Its values, its interview process and the questions it asks, in short episodes. Listen on a walk, then practice.
10 episodes1 free to try
No subscription needed
Behavioral signals at Thinking Machines are graded inside technical conversations and center on ownership, small-team collaboration and judgment under uncertainty. This episode maps those themes to the company's own language and shows how to build a 'why here' answer anchored in things you have actually done with Tinker, Inkling and the research.
Thinking Machines does not publish its process, Glassdoor's 'Thinking Machines' page is a different company, and public candidate reports are rare. This episode separates confirmed signals from informed guesses, lays out the likely loop for research and engineering roles, and shares the kinds of technical and behavioral prompts third-party guides say to expect.
With roughly 100-130 people, flat 'member of technical staff' titles and a habit of publishing, Thinking Machines runs very differently from a big tech org. This episode covers the lab's working norms as it describes them, the community it cultivates through grants and the Tinker Cookbook, and the behaviors candidates should demonstrate.
Thinking Machines has a founder with board control, a chief scientist who co-founded OpenAI, a PyTorch creator as CTO, and has lost roughly a third of its founding team to Meta, OpenAI and xAI. This episode gives candidates the full picture of the leadership story and shows how to discuss it maturely if it comes up.
Thinking Machines says 'infrastructure quality is a top priority' and backs it with gigawatt-scale compute deals, a PyTorch creator as CTO, and detailed public notes on how Inkling was trained. This episode goes through the architecture and training choices, Tinker's infra milestones, and the research-engineer job expectations that follow from them.
In May 2026 Thinking Machines previewed 'interaction models,' systems that listen, watch, think and speak simultaneously in 200-millisecond micro-turns rather than taking turns. This episode explains the architecture, why the lab calls interactivity a scaling problem, and how candidates across research, infra and product can connect their experience to it.
Thinking Machines' research blog, Connectionism, has only seven posts, but they are dense, opinionated and widely read, from 'Defeating Nondeterminism in LLM Inference' to 'LoRA Without Regret.' This episode distills the recurring technical themes and the mindset they signal, so candidates can discuss them intelligently and borrow the lab's standard of rigor.
In 2026 Thinking Machines went from infrastructure provider to model builder, releasing Inkling in July and Inkling 2 Small in November under a staged 'safe path to open weights.' This episode covers what was released, how it stacks up against Chinese open models, the compute deals behind it, and how to discuss open-weights strategy in an interview.
Tinker, the fine-tuning API launched in October 2025, is Thinking Machines' only revenue product and the center of its commercial story. This episode traces one year of shipping, the enterprise results the company touts, the reported $5B raise, and what it means for how candidates should talk about customers and business impact.
Thinking Machines Lab says its mission is to build AI that extends human will and judgment, and it has written an unusually philosophical manifesto explaining why. This episode unpacks that thesis, how it differs from OpenAI and Anthropic, and how to use it to show genuine alignment in an interview.