William Wolf

@taimurabdaal Many many people genuinely enjoy booking trips. It's a rare luxury! Same goes for things like reservations at restaurants.

Really don’t understand the tech obsession with trying to automate the process of booking flights… I think I travel reasonably frequently and it’s really not a big deal to click around for a couple of mins to book a flight…

it's much more than two websites and a chat app. and they serve billions of people. Facebook, Instagram, WhatsApp, Messenger, Threads, Reality Labs. Composed of Reels, Marketplace, Groups, Stories. Best in class Ads platform, business tools, payments. Reality Labs did $2b in revenue in Q1 2025 alone. Huge AI research org. Building and operating a global infrastructure at Meta’s scale (data centers, networking, custom hardware, ML training clusters) is one of the hardest engineering problems in the world... not procurement.

@willium @Amit74582331 @TheGeorgePu Still it's two websites and a chat app. Their hardware salea are near zero. And the the billions in infrastructure is just procurement.

Meta cut 8,000 people today. A survivor wrote about a teammate who slept 4 hours a night for months. Commits at 3am. Commits at 6am. IC4. Strong reviews. No PIP. Cut anyway. Working harder doesn't move you up the layoff list. You don't survive by being valuable to them. You survive by not needing them.

@Spectrumofreas1 @Amit74582331 @TheGeorgePu Meta makes $200B/year, 2000x the HW company you worked at. 2000 * 50 = 100,000. Meta has a ton of properties, infra, billions of users, and advanced hardware.

@Amit74582331 @TheGeorgePu I believe you. I worked at a $100M/year HARDWARE company with 50 employees. Software should have 10X more leverage.

@MikeIsaac record revenue, but capex is up 5x

someone smarter than me relayed the point that in 2023, which MZ called "the year of efficiency," some knew layoffs came amid an overall digital advertising industry downturn (see my '23 article below) contrast that with Meta reporting record revenues last quarter

this, in my opinion, is one of those moments where i think meta execs knew it would be a huge hit morale-wise and probably would be like 2023's "year of efficiency" for a while but could weather it perhaps b/c i'm too close to it re: sources, but vibes feel markedly worse

@nbaschez cool!

Introducing Roughdraft! A new open source project designed to make collaboration with agents better. The idea is to bring commenting and suggested changes to markdown (e.g. plan docs) in a nice interface. Free, local, etc. 👉 roughdraft.md 👈

@nico_laqua @WillManidis I feel like you're saying symbols matter, and Will is saying that it isn't a question of whether symbols matter, but what happens when the symbol becomes the thing itself and nobody really asks what the company does.

It’s not about cargo culting, it’s about increasing your surface area for ambition. If you want to do something world historic, something big, then sometimes the symbolism behind actions signifies something - there’s a reason why signs on the side of the building matter, why beauty has value, why symbols predate history. And the symbolism behind going all in, behind doing big things, makes it such that when important decisions arise, when there’s a fork in the road in which one may choose a more defined path with little variance and ensured profit or a path with uncapped upside and more variance that is harder to pull off, the latter is more likely to be chosen. And those small decisions to be more ambitious, to care more, to respond to that annoying customer service question at 2 am even when you don’t want to, compound and in a nonlinear way make an important company, an important outcome, more likely. If you’re doing anything at all, you might as well do it right. Nor if you notice is it prescriptive that one must go all in on startups or anything else, but right now, most young builders are told that it’s morally wrong to go all in, and that deserves pushback, because being a bit more “hardcore” is a valid option that worked for lots of important “founders”, from Hernan Cortes to Napoleon to Bill Gates. Others probably accomplished just as much enjoying leisure, like Carnegie, and that’s great! But shaming those going all in is a symptom of cultural nihilism.

