@julien_c - AI Dynamics

Data Infrastructure Optimization for AI Model Training

By

–

28 May 2026 18h37

We are starting to be quite bullish about getting in the data infrastructure business.

I just cloned 68 TB (while I only have a 4TB local disk) to my @huggingface training bucket in 1 minute 55 seconds, thanks to Xet deduplication and all our infra optimizations.

You can host… pic.twitter.com/qfm9QvaIdj
— Julien Chaumond (@julien_c) 28 mai 2026

We are starting to be quite bullish about getting in the data infrastructure business. I just cloned 68 TB (while I only have a 4TB local disk) to my @huggingface training bucket in 1 minute 55 seconds, thanks to Xet deduplication and all our infra optimizations. You can host

→ View original post on X — @julien_c,

28 May 2026

104M Image-Text Pair Dataset Released on Hugging Face

By

@julien_c

–

28 May 2026 15h47

With 104M of image-text pairs, this is one of the largest, if not the largest, openly-licensed image dataset

And it's on @huggingface!!

Kudos @heyjasperai https://t.co/mTwGfZUzZU
— Julien Chaumond (@julien_c) 28 mai 2026

With 104M of image-text pairs, this is one of the largest, if not the largest, openly-licensed image dataset And it's on @huggingface
!! Kudos @heyjasperai

→ View original post on X — @julien_c,

28 May 2026

YC Bench: AI Agent CEO Simulation Benchmark for Startups

By

@julien_c

–

01 May 2026 13h46

YC Bench by @CollinearAI
: Benchmark for Agents who play CEO of an AI startup for 1 simulated year via CLI tool use against a deterministic discrete-event simulation. Score = final $$ amount achieved by @nazneenrajani and team Also a good opportunity to showcase this recent hf

→ View original post on X — @julien_c,

1 May 2026

Encouragement to Contribute Upstream to llama.cpp

By

@julien_c

–

01 May 2026 11h26

Don't hesitate to upstream to llama.cpp

→ View original post on X — @julien_c,

1 May 2026

Qwen3 27B Runs Locally Rivaling Claude Opus for Coding Tasks

By

@julien_c

–

24 April 2026 14h03

This is where we are right now. And i’m not gonna lie it feels pretty magical Qwen3.6 27B running inside of Pi coding agent via Llama.cpp on the MacBook Pro For non-trivial tasks on the @huggingface codebases, this feels very, very close to hitting the latest Opus in Claude

→ View original post on X — @julien_c,

24 April 2026

HuggingFace CLI Invocations by Coding Agents

By

@julien_c

–

22 April 2026 22h07

It’s only invocations of the hf CLI by the coding agents. So it’s really only a partial data point.

→ View original post on X — @julien_c,

22 April 2026

Coding Agents Use HuggingFace CLI for Invocations

By

@julien_c

–

22 April 2026 22h07

It’s only invocations of the hf CLI by the coding agents. So it’s really only a partial data point.

→ View original post on X — @julien_c,

22 April 2026

Claude Code, Codex, Cursor: Comparing AI Model Sizes

By

@julien_c

–

22 April 2026 17h54

Currently: – Claude Code is 4x bigger than Codex
– Codex is 2x bigger than Cursor
– Antigravity is almost as big as Cursor (but probably just because all Googlers use it? ) We'll add tagging for other agents asap. (Source: one data point from the @huggingface Hub team. Your

→ View original post on X — @julien_c,

22 April 2026