Alex Zverianskii

ML @withmartian • 3x founder (1 exit)
ex Principal MLE @gett
15+years of NLP & recsys

[Alex Zverianskii @alexML]

@postimortem Imagine how slow it would be to train models to actually use search during token hallucinations

[Modular @Modular]

Mojo 🔥 1.0 is in beta! Beta 1 marks the first step towards finalizing 1.0 later this year, which will bring a new level of language stability. The beta lands safe closures with a new capturing syntax, conditional trait conformance, and major variadic improvements. Plus, Mojo has its own home at https://t.co/j3TzPUkEDr.

[Alex Zverianskii @alexML]

@mitchellh Classic—that’s how I ended up with “gut status” as an alias many years ago

[Alex Zverianskii @alexML]

Now I have something interesting for the Sunday run. Thanks!

[CodeRabbit @coderabbitai]

Think OpenAI and Anthropic own the AI game?

Open-source models already handle 50%+ of real production requests!

New pod episode with our own Erfan & @shriyashku from @withmartian on the DeepSeek moment, vendor lock-in risk, and the fine-tuning cheat code.

[Alex Zverianskii @alexML]

@mwilliammyers Bots use the product!
ENDLOOP

[Alex Zverianskii @alexML]

@jhoysbou That's my favorite kind of PR!

[Alex Zverianskii @alexML]

It’s time for factory-level quality control and ISO 9001 compliance!

[Martian @withmartian]

The software factory is already here.
We're seeing bots write code, bots review it, and humans reduced to dispatching the next tool in the chain.

Using 500k+ PRs from Code Review Bench, we looked at one question: can the human leave the loop yet?

[Alex Zverianskii @alexML]

@withmartian 52.8% of AI code reviews get zero human follow-up. we've accidentally built RLHF with no HF!

[Alex Zverianskii @alexML]

@AshleyZhang110 Are you going to yell at @Narmeen29013644 , the maintainer of the k-steering repo?

[Alex Zverianskii @alexML]

@jhoysbou The words of a man, not a boy!

[Alex Zverianskii @alexML]

@withmartian Environmental degradation is a great example for steering LLMs!

[Alex Zverianskii @alexML]

Controlling LLM inference has never been easier thanks to @Narmeen29013644 k-steering. And the name is just as brilliant as all those other game-changers: k-means, k-ubernetes, k-pop!

[Martian @withmartian]

We’re open-sourcing a new tool to control how LLMs behave: k-steering. In just 10 lines of code, you can control multiple aspects of LLM behavior at the same time without any fine-tuning or prompt engineering.

Here's how 👇

[Alex Zverianskii @alexML]

@thejacobclyne It looks a lot nicer than vertical scaling.

[Alex Zverianskii @alexML]

@shriyashku Out Of Evolution

[Alex Zverianskii @alexML]

Now I see why no code review company is called "CodeMonkey”

[Martian @withmartian]

Traditional Precision-Recall curves tell you how your code review tool performs on static benchmarks.

They don't tell you how it performs against a Hawk.

Introducing Fight Index (FI).

[Alex Zverianskii @alexML]

@shriyashku @withmartian

[Alex Zverianskii @alexML]

@withmartian This turnaround is definitely faster than what @shriyashku usually takes for a review, so I’m all for AI.

[Alex Zverianskii @alexML]

@joshgreaves_ml Well, now you can rest assured that every typing error will be caught!

[Alex Zverianskii @alexML]

@rnsmith49 Most importantly, we don’t want to end up in a reality where humans are stuck in a slow loop. "Strange loop" is fine btw