what fulfills me.8 That is, as long as my day job doesn’t get in the way…

Misc. Links

in case you wanted more

• NY tech scene (vs. SF)

• Dive into Mondragon, the mega cooperative from Spain

• Matt Klein on Curators

• Language and Writing as it Relates to Progress

• Interactive slime mold modeling

• Skate Oregon 

• Unconventional mobile forward design site

• eInk camera

• Germans are installing solar panels on their balconies 

• Munich 72 Olympics branding (other) (also Mexico 68) is cool! 

Conclusion

That’s all for this week! Thank you for reading, take care :)

-Andrei

Updates

Sfant nº2 last Saturday was a success! We had a good turnout, featured a local coffee roaster (@thoughtfulcoffee) and raised $90 for the NY Public Library and Doctors Without Borders. This time around we got to feature our new living room; Lily and I demo’d our flex wall a few months ago and have been re-doing our living space. The space offered a much calmer environment and people had a good time enjoying our baked goods!

The tennis ladder I joined for the summer has begun! It has felt good getting back into the rhythm of playing tennis consistently, but I am definitely rusty in my matches. Hopefully I can make the top 16 and play in the end-of-season playoffs! Would be fun to have some fans  friends come watch 😛.

Excited for two classes (riso [!!!] printing and a dye workshop) in early June. If anyone has anything similar scheduled later this summer, lmk! I am eyeing an online French class or perhaps a typography intensive at Cooper Union 👀

Work has been unfortunately depressing the last 4 months, and this week I realized it will continue to be this way for at least the next year and a half as things stand. Most irritating is the feeling that I have stopped learning (beyond corporate politics, that is) and won’t have a real opportunity to learn what I hoped I would anytime soon. I have therefore started putting out feelers to see if there are any better opportunities for me… if you know anyone working on a project you think I’d find interesting, hmu!

Athlete (Prosumer?) as God Follow Up

Furthering this blog, if you haven’t read it before:

The Sinking (?) Icebergs

Incumbent brands have continued to fall off the pace in the months since my last post. Nike has had layoffs, seemingly caught in the middle of two opposing directions. Under Armour has brought back a former CEO after its stock continues to slide - down 5% this year, and 70% in the last five…1

New Balance, on the other hand, has continued their growth. It seems that New Balance is connecting with premiere prominent young athletes (like Bukayo Saka, Coco Gauff, Shohei Ohtani, Sydney McLaughlin-Levrone, Jamal Murray, Tyrese Maxey, and Cameron Brink most recently) in a way Nike used to be able to. They are increasingly building a portfolio of prominent cultural brand ambassadors as well (like Action Bronson, Amine, and Chito Vera to name a few). This is unique from Nike, especially in the way New Balance is combining these portfolios. 

New Balance have begun synergizing their athlete and cultural brand ambassador relationships: Bukayo Saka of Arsenal fame has recently teased an ‘ALD Cleat.’2 This diverges from Nike - although Nike did play at this prestige… Federer’s gold trim blazer comes to mind3 - because of these talent brand-linked limited releases, giving the talent direct upside in the brand’s performance. It will be interesting to see if New Balance can maintain this connection with ‘the culture’ beyond a few cycles. 

Leave a comment

On Spreading Like Slime

On has expanded in a number of ways since my last post. They have entered new sports, expanded unique programs, and have also begun leveraging their talent with their partners. 

On uniquely (at least as far as I’m aware) offers a subscription shoe (Cyclon) model.4 One used to be able to order a special shoe, wear it, and send it back to On when it was worn out. They would send you another and recycle the worn shoe, claiming they were working towards a circular shoe economy. The Cyclon program now includes three subscription shoes and a purchasable t-shirt. I am curious to see how this program continues to expand, I think it is genuinely quite innovative. 

Culturally, On has been following the New Balance playbook in their own way. They have recently brought on FKA twigs as a Creative Partner. FKA twigs will be the face of the FW24 training collection, which they are helping develop.5 They have paired this by using their signed athletes (Ben Shelton, Masato Yokata, Alexandra Burghardt, and Aaliyah Mille to name a few) in a collaborative campaign with Spanish and LVMH luxury house Loewe.6 This positions them well into the ‘luxury’ space of technical clothing that they aim for, and away from New Balance’s ‘street’ and street-adjacent influences. 

