Charlie Marsh✓

@charliermarsh

@OpenAI. Building Ruff, uv, ty, and other high-performance Python tools with the @astral_sh team.

992 Following53K Followers9.7K Posts
  1. Charlie Marsh✓@charliermarsh
    We built tar-codec: a fast and secure tar parser for the Rust ecosystem. It's roughly twice as fast as our previous tar implementation, but with much lower exposure to parser differentials and other security risks. Here's the state machine btw.
    Download
    1131173.8K
  2. Charlie Marsh✓@charliermarsh
    Posts @charliermarsh
    All credit to William Woodruff on this one. We're going to continue building what we view as fast, secure foundational crates for the ecosystem.
    106472
  3. Charlie Marsh✓@charliermarsh
    Posts @charliermarsh
    005449
  4. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @TaKOBKi: My latest rustc PR introduces adaptive sparse encoding for metadata tables. Rustc-perf results show a 20.18% average reduction in crate metadata size.
    010138
  5. Charlie Marsh✓@charliermarsh
    I'm seeing benchmark methodologies you wouldn't BELIEVE in some of these agent-authored projects
    12012110.2K
  6. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @steipete: I spent some time on the weekend to test bun's AOT compiler and results are nothing short of amazing. Performance up 2-3x, memory use less than half.
    Jarred Sumner @jarredsumner
    Bun AOT also introduces `--trim` which deletes unused parts of Bun from the compiled binary On macOS, a binary that uses Bun.serve’s HTTP server goes from 89 MB to 25 MB
    0240157
  7. Jarred Sumner✓@jarredsumner
    Bun AOT also introduces `--trim` which deletes unused parts of Bun from the compiled binary On macOS, a binary that uses Bun.serve’s HTTP server goes from 89 MB to 25 MB
    Download
    55391.4K93.6K
  8. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @polynoamial: Only 2.5 years ago, AI labs would highlight their frontier LLM performance on grade school math benchmarks (GSM8K). I can't wait to see where we'll be in another 2.5 years.
    Aashish Reddy @_AashishReddy
    It just never gets old
    0520168
  9. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @wtgowers: For those who don’t know, ETH is the hypothesis that 3SAT needs exponential time. Very surprising and consequential that it is false.
    Ryan Williams @rrwilliams.bsky.social @rrwilliams
    An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
    01110194
  10. Ryan Williams @rrwilliams.bsky.social@rrwilliams
    An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
    2716646.5K3.1M
  11. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @danintheory: We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
    01K0200
  12. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @mweinbach: GPT 6.1 Terra is out! I had early access, it's a really model for writing and basic use, as well as computer use. It's quick and overall a great value for what it is. It also works extremely well as a subagent.
    080208
  13. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @OpenAI: GPT-6.1 Terra: delivering fast and capable computer use at a lower price. GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
    02.4K0145
  14. Charlie Marsh✓@charliermarsh
    SF this week with the Astral team ✌️
    30645.6K
  15. Charlie Marsh✓@charliermarsh
    That frontier models can find infinite bugs in any piece of software has shaken me and my naivete
    50555228.4K
  16. Charlie Marsh✓@charliermarsh
    Be careful what you read on this site, they let you post whatever you want around here
    1571668.5K
  17. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @gdb: “On a technical level, the paper is unbelievable, maybe beyond what a human could have come up with but not beyond human understanding”
    Scott Armstrong @scottnarmstrong
    Since I'm now blocked by bluey from replying to this thread, I answer here by quote tweet the openAI construction has more to do with convex integration than anything else. In convex integration, the oscillations are typically built by hand and the iteration procedure creates wild weak solutions. The non-uniqueness construction of Albritton–Brué–Colombo used an unstable vortex to split solutions apart. The Córdoba–Martínez-Zoroa singularities amplify vorticity across scales by strain. In the openAI paper, a fluid instability is clever made so that the flow itself generates the oscillations: they grow by drawing energy from the background shear. The Reynolds stress then becomes the force that creates the collapsing smooth vortex. Using a physical instability to create the stress in this way has no precedent, and it is what produces a smooth force rather than a merely rough one. Some on these ideas are in the literature (e.g. Leibovich-Stewartson 1983) but the way they are combined is definitely insane and definitely nowhere in the literature previously. I also don't know of anyone suggesting that convex integration would play a role in the creation of a blow-up solution like this. On a technical level, the paper is unbelievable, maybe beyond what a human could have come up with but not beyond human understanding: there are several different groups of pde experts working on understanding it who say that they more or less understand now. I expect digestions soon td;lr: anyone saying that this OAI result is unsurprising is bullshitting sorry!
    0810171
  18. Scott Armstrong✓@scottnarmstrong
    Since I'm now blocked by bluey from replying to this thread, I answer here by quote tweet the openAI construction has more to do with convex integration than anything else. In convex integration, the oscillations are typically built by hand and the iteration procedure creates wild weak solutions. The non-uniqueness construction of Albritton–Brué–Colombo used an unstable vortex to split solutions apart. The Córdoba–Martínez-Zoroa singularities amplify vorticity across scales by strain. In the openAI paper, a fluid instability is clever made so that the flow itself generates the oscillations: they grow by drawing energy from the background shear. The Reynolds stress then becomes the force that creates the collapsing smooth vortex. Using a physical instability to create the stress in this way has no precedent, and it is what produces a smooth force rather than a merely rough one. Some on these ideas are in the literature (e.g. Leibovich-Stewartson 1983) but the way they are combined is definitely insane and definitely nowhere in the literature previously. I also don't know of anyone suggesting that convex integration would play a role in the creation of a blow-up solution like this. On a technical level, the paper is unbelievable, maybe beyond what a human could have come up with but not beyond human understanding: there are several different groups of pde experts working on understanding it who say that they more or less understand now. I expect digestions soon td;lr: anyone saying that this OAI result is unsurprising is bullshitting sorry!
    32991.2K422K
  19. Charlie Marsh✓@charliermarsh
    Charlie Marsh
    RT @thenanyu: By the time it shows up in the data or UXR interviews it’s too late. The damage has been done and the reputation tarnished
    Charlie Marsh @charliermarsh
    Caring is the moat, you should feel visceral pain when things are bad or broken and stop at nothing to fix them
    080232
  20. Charlie Marsh✓@charliermarsh
    Caring is the moat, you should feel visceral pain when things are bad or broken and stop at nothing to fix them
    602542.3K108K
  21. Charlie Marsh✓@charliermarsh
    I think this is like indisputably true for rewrites now. For new things, I think you still have to choose between something like: (1) a bottleneck on human judgment, or (2) give up control, stop caring about any of the details (may not even realize this is happening).
    Jarred Sumner @jarredsumner
    a growing part of how I feel about bun is this whole category of software - linters, formatters, package managers, type checkers, test runners, bundlers this used to be hard software to build and now it can be prompted into existence in a day or two
    311149151.1K
  22. Jarred Sumner✓@jarredsumner
    a growing part of how I feel about bun is this whole category of software - linters, formatters, package managers, type checkers, test runners, bundlers this used to be hard software to build and now it can be prompted into existence in a day or two
    87251.4K128K
  23. Charlie Marsh✓@charliermarsh
    Posts @charliermarsh
    Honestly I'm still very conflicted on (1) and (2), I think a lot of amazing software will be built with (2), but it's truly a blackpill
    20553.6K
  24. Show more posts