Steven Yin
@stevenydc
Followers
398
Following
562
Media
4
Statuses
177
macrodata refiner @openai prev: HFT @weHRTyou blog: https://t.co/aLvLWaCzog
Joined September 2010
Ads aren't inherently good or evil. They are fundamentally a market for attention. Attention is humanity's ultimate scarce resource. Our goal should be to allocate it as efficiently as possible.
2
2
51
Hype
Over the past few months, OpenAI models crossed a threshold: we’re seeing early/small-scale but repeated examples of GPT-5 meaningfully contributing to novel research. AI is the next great scientific instrument, and it benefits every field. Progress accelerates when researchers
0
0
1
🦄
Here at @OpenAI we've cracked pretraining, then reasoning, and now we're experimenting with a new set of techniques that maximally leverage their interaction. GPT-5 is just the first step in this direction, and we're incredibly excited to see where scaling this up will lead us!
0
0
4
With AI assistants (like ChatGPT) becoming our main gateway to information, the isn't about eliminating ads, it's about designing attention markets that are maximally efficient. New blog post on AI and advertising:
syin.bearblog.dev
Ask someone if they like ads, you'll almost certainly get a "no." Nobody wants their YouTube videos interrupted by a gecko. But directing people's attenti...
1
0
4
Ads also fund the free digital public goods: search, news, videos... Think of ads as an attention tax, they fund digital public goods the same way your income tax funds schools and roads.
2
0
6
Directing attention to businesses generates real economic value. When done well, ads guide you toward products and services you genuinely benefit from, and they help new products and services get off the ground.
1
0
5
Ads create a mechanism where businesses compete openly: those who value your attention the most bid the highest for it. A fair, competitive market is a necessary condition for an efficient attention market.
1
0
5
There are regulations, and there are etiquettes. This exists in sports, finance, AI research job hunting. These things are technically not in the rules, but if you break them people are going to be mad at you. The alpha is in knowing when to break them. (not financial advice)
0
0
3
Meta is giving classic CitSec vibe these days. Zuck is the new KennyG
1
0
3
We needed internet scale data for pre-training, now we need planetary scale environments for RL.
Why you should stop working on RL research and instead work on product // The technology that unlocked the big scaling shift in AI is the internet, not transformers I think it's well known that data is the most important thing in AI, and also that researchers choose not to work
0
0
7
There are times when it feels like we've been doing thousands of years of philosophy just to prepare for the current moment.
76
49
745
Humans started with an intuitive understanding of "water": feels wet, becomes solid when cold, tastes like nothing etc. But eventually we formalized "water" with H2O. We all have an intuitive feel for "intelligence", one day we are going to formalize it into ... ?
0
0
2
Full post dives deep into how compute and intelligence futures markets would transform AI from speculation to calculation:
syin.bearblog.dev
OpenAI is about to spend $500 billion on infrastructure they might not need: if researchers aren't ready with experiments, or if demand for intelligenc...
1
0
7
Even better: intelligence futures. Contracts for "1 billion GPT-5 tokens in 2026." The day these hit $10k/million tokens is the day AGI has found product-market fit: we would all happily pay if they are able to solve some truly economically valuable tasks.
1
0
3
Labs making breakthroughs → buying every contract → prices moon Progress stalling → dumping excess capacity → prices crash Want to know if AGI is near? Check the compute futures curve. The market would know before anyone else.
1
0
3
Imagine compute futures: "100 million H100-hours deliverable March 2026" Suddenly AI labs can hedge costs. Plan infrastructure. Lock in supply. But here's where it gets more exciting: these futures prices would be a real-time prediction market for AI progress.
2
0
3
AI companies are the new utilities. Compute goes in → intelligence comes out → distribute through APIs. But unlike power companies who can stockpile coal, and hedge natural gas futures, OpenAI can't stockpile compute. Every idle GPU second = money burned. Use it or lose it
1
0
13
What if the futures market could tell us exactly when AGI will arrive? Not NVDA stock. Not Twitter takes. An actual market that prices AI progress in real-time. This market should and can exist, here's how.🧵
2
1
19
Race to the bottom in AI pricing is a sign of no product market fit: if we have found truly economically valuable tasks for AI to do we would all happily pay $10,000/million-tokens. People obviously wouldn't want to pay if all they are doing is solving AIME puzzles.
1
0
4