This prestige is being furthered with mini-documentaries of their highest performing athletes, like record-breaking and world #1 tennis star Iga Swiatek and runner Hellen Obiri.7

On has also branched out into new sports in major ways this year. They have launched the ‘Cloudpulse,’ the first On shoe built solely for the gym (verging on CrossFit). They have expanded their trail running collection of shoes and added new trail related clothing. On has also finally released their tennis apparel to the public, which sold out in days. 

Bandit On A Rampage

Bandit has been on a tear in 2024 with their spring and summer collections. They sold out of nearly the entirety of their spring collection in the first week, Their summer collection is currently sold out of popular sizes after a week of sale. 

Bandit have continued their more social media savvy ‘influencer as partner’ model to grow their audience. Joe Greer, a Bandit sponsored runner and brand photographer (700k IG followers, 100k on a dormant YT) continues to publish ‘behind the scenes’ content of their upcoming collection photography projects. Recently, this has included videos by other prominent NYC photo/video creators like Chris Chiu (200k followers cross platform). This has continued amplifying their brand to adjacent communities. 

Subscribe now

Continuing on that trend, Bandit is opening their second brick-and-mortar location in the West Village. This will be a shared space with trendy Greenpoint cafe Rhythm Zero… see for yourself on their IG.8

Conclusion

The graph of aesthetics among these brands and other new companies in their locales may be the topic of a future piece. Bandit has focused heavily on short clips posted across their social channels to gather momentum behind their releases and announcements. Fellow Brooklyn brand The Browser Company has done the same for all new feature announcements, and the aesthetics are more similar than different…!9 Perhaps the aesthetics of near-waterfront Brooklyn startups is worth analyzing in a future post (here come the millennials).

Can’t help myself… a few fun links 😛

• Video about track and field athletes competing to represent India at the olympics 

• Lupe Fiasco (!!) wrote for Wired how tech is transforming rap beef 

• Cool infill lot house in LA - clad in cork!

Outro

That’s all for this letter 🙂– crazy to think that I am nearing a year of working full-time… I am excited for the fun and learning I have planned this summer and hope you can all enjoy some good weather. 

Take care,

Andrei

My top articles to start the year (a little late 😉)

1. Why The Age of American Progress Ended

   1. I generally agree with much more than the misguided 'it's time to build' doctrine much of tech seems to align with. 'Great Man' theory is not real!

2. How Colleges Make More Money Than God By Giving It Away

   1. Illuminating piece on the financial logic of college tuitions.

3. Nameless Feeling

   1. Lily and I have been picking up these themes repeatedly these last 6 weeks or so. How much of this do you see in yourself?

4. No Good Alone

   1. Have found myself thinking about relationships as I’ve spent time with some older ‘millennial’ people of late (at work, friends of friends, etc.) and think of this piece.

5. Why We Need New Words for Nature

   1. I was on a linguistics-theory reading kick last winter, and this piece stood out. While this piece focuses on how we speak about nature (or don’t), I found it interesting to extrapolate this point to other dominant themes in American life, like finance, education, politics, etc.

and some books!

1. Three Body Problem Series, Cixin Liu

2. The Glass Bead Game (Magister Ludi), Herman Hesse

3. Finding the Mother Tree, Suzanne Simmard

4. 21st Century Monetary Policy, Ben Bernanke

Coffee Shop Popup in my apartment!

I started a coffee popup brand, named Sfânt, and hosted an event in my apartment! Great turnout, lots of fun, will definitely pursue this more :). I’ve linked the Instagram above, here was the flyer I sent around, among some other photos,

Time, Innovation, Political Economy

• Pace Layering - been thinking a lot about perceptions of time recently

• Climate Economics Has Let Us Down

• How Social Engineering Drives Technology - Samo has been one of the more interesting thinkers who I follow

Misc. Personal

• Tit for Tat Game Theory - I remember learning about this and running a little competition like this in high school CS. Nice summary video, did anyone else do this in class?

• Richard Sapper - His Tizio light is so nice

• James Dyson is working on farming… and it’s kinda sick???

