jeffreysmith
Joined 11 years ago
Karma
150
Submissions
70
Member since
Mar 16
Recent submissions
Comment · 14 days ago
I went around and asked a bunch of them with a similar idea (see the above comment of mine), and you'd be surprised to hear that they all hated the idea. They mostly just don't want to deal with anything where they fear that the person vibed the submission. Just no desire to engage with at all, because they've got so many bigger problems with the avalance of code coming their way.
Comment · 14 days ago
I think it's a real quote, but from Karpathy, not Sunny T.
Comment · 14 days ago
We built something quite similar 6 months ago called PEAK (Proof of Effort And Knowledge). Blog post and YT video here: https://substack.com/@jeffksmithjr/p-195763168
Our version has its own TUI and was intended for semi-adversarial scenarios like OSS contribution and enterprise. Thus a lot of anti-cheat tech under the covers (including invalid questions).
My guess is that a lot of folks have built something in this vein. It really made sense earlier in the coding agents cycle. These days, I dunno. Seems like we need to more fundamentally rethink the relationship between coding agents and human understanding of the implementation.
Comment · 4 months ago
Super cool work. I love seeing this direction taken all the way to hardware.
I'm a big fan of KANs. The really seem like the start of something big and new. We've got a couple of papers out and in the works on KANs. The most relevant to OP's is this one: https://arxiv.org/abs/2512.15742v2
And we just put up a general primer on KANs on YT: https://youtu.be/wgcSsJ69x1c?si=fiUl1YGTgaTt_bn9 Fun stuff if you want to get into the weeds.
And if you are really interested in KANs, you should really check out Ziming (KAN creator)'s blog: https://kindxiaoming.github.io/blog/
Comment · 4 months ago
I played a somewhat unusual role in this whole story. I was the guy who acquired in the original Papers with Code and managed them after they joined Facebook/Meta.
It was super sad to see FB/M abandon the original mission of what PwC was building towards and let the original community resource rot. During the good times, we always talked about how PwC related to HF. So, I think there is a sort poetry to PwC winding up as part of HF, where they probably always belonged. No company is perfect, but HF has been a better than average steward of open source and community resources.
For the younger folks on this thread, you probably have no real feel for just how frustratingly inefficient AI/ML research used to be before people like Robert and Ross of PwC came along to start to bring structure, sanity, and reproducibility to the information needed to work of this kind. And of course, Clem, Julien, and Thomas of HF kicked off an even bigger effort to tame the previously scattered workflow of open AI research into some sort of sane stack.
It's clear that, in 2026, what PwC could be is something much more evolved than what we were able to do back in the day. LLMs + PwC is a huge design space. I hope nielz_r and friends at HF are able to make something truly useful for the community. AI research has both gotten way easier and much harder. e.g. We have a Fable, but Anthro won't let us use it forward our science. Community resources for research are still very much needed.
Best of luck Son of PwC. May you thrive.
Comment · 6 months ago
Burnin (London or NYC) | Founding Research Scientist + Founding Engineering Lead | Hybrid (Some on-site) | Full-time
We are building the trust stack for AI code generation, targeted at high-stakes computing where wrong numbers cost real money: a statically typed functional language designed for coding models, a compiler whose guarantees double as audit-grade trust infrastructure, and a coding model fine-tuned on the language with compiler fitness as the training signal. Language and model are co-designed.
Small team out of PyTorch, FAIR, and Meta. Strong institutional VC support from our pre-seed; active investor interest heading into seed.
Founding Research Scientist: own the agenda across the language and the model. Type system design, compiler analyses that produce useful fitness gradients, and fine-tuning open-weight coding models on a language with no pretraining footprint. Publish in PL and ML venues. Strongest fits cross between machine learning, programming languages, and formal methods. PhD preferred, equivalent output equally fine. Stack is Rust, Lean 4, Python.
Founding Engineering Lead: own engineering across compiler internals, language runtime, GPU backends, notebook and library tooling, and the AI infrastructure around training, evaluation, model release, and likely public inference. We already support x86 and ARM, CUDA and AMD, macOS and Linux, and the matrix grows. Real feel for statically typed functional programming expected; the kind of engineer who picks up Lean or Haskell on a weekend because they wanted to. Have led engineering before, formally or not. Stack is mostly Rust with Python where it earns its place, primarily on AWS.
Both roles: comfortable with early-stage ambiguity, define your own roadmap, defend it with evidence.
Apply: RS: https://wellfound.com/l/2Carrr Eng Lead: https://wellfound.com/l/2CewDC
Socials: YT: https://www.youtube.com/channel/UC4DLS_emqwKXO7A9qaOtxlw SS: https://burninai.substack.com/
Comment · 7 months ago
OP here. This is the sequel to the original Good Egg post (https://news.ycombinator.com/item?id=47151678).
We ran three experiments since v1 that changed the model substantially.
We tried to detect suspended GitHub accounts from behavioral signals (merge rate, network centrality, TF-IDF on PR titles, LLM classification with ~31K Gemini calls). Best individual AUC was 0.619 on a 1.9% base rate. The merged-PR population is too homogeneous. Accounts that pass code review look like everyone else. The interesting finding: the suspension rate among contributors with merged PRs is under 2%. The review process is a better filter than the discourse around AI slop suggests.
That led us to question the scoring model. The graph score (bipartite construction, personalized ranking, language normalization, the whole pipeline from v1) actively hurts predictions for the contributors who actually need scoring: unknown people with a handful of merged PRs. Merge rate alone outperforms merge rate plus graph at every tier we tested. The new default model is merged / (merged + closed). We also pulled account age out of the score into a separate advisory after DeLong tests showed it adds nothing once you condition on merge rate.
The post has the full data, including the tables.
Next we're working on content scoring (does this PR fit this repo's conventions?) and cold-start tooling (helping new contributors understand project expectations before they submit). Contributor reputation is one input to review triage. The PR itself carries more signal.
Repo: https://github.com/2ndSetAI/good-egg
pip install good-egg
Or just run it via uvx.
Comment · 7 months ago
They definitely are a powerful option for smaller scale runs. Very much optimized to have the unit economics and turnaround time work for smaller brands.
I don't really know the answer around supplying your own yarn. I'd assume that's the abnormal case, but just a guess.
Comment · 7 months ago
I interviewed these guys for an article on the use of seaweed in yarn and fabric. And I bought the 3D knit seaweed sweater. Great team, with a lot of heart and good intentions.
I'm also a hand knitter, and I don't really see any conflict between what they're doing and hand knitting. The grist of the yarn that you use as a hand knitter is generally much thicker than these machines commonly use. Commercial 3D knitting machines can do all of the stretchy, thin, and light stuff that the modern wardrobe is built around.
As folks note, this technology was really pioneered by Shimaseki's work in Japan just decades ago. What OC and the similar Brooklyn-based Tailored Industry are really innovating on is the business model and connection to production process. Folks like this are really serious about not producing all of the waste that comes with most fashion production processes, and it shows up at several levels of the stack.
For the HN crowd, TI's platform gives you more of a sense of why this sort of tech is really like the cloud for knitwear: https://tailoredindustry.com/platform
Really a fascinating part of the global fashion production world, and one we would all benefit from seeing grow.
Comment · 8 months ago
And I filed that suggestion as an enhancement issue on the repo: https://github.com/2ndSetAI/good-egg/issues/43
Thanks for the idea.
Comment · 8 months ago
Thanks!
Yeah, scanning non-GitHub is on the roadmap and really should be done. I expect there would be value in understanding all of the current GitHub competitors. And I think the forecasts of new GH competitors getting launched (likely by AI companies) will become relevant in the near future.