Understand them deeply.
Everything our team needs to be aligned on Thinking Machines from day 0 — who they are, what they've shipped, the science they publish, and how they intend to make money. Every claim links back to their own writing or reputable press.
Public information only. Reported figures are labeled; nothing here implies a relationship.
The facts, cited.
“We're building a future where everyone has access to the knowledge and tools to make AI work for their unique needs and goals.”— their mission, thinkingmachines.ai ↗
| Founded | February 2025, San FranciscoWikipedia ↗ |
| Founder & CEO | Mira Murati (former OpenAI CTO)Wikipedia — Mira Murati ↗ |
| Chief Scientist | John Schulman (OpenAI co-founder)Connectionism ↗ |
| Funding | Reported ~$2B at a ~$12B valuation, led by Andreessen Horowitz (mid-2025); reported investors include Nvidia, AMD, Cisco, and Jane StreetreportedTechCrunch ↗ |
| First product | Tinker — a fine-tuning API (launched Oct 1, 2025)Announcing Tinker ↗ |
| Research blog | Connectionism — "shared science from the team"Connectionism ↗ |
Tinker — teach a model your own expertise.
A managed API for fine-tuning open-weight language models. It runs distributed training on their clusters while deliberately keeping the low-level knobs in the user's hands — using LoRA to share compute efficiently.
The primitives: forward_backwardoptim_stepsamplesave_state
Who it's for: Researchers and developers who want full control over algorithms and data without managing GPU infrastructure.
Access: Launched Oct 1, 2025 in private beta, free to start, with usage-based pricing to follow.
Connectionism — shared in the open.
| Date | Post | What it shows |
|---|---|---|
| Jul 10, 2026 | The Future Worth Building Is Human ↗ Thinking Machines | The vision: distributed, customizable AI shaped by human knowledge, will, and judgment — not a single centralized model. |
| May 11, 2026 | Interaction Models: A Scalable Approach to Human-AI Collaboration ↗ Thinking Machines | Real-time multimodal collaboration — handling audio, video, and text in time-aligned micro-turns instead of turn-by-turn. |
| Oct 27, 2025 | On-Policy Distillation ↗ Kevin Lu et al. | Train a student on its own outputs graded by a teacher — reportedly matching RL results at a fraction of the compute. |
| Sep 29, 2025 | LoRA Without Regret ↗ John Schulman et al. | LoRA can match full fine-tuning when applied across all layers within capacity — the science behind Tinker. |
| Sep 26, 2025 | Modular Manifolds ↗ Jeremy Bernstein | A framework for constraining weight matrices to manifolds, with principled per-layer learning rates. |
| Sep 10, 2025 | Defeating Nondeterminism in LLM Inference ↗ Horace He et al. | Identifies batch-size variation as the true cause of nondeterministic LLM outputs, with batch-invariant kernels for bitwise-identical results. |
The through-line of their work.
Reproducible, deterministic inference
Batch-invariant kernels so an LLM returns bitwise-identical results — reliability as a first-class research goal.
Defeating Nondeterminism ↗Efficient fine-tuning anyone can do
LoRA done right, so a small team can teach a model their own expertise without full-scale infrastructure.
LoRA Without Regret ↗Cheaper, stronger post-training
On-policy distillation that reportedly matches reinforcement learning at a fraction of the compute.
On-Policy Distillation ↗Real-time, multimodal collaboration
Interaction models that listen while they talk — audio, video, and text in time-aligned micro-turns.
Interaction Models ↗Customizable AI, shaped by people
A future where AI is distributed and adapted by the people who use it, guided by open science.
The Future Worth Building Is Human ↗They publish the code, not just the paper.
Public facts, cited.
- Thinking Machines Lab — official site ↗
- Connectionism — their research blog ↗
- "The Future Worth Building Is Human" (Jul 10, 2026) ↗
- Announcing Tinker (Oct 1, 2025) ↗
- Tinker — product ↗
- Thinking Machines Lab — GitHub ↗
- Wikipedia — Thinking Machines Lab (founding & funding) ↗
- TechCrunch — $12B seed (Jul 15, 2025) ↗
This is a public study and an open, honest invitation from 🤫 hushh — an admiration of Thinking Machines' public work and a proposal to explore a partnership. It is not a claimed deal, affiliation, endorsement, or sponsorship. Every fact here is drawn from Thinking Machines' own public writing or reputable press; figures that are press-reported are labeled "reported", and quotations are verbatim from the cited source. Companies and facts change — verify against the primary sources before acting.