Atlas

The people inside Anthropic, OpenAI, Google DeepMind, and Meta AI building Claude, GPT, Gemini, and Llama. Andrew Lampinen and Jasmijn Bastings on what DeepMind's models can and can't generalize, Sam Bowman at Anthropic on alignment work, Sara Hooker on whose research gets cited, Soumith Chintala on Meta AI infrastructure. When this neighborhood stops posting about a release, that silence often carries more signal than the launch.

karpathy Jacob Eisenstein David Pfau Sara Hooker Charles Sutton Andrew Lampinen Jasmijn Bastings Sam Bowman

The people who write the rules and audit the labs. Meredith Whittaker at Signal on encryption versus surveillance, Emily Bender on what "reasoning" actually means, Abeba Birhane on power and accountability, Arvind Narayanan calling out AI snake oil. They fact-check the press releases, translate model cards into legislation, and feed evidence to the journalists who follow. One post here crossing a thousand reposts usually means a policy fight is days away.

Margaret Mitchell Meredith Whittaker Emily M. Bender Arvind Narayanan Abeba Birhane Dr Sasha Luccioni Thomas Dietterich Deb Raji

The largest community in the directory, by far. University faculty whose papers the labs cite in keynotes years later. Melanie Mitchell at Santa Fe on what models actually understand, Mark Riedl at Georgia Tech on AI storytelling and safety, Christopher Manning at Stanford on the NLP stack that became the LLM stack, Hal Daumé III at Maryland on fairness in ML. The frontier labs publish faster; this neighborhood remembers longer.

Melanie Mitchell Mark Riedl Naomi Saphra Maria Antoniak Hal Daumé III Eugene Vinitsky Christopher Manning Sasha Rush

The people shipping AI products outside the frontier labs. Jeremy Howard at Answer.AI on practitioner-first ML and fast.ai, Thomas Wolf at Hugging Face on the model hub that hosts everyone else, Simon Willison on Datasette and his LLM CLI, Sebastian Raschka on the textbook everyone learns from. Indie devs, founders, and infra engineers who got burned enough times to know which lab claims to discount.

Zachary Lipton Jeremy Howard Thomas Wolf Jess Hamrick Simon Willison Simon Willison Vicki Boykis Jay 🦋

The cross-cutting ML research community outside the frontier labs. Nathan Lambert at Ai2 on what post-training really does (and writes Interconnects), Stella Biderman at EleutherAI on open model science, Yoav Goldberg at Bar-Ilan on NLP, Luca Soldaini and Kyle Lo on open data. Less institutionally anchored than academia or the labs; more likely to start a methods debate that lasts a week.

Nathan Lambert Yoav Goldberg Luca Soldaini 🎀 Stella Biderman Valentina Pyatkin Kyle Lo Ida Momennejad Gaël Varoquaux

The reporters covering the labs and the substack writers who outscoop them. Kashmir Hill at NYT on AI's privacy fallout, Casey Newton at Platformer on the platform/AI overlap, 404 Media on the stories the labs don't want surfaced, Brian Merchant on the labor angle, Taylor Lorenz on the AI-political astroturfing beat. They turn a 60-page paper into a sentence the rest of the directory then argues about.

Casey Newton Meredith Broussard Molly White Will Oremus 404 Media sarah jeong Mike Masnick Brian Merchant

The connective tissue. Rodney Brooks on robots being harder than you think (forty years running), Anil Dash on tech accountability, Hypervisible on labor and surveillance, Mar Hicks on the computing history we already lived through. Plus the institutional accounts NeurIPS, ICLR, and Hugging Face that every other neighborhood follows. Voices that don't fit one cohort but get reposted by all of them.

Chris Albon NeurIPS Conference Hugging Face Mar Hicks Catherine Breslin Rodney Brooks ICLR Conference Ethan Zuckerman