You can run your company taking lots of breaks, “working” from home, showing up 2/3/4/5, however many days per week you want, it’s your company, not mine. Writing a long ass article attacking people who care and push themselves hard to create good products comes from a place of insecurity dressed up as intellectual rigor. Working hard and caring is ok, actually. Maybe you can be a pro athlete going to the gym 1 day per week, maybe you can be the best military general not marching on weekends. But attacking those that go to the gym 6 or 7 days per week because they care and because xyz athlete never worked out and did great comes from a bad place, and is bad for the industry, bad for progress, and bad for the soul. It’s funny how those that don’t push themselves have an absolutely visceral reaction to those that do - I certainly don’t write long articles against working from home or being slothful. Bill Gates was always in the office, and I think Microsoft is pretty cool.

Very cool to see OpenAI using Vega to make their charts :)

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

@heysaik +100000000 I spent the last couple years building for older adults. people in SV just don't realize that most people don't have calendars or inboxes like them. and the things people want to automate are probably not what you think either (e.g. they enjoy trip planning)

As someone building in this space for almost a year, the biggest finding is how infrequently the avergae consumer uses their calendar the way you think. Plans with family and friends tend to happen in person or over text message. Calendar is sparsely used. Then there's also folks who use their calendar as a reminders app adding blocks like "FOCUS" or "PLAYTIME" where the person knows what it means but an LLM would have no idea and hallucinate. But unfortunately, unless we all consent to 24/7 surveillance with personal AI wearables always recording, a calendar is the best way to understand a user's life from a chronological lens to generate proactive suggestions. I have more thoughts on the general idea of proactivity in the average consumer's life. Maybe I should write an article

@lvwerra so cool

We are releasing Carbon: a crazy fast DNA model Carbon is 275x faster than the next best model. So fast you can process the whole human genome on a single GPU in <2 days. Here are the tricks we used: When modelling DNA sequences a lot of the performance comes down to tokenizing the sequences in a smart way. BPE tokenizer struggle because there are no whitespaces and character (called base in DNA) level tokenizers waste a lot of compute on too many tokens. Carbon is built with a unique tokenizer: we split sequences in chunks of 6 bases, but during both training and inference we can work with single base resolution. That's similar to having word tokens but resolving them at the character level. All possible thanks to the DNA tokens unique structure. The architecture combined with the tokenizer makes the model 275x faster than the previous SoTA (Evo2) at this size. We built an interactive demo so you can explore how the model can generate DNA sequences, investigate the structure of genes, predict the effect of mutations, generate and fold proteins and even reconstruct parts of the tree of life. huggingface.co/spaces/Hugging…

@MarcusMilione legend

Sold out in 5 minutes 🙏 what a crazy rush, just sit here trying to deal with it all forever grateful for anyone thats supported at any point along this journey back to work

@ankitkr0 my friends are too cool to like my slop. they keep staging interventions! not cool

this almost sounds very dumb but the key to a good launch on X is just being friends with like 100 people, if you have that and ship a 4/10 product, you will most probably hit 100k views on your video/demo type shit

it's something about the aquavit, gives swedes superpowers. they do not fuck around.

@willium @bernhardsson

Over the past 2 years I've seen the smartest technical minds in the world waste half of their time trying to go viral and win internet attention. It needs to stop. Help me unfuck marketing!

@wuseokjung_ maps.app.goo.gl/htoafmwBMMoDMK…

we’re opening a corgi office + cafe in nyc. we move fast and have a couple asks. let us know if you know any cool: - office space - cafe space (prev f&b) - gtm/marketing/finance friends separately, if you're interested in joining us here, comment/shoot me a dm!

@itunpredictable @yrechtman Double, double toil and trouble; Fire burn and caldron bubble.

@willium @yrechtman Toil in obscurity. It's great.

@lenagrundhoefer chews me up and spits me out every few years

living in nyc: year 0-1: i like it here year 1-5: i love it here year 5-7: i like it here year 8+: i need to split time or get me out of here

@itunpredictable @yrechtman The waters are polluted. The algorithms (X and LinkedIn especially) reward the rage and FOMO and last ticket out of Saigon frames. No one is in it to solve problems anymore, they're in it to catch a vibe and be relevant. Feel relevant. Feel something.

@willium @yrechtman's slop cannons

@shadersweden god damn

It’s live. Plugged in. Ready to execute. shader.se