How Maxime thinks
Reject monolithic models
Rejects monolithic frontier models in favor of intelligence that is cheap to meter everywhere.
For you: Consider how you can design AI that stays affordable at scale.
Agentic efficiency
Shows that agentic models can be extremely fast and competitive with much larger models.
For you: Think about leveraging agentic approaches to boost performance without scaling size.
Decision models value
Highlights decision models as a useful piece of the AI stack, offering speed over traditional LLMs.
For you: Explore where decision models could speed up your applications.
Ask this first
- How do you keep agentic models fast while staying competitive?
- What motivated the shift to cheap‑to‑meter intelligence?
- Can you explain the benefits of decision models over traditional LLMs?
Maxime's brain
Recently updated
In mid‑October 2026 I announced LFM2.5‑24B‑A2B, the largest member of the LFM2.5 family, positioned as an efficient local model for laptops…
In September 2026 I also outlined a European Transformative AI strategy. The plan calls for a three‑year political compute pre‑commitment…
From October 1‑9 2026 I posted on X (formerly Twitter) about the open‑weight d1 decision‑model family, highlighting releases, performance…
In September 2026 I shared a series of LinkedIn updates covering my move back to the office in San Francisco, progress on MiMo‑V2.6‑Pro…
In early October 2026 I posted a series of LinkedIn updates announcing new tiny multimodal models, sharing my upcoming panel at the Science…
I maintain a concise record of my scholarly output, professional service, recognitions, language skills, and intellectual property. This…
I chose Node.js for ToldYa to leverage its asynchronous I/O for blockchain API calls.