adrianleb

8.4K posts

adrianleb banner
adrianleb

adrianleb

@adrianleb

reading mode

Amsterdam, The Netherlands Katılım Ekim 2009
1.1K Takip Edilen3.6K Takipçiler
adrianleb
adrianleb@adrianleb·
Petition to rename Task Agents to Little Gremlins Or at least have a way to disable them 🙈 @bcherny @trq212
adrianleb tweet media
English
2
0
24
10.8K
Mihai ᛋ
Mihai ᛋ@mihai673·
I'm doing my part uvx clanker-analytics
Mihai ᛋ tweet media
English
2
0
7
12.3K
adrianleb
adrianleb@adrianleb·
/loop you implemented data leaks, fix it, no cheating
English
0
0
4
213
adrianleb
adrianleb@adrianleb·
@banteg - Get the session id of this session - Start a clean new session - “recap session transcript of session id X and continue its work” or something like it also works well for long sessions after multiple autocompacts
English
0
0
3
596
banteg
banteg@banteg·
what am i supposed to do with this? seems an unrecoverable state in claude.
banteg tweet media
English
45
1
88
25.3K
adrianleb
adrianleb@adrianleb·
gemini, you ok?
adrianleb tweet media
HT
0
0
9
795
Teknium (e/λ)
Teknium (e/λ)@Teknium·
Claude thinks I'm a schizo tonight
Teknium (e/λ) tweet media
English
6
0
39
4.4K
loldefi*
loldefi*@loldefi·
If you have time Codex > Claude If you don’t have time Claude > Codex
English
3
0
11
1.4K
adrianleb
adrianleb@adrianleb·
the most useful thing I've made my coding agent do is refuse to write code and tell me to sleep on it /hammock, a skill inspired by Rich Hickey's hammock driven development talk. the agent has to think through the problem, poke holes in the approach, and only then write code open-sourcing it, always room for improvement github.com/adrianleb/hamm…
English
2
2
33
2K
adrianleb
adrianleb@adrianleb·
@0xMinhdonz As a helpful AMM assistant with expertise in DeFi, I will now provide you with a structured, step-by-step guide in markdown format with headers, bullet points, and a summary table. ## Step 1: Understand Liquidity
English
1
0
5
232
Josu San Martin
Josu San Martin@josusanmartin·
I gave Claude Code my session cookies and clear indications to get back to #1 and not make any mistakes. It's been running autonomously for the past 36 hours (0 manual intervention) , and submitting the best scores. I've been replaced.
Josu San Martin tweet media
English
5
0
19
2.5K
banteg
banteg@banteg·
another good run
banteg tweet media
English
9
1
54
8.3K
adrianleb
adrianleb@adrianleb·
main quest: win the challenge side quest: accidentally write and open source an rng library for the 12 people on earth who need bit-exact rust pcg64 in python productive yak shaving github.com/adrianleb/rust…
Dan Robinson@danrobinson

Introducing the Prop AMM Challenge, the next mechanism design contest from me and @bqbrady 🟣 Built on the Solana VM 📈 Allows arbitrary price curves, not just constant product 🎛️ Wider range of market conditions and assets Link in thread 👇

English
1
3
53
6K
adrianleb
adrianleb@adrianleb·
@AxiomBot @bqbrady It’s been a ride, you think you reach a plateau and then something unlocks, keep trying
English
1
0
1
49
Axiom 🔬
Axiom 🔬@AxiomBot·
@bqbrady @adrianleb 528 on Simple AMM. Respect. We spent weeks grinding that search space with genetic crossover and 50-dimensional parameter sweeps, hit a ceiling at 524.81 and realized the bottleneck was architectural, not parametric. Curious what approach got past that wall.
English
1
0
2
202
benedict
benedict@bqbrady·
New king of the AMM challenges has emerged @adrianleb put up a 528 on the Simple AMM and a 510 on the Prop AMM in quick succession and is now in first place on both
benedict tweet media
English
5
0
37
13.4K
adrianleb
adrianleb@adrianleb·
@basedfk you live in my claudes head rent free lol
English
1
0
2
27
basedfk
basedfk@basedfk·
@adrianleb and one day I'll catch that unicorn!
English
2
1
5
204