@stablechen

AI R&D @IdyllicLabs, prev: dev rels @terra_money, head of R&D for wasm devx @cosmology_tech & @terran_one, @ucberkeley ('19 dropout)

San Francisco
Joined October 2015
2 forces - the push to make things harder and challenge yourself, and the pull to make things easier and systematize I often find insisting that things should be easy from the start, rejecting anything manual, building systems rather than engaging in reality - is a mistake this has held me back
2
388
Will Chen retweeted
Jev + Astra beats the Ender Dragon in Minecraft in 8 minutes 43 seconds! ⏱️ Cost less than $1 ($0.01 Jev, $0.96 Astra) I open sourced the code and explain the harness setup below. This type of movement is only possible with Jev's near instant decisionmaking, and some continually learning skills from Astra.
148
344
174
5,325
885,908
tokenized time
businesses on @calcom sold $17M worth of services creatives, lawyers, accountants, coaches, … the service industry lives on the calendars and earns on cal.com
2
5
1,783
there’s a common misconception that thoughts are virtual and separate from their physical substrate we treat them like neutral electrical signals because we experience their meaning rather than the biology producing them
Repetitive negative thinking is linked to Alzheimer’s protein buildup in the brain.
5
1,825
creative ways people are converting text problems (large search space) into enum classification problems!
All of the coolest Jev projects I could find on X today 🧵
4
1,908
Jev picking the next command from shell history
33
57
13
1,773
118,073
weekly limit is the new 5-hour limit
3
11
1,979
No he’s right; intelligence can naturally exploit any channel or combination of phenomena where causal communication is physically possible, even ones that seem absurd / impractical. This reaction to it just highlights humans are prone to underestimating AI
Okay now they are just making shit up
1
4
1,895
Curious wen Jev pragmas in openprose @irl_danB
I've seen people describe Jev as an "AI if statement". But what if it actually WAS an if statement? Introducing Probably: a programming language powered by Jev: probably-lang.southpolesteve… Jev baked into the language. “feels” asks a question. “match” routes between descriptions. “while” keeps going until something stops feeling true. This is obviously a toy, but it's fun to think about what something like Jev unlocks. Jev makes the decisions, an LLM does the writing, and a little program ties it together.
6
1,921
i'm drafting a small book covering the main mental models for designing agentic systems, from first principles the book starts with a single LLM call and walks you through how each engineering constraint leads to new structures / patterns and design decisions feedback greatly appreciated! idylliclabs.com/projects/ele…
3
1
10
1,889
What is memory? I used to wonder how the brain could store a feeling. What could be recorded in cells that would let you feel an old fear again? Suppose a pan catches fire in your kitchen. Your brain processes the flames, the smell of smoke, the alarm, and your fear through different but connected systems. The experience changes connections between neurons, associating these parts of the event. Those changes can remain after the fire is out and you have stopped feeling afraid. Later, smelling smoke can bring back the kitchen, the alarm, and some of the fear. A reminder activates part of what was learned, which helps activate other associated parts. What persists is a physical change. What produces the remembered experience is that change acting on the brain. The brain already has the ability to process sights, smells, and emotions. The stored information draws on those abilities rather than having to supply them from scratch. This means memory depends on two things: storage and retrieval, and the ability to interpret or express what was stored. An agent might save: The front door is locked. When it retrieves this sentence, it can decide to try another entrance. But the sentence does not explain what a door is, how locks work, or why another entrance might help. The model supplies that knowledge. Put the same sentence into a program that can only display text, and it will never use it to choose an entrance. The amount you need to store therefore depends on what the interpreting system already knows and can do. A small record can produce a rich response because much of what is needed already exists in the system. In the brain, storage and interpretation are intertwined. A changed connection both preserves a consequence of the past and affects how new activity travels through the brain. There need not be a separate file that gets loaded. So remembering a frightening event does not require fear to sit somewhere in storage. It requires lasting changes that allow a reminder to engage the brain’s existing abilities to reconstruct the scene and feel something about it.
1
7
1,784
to defeat ego: - recognize when you don't know something (mind blanks / is unable to produce specific paths or evidence) - stop attributing everything to your own sense of authorship (we live in a society) - study solutions of others without feeling bad for not coming up with it (we have limited information and resources, someone in the set is bound to find something before you)
2
11
1,971
Will Chen retweeted
i don't want AI in your app i want your app in my AI
151
138
63
1,551
102,138
Will Chen retweeted
only GTA VI can slow down the AI progress at least for a couple of weeks
2
1
11
1,780
I cannot stress how damaging it is to mental health to stay at home. Do something, anything, but get out of the houses. Weekends in particular are for doing stuff, going places. Staying home tires you out and makes you anxious. Climb a mountain or jump in a lake, anything.
is this genuinely the explanation for decline of everything - friendships, excitement, spontaneity, dating?
750
1,391
183
15,394
1,949,800
Will Chen retweeted
Replying to @dlouapre
You are missing several points: 1. By knowing where the destination is & teleporting there, you can easily walk back home. It's faster & easier to find the path between two known points, than having to search for an unknown point in the first place. 2. The purpose of mathematics is not to provide jobs to mathematicians, but to advance humanity's mastery of the universe. We all use tools everyday without mastering their inner workings. That's fine. Even the best mathematicians today don't fully intuitively understand concepts like spinors, Hopf fibration or E8. 3. Human understanding is not guaranteed, and requiring it might be a limiting factor. nitter.cf/timsoret/status/209857…
Replying to @timsoret
Nothing guarantees that the universe is operating at a level of complexity that is conveniently understandable by our constrained biological brains. So insisting that every discovery remain completely human-legible is forcing an arbitrary ceiling on science.
5
5
2
33
8,372
Strong Pokemon. Weak Pokemon. That is only the selfish perception of people. Truly skilled trainers should try to win with their favorites.
267
494
77
7,636
247,415
chat UX -> tinder UX
Introducing Agency. Never prompt again. AI prompts you 🫡 Get 100x more done. Your agent finds useful work. You approve. > full context of your life with me.md > catches things you forgot > fixes issues before you see them > never open Gmail, Chrome or Slack again Runs locally, free and 100% OSS. New interface for everything. ⏩ Tell your agent: start Agency. github.com/browser-use/agenc…
2
5
2,352
If I did not believe in the intrinsic and underrated usefulness of mathematics, I would be somewhat sad about what AI is doing to it. I think Jevons paradox will strike again: much cheaper and deeper mathematics will become even more of the fundamental infrastructure of the future than it is today. Math will no longer be centered on conjectures posed by humans. It will be continually postulated, solved, processed, and utilized by AI systems as they work on the hard scientific, engineering, and computing problems of the future.
27
47
17
478
43,718
Something bad for the (existing) class of mathematicians as they lose their privilege, but not necessarily bad for mathematics & our body of knowledge?
24 Fields Medalists speak out against "A Severe Misalignment of AI in Mathematics" mathandai.org/
153
174
42
2,258
403,981