ChimeraDefi.eth - muShanghai

Defi dev | founder @ | ex - @meta @uwaterloo | Medium @chimera_defi | Email [email protected] | GH chimera-defi |

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

Built token-reduce: a workflow skill for Claude Code, Codex, and MCP clients that reduces total token burn across real coding tasks. 1/5

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@ThinkingUSD Try “helicoptour” if you ever go to Argentina. My Colombian/argentinian gf found it and it was dope af
I think you’ll like it if you’re into this shit

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

Crashed my bike like 3 times on the way home so I’m crippled now. Hurts to type
But hey I can still trade lmao using my voice
My voice is also fkd due to the pack of smokes but I have faith

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@Cryptoyieldinfo Incredible retirement arc lmao
Anything you wanna see on rh chain ? Any easy forks ?

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@macrocephalopod @AgustinLebron3 @TheOneLanceB @grok how realistic is it to get a sharpe >2 or >3? Any historic traders who maintained that?
Are they using PIT sharpe and picking a peak in this convo or excess hac sharpe?

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@adocomplete What? Like this finally lets Claude spin up remote sessions via the mobile app on a self hosted server? Like codex has been able to do for a while?
I’ve been using a hand rolled Claude skill to spawn remote self hosted runners for a few months now

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

Yea just like humans. How many times have you heard good engineers say they choke up and struggle in interviews cos they gotta verbalized their cot? It’s pretty common

[Tenobrus @tenobrus]

okay, so in fact despite all of openai's PR chaff and misdirection over the last few days Astra is a near 10x in no-CoT capabilities, meaning it can do massively more without verbalizing thoughts or intention, meaning it is in fact quite clearly less monitorable.

[James Ross @jamesrosst]

There is a new LP meta game on @RobinhoodCrypto of using @Uniswap uniV4 hooks as a substitute for mempool JIT (just in time) liquidity as this isn’t possible on arbitrum orbit with to fifo ordering

So you use a v4 hook to mint a tight range from inside it and get infront of the swap. Sometimes with a 5% pool fee lol

Especially useful for tokenized equities if you have access to a nasdaq feed or powerful predictive models like we do @SynthdataCo

Gg

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@cryppimagic Is it possible to get historical data on these? I’d like my clanker to review it all
Posts, txs etc

[Weilin (William) Li @hklst4r]

The @NotionalFinance attack repported by @SpecterAnalyst appears to be caused by an integer overflow.

The ERC1155 implementation allows minting a pair of positions (one long one short) when calling "safeTransferFrom" where from does not have enough position to be transferred.

The attacker minted 1 unit of position first: 1 long goes to a helper contract A and 1 short goes to the main contract.

Then mint another 2^128-1 uint of position: 2^128-1 short to main contract and 2^128-1 long to helper contract B.

These sum up to cause overflow in the main contract, allowing the attacker to end massive long position with no debt in helper contract B. It is then used as collateral to drain the tokens.

The attacker pocketed ~$1.7M in DAI and USDC and deposited to tornado cash.

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

Very cool. Ended up building something personal like this this. Got an mcp ChatGPT can use to direct codex and run parallel jobs on my server. Also have a coordination bus like yours and a GitHub issue as a backup bus.

I’ve found the mcp to be super flaky in ChatGPT where I need to fork the convo every few turns to get the custom dev mcp to show up again. Have you seen any such flakiness?

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@erinmagennis Congrats!

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

Marked important enough to get x password reset emails

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@pastienot Wasm runs in the browser no? I want something compiling down to binary with no gc for the performance

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

I think we need something above rust. Something more readable closer to JS or python so you can grok the business logic that transpiles down to rust in case you want to go deeper into the optimizations, types etc

The LLM can work on optimizing the transpiler / compiler in parallel with whatever business logic gets implemented

Good points tho

[Justin Schroeder @jpschroeder]

All code should be rewritten in Rust.

Rust was such a horrible language for humans to write, and that's precisely what makes it incredible for agents. If you're not steeped in languages like I am (was), here's why this matters...

All languages fall into one of two camps: dynamic memory allocation and manual memory allocation. Sometimes languages allow for both, but programmers either needed to manage carving out chunks of memory for their program themselves or to delegate that responsibility to the machine.

The tradeoff between the two is roughly performance vs. pain and safety. Manual allocation is generally faster, but sucks to program and, if not done carefully, can leave security vulnerabilities (C, C++, Zig…). Languages that manage your memory are, generally, much more delightful to write but a bit slower (JavaScript, Python, Ruby, OCaml).

Rust tried to solve the security concerns through, essentially, the most complex language constructs known to man. It was fast. It was safe, and broadly it was the wrong decision for most projects because it was too hard to learn, too hard to change, and hard to hire for.

Not anymore. Programmers no longer program. They don't write syntax, they don't read code, and frankly most don't know a thing about security. Vibe-coded projects written in Rust are faster, safer, and more likely to work the first time (due to an insanely complex compiler).

It's the perfect tool for the moment. Only the Rust Foundation could miss this layup.

**I will leave one small caveat here: TypeScript is the inevitable language for agent development, deployment, and distribution, but thats deserves a separate post.

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@pastienot What? Yea I want one that ends with rust as the target since it’s the easiest language without gc. That’s the point lmaoo
We’ve had languages with other backends for a minute.

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

From callback hell in nodejs to tail recursion to goal driven workflows or loops in AI models. This industry just loves to keep reinventing the same things with new names

Wonder what OOP in AI will be called

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@phantom plz fix
All your tokens in the volume page are RH chain tokens and you give me the buy button but it doesn’t work. Just says no liquidity. Plz lemme gamble on robinhood chain too

[ChimeraDefi.eth - muShanghai @ChimeraDefi]

@forloopcodes Lucky if that is all that happens. My agents like to also stomp and revert each others work so half the time I’m like yo wtf happened to the other changes