Daniel Bate

probably building

[Kyran | K1 R1 @K1_R1_]

My CTO:

[Aksel @akseljoonas]

Introducing ml-intern, the agent that just automated the post-training team @huggingface

It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem.

It can pull off crazy things:

We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%.

In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%.

For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on https://t.co/udm7xGpNzR, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously.

How it works?

ml-intern makes full use of the HF ecosystem:
- finds papers on arxiv and https://t.co/brvCC7fLPa, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on https://t.co/hrJuRkRyzi
- browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data
- launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains

ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like.

Releasing it today as a CLI and a web app you can use from your phone/desktop.
CLI: https://t.co/l3K1PslZ1n
Web + mobile: https://t.co/orko5srL4H

And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.

[Daniel Bate @danbatedev]

Smoke testing with agent-browser is a cheat code

[Fin Moorhouse @finmoorhouse]

The hyperscalers have already outspent the most famous US megaprojects

[Daniel Bate @danbatedev]

Even without the big AI deals it was ~$112 billion quarter

The time to build is now anon

[a16z @a16z]

Q1 2026 was the largest quarter for venture investment ever recorded.

More charts: https://t.co/OgJQcBSNjy

[Daniel Bate @danbatedev]

@K1_R1_ b2b Thomas Bangalter at Alexandria Palace? Truly music of the people

[Daniel Bate @danbatedev]

If I have to scan my iris just to shitpost on the internet I’m out

[Kyran | K1 R1 @K1_R1_]

Sorry sir your proof of humanity has been declined

The retreat to a more comfortable scale of dysfunction does not excite me

I seek what I have always sought: proof of insight, proof of wisdom

[Daniel Bate @danbatedev]

Segregation isn’t a holistic fix and I’m not handing over my biometrics to any tech or gov platform.

Current incentive structures break in an agentic world. PoH is solving the wrong problem entirely.

[Marc Andreessen 🇺🇸 @pmarca]

It’s time for Proof Of Human.

[Daniel Bate @danbatedev]

Ahoy @K1_R1_

[Dan Shipper 📧 @danshipper]

new model for engineering team structure in 2026:

2 people only

one pirate and one architect

the pirate's job is to move as fast as possible to develop valuable, shipped product features by vibe coding.

the architect's job is to turn the product surface discovered by the pirate into a reliable, structured machine—also by vibe coding, but at a slower, more well-reasoned pace.

every product needs a pirate but most product's only need an architect once they some form of PMF, and in that case they usually don't need one full-time. architects can work across many codebases and solve interesting technical challenges. pirates go hard on a product that they own end-to-end.

[Daniel Bate @danbatedev]

The most important article I've read this year

[0xSero @0xSero]

https://t.co/txDnDa0Flf

[Daniel Bate @danbatedev]

Touching grass (tarmac) + Londonmaxxing

[Kyran | K1 R1 @K1_R1_]

The Londonmax thesis
Monitoring the situation closely

[Jarred Sumner @jarredsumner]

now that I’m no longer doing a startup and won’t for many years, some early stage startup fundraising advice:

don’t spend any time at all with investors until you’re ready. tell them you’re too busy. do not meet with them. yes especially if an associate emails you 5 times

[Daniel Bate @danbatedev]

everyone Londonmaxxing rn but they're all paying their taxes overseas

[Daniel Bate @danbatedev]

@K1_R1_ Might as well be using pen and paper

[Daniel Bate @danbatedev]

After far too long I've finally ditched Notion

Obsidian + git + Claude

This is the way

[Daniel Bate @danbatedev]

@K1_R1_ If it doesn't work with my Claude max subscription then it's getting uninstalled

[Daniel Bate @danbatedev]

@K1_R1_ @bcherny +1

[Daniel Bate @danbatedev]

Building proof of humanity at this stage of the AI timeline feels like a very luddite mindset smh

[Daniel Bate @danbatedev]

>find industry still using fax machines
>create AI wrapper
>add nice UI
>call it "transformative AI"
>billion dollar valuation
>mfw

[Niko Ludwig @Collateral_com]

AI Legal startup Harvey is raising $200M at an $11B valuation

They've reached:
$190M ARR
1,000 customers
100,000 lawyers on platform

Below is their 2025 in review deck: