The Extended Brief
Inside the Model Factory — Eiso Kant, Poolside AI
Brief by The AI News AI newsroom · Jul 30, 2026, 6:33 PM EDT edition
Original reporting by Latent Space · published Jul 23, 2026, 1:09 AM EDT
Poolside’s 'Model Factory' approach of running 20,000 experiments a month with agents modifying training pipelines reveals the new operational baseline for competitive model development.
Key points
- Poolside AI released Laguna S 2.1, outperforming Thinking Machines models nearly ten times larger.
- Poolside AI recently secured a $500 million funding round.
- Their Model Factory accelerates development from pre-training to release in just eight weeks.
- The system executes up to 20,000 monthly experiments using streaming data and low-precision compute.
- Co-founder Eiso Kant spent $12 million early on developing code-focused language models.
From the source
“So I would say that our view from very early on in the company was that model building is ultimately 90% engineering.”
“We’re less than 70 researchers, another 35 engineers.”
“and we are running, I haven’t checked the latest count, but far more than 10,000, maybe 10 to 20,000 experiments a month that we cut.”
“The model that we’re gonna talk about today was eight weeks from start of training, to launch.”
“but I think it all just came down to one thing, and I’ll stop the monologue, is the fact that I rather live in a world that has 100 foundation model companies than a world that has five, even if I was one of the five.”
Practical applications
- Model teams can benchmark Laguna S 2.1 against much larger models to verify the claimed efficiency advantage on their tasks.
- Research leads can compare their own experiment throughput against Poolside's reported 20,000 monthly experiments to gauge infrastructure gaps.
- Teams building training platforms can study the Model Factory pattern — streaming data, low-precision compute, agents modifying pipelines — as a target architecture.
- Investors and strategy teams can factor the eight-week pretraining-to-release cycle into assumptions about how fast well-funded labs can iterate.
Who should care
ML researchers, training-infrastructure engineers, and lab leadership deciding how much to invest in experiment automation and training-pipeline tooling.
Context
Competitive model development increasingly hinges on how fast a lab can run and learn from training experiments, not just on raw compute. Poolside AI, a code-focused lab that recently raised $500 million, describes a 'Model Factory' that compresses pretraining-to-release to eight weeks and runs up to 20,000 experiments a month using streaming data, low-precision compute, and agents that modify training pipelines. Its new Laguna S 2.1 reportedly beats Thinking Machines models nearly ten times larger, and co-founder Eiso Kant says he spent $12 million early on developing code-focused language models.
What to watch
- Independent evaluations of Laguna S 2.1 confirming it outperforms Thinking Machines models nearly ten times larger.
- Whether other labs publicly adopt factory-style automated experimentation, making high experiment throughput the industry norm.
Editorial score 4.0 / 5 · significance 3.5 · novelty 4.0 · edge 3.5 · perspective 5.0
Desks: Research · Engineering · Tags: research, models, tooling
Evidence basis: Reviewed from a feed excerpt
This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.