• on beauty and self image

• emotion labor isn’t recognized all that much imo by people my age… at least sometimes

• thought these stories were inspiring (from VC Hall of Fame member Bill Gurley)

Photo and Architecutre

• It is interesting how Ed Templeton decides to stage his photos

• B&W tones :O

• pretty photos

• I love the glass exterior and bookshelves inside

• pretty cabins :))

Conclusion

Hope you are all finding the start of the year to be peaceful and positive :). I leave you today with this:

I met a traveller from an antique land,

Who said—“Two vast and trunkless legs of stone

Stand in the desert. . . . Near them, on the sand,

Half sunk a shattered visage lies, whose frown,

And wrinkled lip, and sneer of cold command,

Tell that its sculptor well those passions read

Which yet survive, stamped on these lifeless things,

The hand that mocked them, and the heart that fed;

And on the pedestal, these words appear:

My name is Ozymandias, King of Kings;

Look on my Works, ye Mighty, and despair!

Nothing beside remains. Round the decay

Of that colossal Wreck, boundless and bare

The lone and level sands stretch far away.”

- Percy Bysshe Shelley

Intro

Hey everyone! In this edition, I discuss some dramatic updates relating to my last post and talk about camera design. Hope all is well, and that you are able to enjoy Thanksgiving this week! With that…

Top 3 Picks

1. TikTok's Eurocore Summer - I think people in the US who want things to change (that I have been around, at least) like to imagine the positives of other realities without considering the tradeoffs. Perhaps this is because they wouldn't be tolerant of the tradeoffs... I find this pining frustrating when they would not be okay with changes necessary to reach them here in the US.

2. Longtime area photographer’s tour of Manhattan's Chinatown (Nowness produced)

3. A recent New Yorker re-look at WKW's In the Mood for Love with an interesting scroll format :)

Misc. Links

• Spain lives in flats - very nicely told visual narrative on why and how so much of Spain lives in urbanized flats.

• The Netherlands using stable diffusion to advertise their urban design and promote tourism… sick. Here is my street done up pretty:

• Are you an ethical tourist?

• Benedict Cumberbatch reads Sol LeWitt's letter to Eva Hesse

• The tradition of knowledge behind ASML - relationship between the social democracy welfare state and industry is interesting

• 032c interview with Ouana Stǎnescu - interesting commentary on her generation's faults, Romanian POV, and this quote (my own emphasis):

  "Quite often in the States I find, with the hierarchy of the school systems, and capitalism, and the competitive nature of everything, in spite of this abstract notion of freedom, the systems are suggesting the opposite. They set everyone up in a very narrow rat race. And you are led to believe there's only one way of studying architecture or being an architect and that's fundamentally untrue — not for anything. Not for music, not for arts, it is true for absolutely nothing, right? Because it’s less about excelling in your own way, but it’s always comparative. So, again in the face of adversity it's important to not let it reflect back to you, to not start believing other people's bullshit, it's their own problem, it's not yours. It’s hard."

Update on my sporting brands piece

First, this Alexandre Arnault Q&A at The Oxford Union taught me new perspectives on how LVMH views luxury, and how they view building their consumer base.

Second, Joe Greer, the photographer/runner behind Bandit's media work this year, called out On Running for copying his work... here are his Instagram stories:

I don't think this is surprising given On's position between these smaller niche-brands and bigger incumbents. It is a common tactic to lift the creative/culture from smaller brands in fashion, after all.

Third, David and I have been trying some of the events the running space had on offer around the NYC Marathon! We went to NYC's 'Sprint Collective' partnered event with Saucony and Brooklyn Running Co. Around 25 people showed up, got free food, and the chance to do a sprint workout in Saucony's speed shoes. Pretty fun! We also went to a Bandit x Shopify popup event at Shopify's Soho space. Joe Greer and Ardith Singh (Bandit Co-Founder & Chief Design Officer) spoke about Bandit's creative direction and Joe's relationship to running. I thought the way they use their location in Greenpoint/Brooklyn to inspire their creativity less shallow than I initially thought.

Camera Bodies (film vs. digital)

I have recently sold my digital camera body, and replaced it with two film cameras :). I am now a #onlyfilm photographer, lol. My main issue with my digital camera is that, although it was a great digital camera, it just didn't get me excited to make photos. It was heavy and needed to be charged. What annoyed me the most is that the pictures always needed to be edited after the fact. I have opted for a largely automatic (yet still high quality) small 35mm camera (a Contax G1 with a 28mm Zeiss lens) and a compact medium format camera that offers more manual controls (the Fuji GS645 Pro). I am excited by the results I have already had with these cameras!

