Denis Yarats' AutoJev: an agent swarm autonomously trained a model in 20 hours for $3.1k
NirantK · x · 2026-09-22
Denis Yarats shared a weekend experiment, AutoJev: using his team's internal agent swarm to fully autonomously train a competitive Jev-like model — and it "gets quite far." The swarm (mix of astra and fable, with sol and luna for synthetic data and filtering) ran on a single H200 devbox for 20 hours, costing $3.1k total. A notable demo of multi-agent systems taking over the full training pipeline end to end.
Related event: Agent Swarm Autonomously Trains Competitive Model in 20 Hours for $3,100(2 posts)→
More from coding & agent
- Dell expands AI Data Platform: agents have a data-context problem, not a model problem — DavidLinthicum · 2026-11-12
- Dev's favorite AI use case: Claude monitors competitors and emails alerts — doooyle · 2026-11-12
- LLM controls vocal synth to perform Ave Maria in 4-part harmony — AAAzzam · 2026-11-12
- ArcGrav: a cross-model Agent Control Language to cut token overhead — ArchitecturalGravity · 2026-11-12
- Claude watches competitors around the clock and emails me what matters — GlenBradley · 2026-11-12
- Developers say Gemini 4 Carbon is notably better at favicon design and micro-interactions — AI_Andrew · 2026-11-12