While I was looking at film cameras, I was struck by the huge amount of varying designs - and by how different they looked from today's cameras.

Look at these designs! Rather disappointing that the tools of today don't inspire as these film cameras (at least to me they really don't). I find the Pixii camera an interesting bridge between the eras - definitely a startup worth following if you are a photographer.

Conclusion

That’s all folks — cheers. I’ll leave you with this poem to end, found it in a reflection on the conflict in Palestine…

Intro

Hi everyone! Sending you all a short update this month -- I found some links I wanted to share, relating to 'design' more generally. I have found myself excited to learn more about great design recently - more on that below - and playing games like CS2 and Mortal Kombat 1, and watching 90 Day Fiancé!

I am stoked to race the half marathon I have been training for this Sunday! I am also eager to learn more about coffee in a Roasting Masterclass I am enrolled in next week, and potentially host a coffee popup in my apartment with some friends a few weeks after.

Work has been going smoothly. I am working on quite a few projects, relating to live sports on the legacy WarnerMedia 'TNets' (TNT, TruTV, TBS), the legacy Discovery ATVE brands (Food Network, HGTV, TLC, etc.), and of course the new product we are working on behind the scenes :). It is becoming easier navigating my 'work persona' in the office and am glad I feel I can at least be a little bit more like myself with my immediate team.

Important Links, Ranked

1. Charles and Ray Eames Design Q&A (1972) honestly the last answer is so clever... can't believe I haven't seen this before. Becoming a bigger fan of the Charles and Ray Eames the more I learn about them :^)

2. This is probably my favorite space featured on the NTS channel - been revisiting their videos. I think it is incredible how you can see his style so clearly throughout the space, and I admire how fine tuned the space is to his needs.

3. Are.na board titled 'life is weather, life is meals'

4. Thought this profile on Maya Man was interesting - more current aesthetics.

General Links & Notes

• New paper analyzing the results of over 3,000 antitrust lawsuits by the DoJ brought forward between 1971 and 2018. They found that the suits increased economic activity... interesting context to Lina Khan's recent suits.

• Andrew Yeung's '10 things from 27'

• The Nation Op Ed on how we've watered down 'gentrification' to a near meaningless point.

• Reading

  • I finished Ben Bernanke's 21st Century Monetary Policy recently and have mixed feelings about it. I thought the way he walked through the Fed's last 80 years, adding his own commentary and opinions, was enlightening, but thought the conclusion and broader message at the end was a little weak.

  • I also recently finished Liu Cixin's Three Body Problem (thanks for the rec Nick) and liked it a lot - it's being turned into a Netflix movie soon. I am midway through the second book and for a while felt like I was reading a sci-fi, more 2020 modernized Murakami novel, which is a good thing for me!

  • Julia Watson's Lo-TEK: Design by Radical Indigenism has also been very fun to read, there are some great illustrations that help visualize what she's talking about.

On AI rn…

I was talking to Eric about how disappointing we have found many AI products to be in our own usage so far. Naturally, given the hype, I tempered my expectations, but still feel like many applications of the technology haven't been placed where they can hit their stride. The 'falling behind' (FOMO-adjacent) culture in Western technology companies has furthered the hype and lazy adoption of AI tech even more.

I think it is honestly worth giving Apple credit here. They have seemingly been slow to add AI features, at least as of now. But, looking back, they have been consistently adding many subtly applied AI features where it actually ads utility to various parts of macOS and iOS for the last few years. Those are my two cents for now at least ¯\_(ツ)_/¯ do any of you have any products you especially enjoy so far? I have personally found tools like Careful Words to be more useful than many AI editors... but maybe my intentions are different than what AI copy editors mean to provide... (less sales driven). While I understand why, I still think it is sad that AI is being explored in such a profit-ingrained way. I feel that a more imaginative and explorative purview could create much more exciting designs.

Conclusion

That's all for now... for the next edition, expect some more in depth writing on *camera bodies*... 😉

Take care everyone!

Andrei