doing things @AnthropicAI

Joined April 2009
It'd be nice to have an LTT who isn't audience captured by gamers
3
1,059
Its visibly noticeable with the naked eye how many more satellites are in the sky
1
1
1,113
Thomas Millar retweeted
"God gave me a sign" (remembering i live in San Francisco) "I mean, I was acasually influenced by the ASI at the end of time to maximize EV for humanity"
65
31
1
1,091
141,948
Thomas Millar retweeted
life would be awful without tailscale
33
26
1
711
56,439
Parents visited SF for the first time in a decade, they tried Waymo for the first time, then tried FSD in my Tesla, they much preferred the FSD, way more confident of city driver
5
2,040
Thomas Millar retweeted
New Anthropic research: Natural Language Autoencoders. Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read. Here, we train Claude to translate its activations into human-readable text.
581
1,662
615
16,473
2,600,256
Tencent sent hundreds of thousands of requests to my personal site yesterday...? Cost me like $20 in Vercel function time. Anyone else seen this?
2
1
12
3,820
Thomas Millar retweeted
Bryan Johnson and I have a lot in common. He just had sex with Kate. I just got f*d by k8s. We both spent hours looking at the health checks after.
Just had sex with Kate. Goodnight everyone.
65
105
11
3,412
197,558
Thomas Millar retweeted
I've built a new JavaScript runtime that runs inside the BEAM. Every JS runtime is a GenServer with its own OS thread. No JSON anywhere — JS objects map to BEAM terms natively through a lock-free queue. What makes it different from running Node/Deno/Bun alongside Elixir: → JS runtimes live in supervision trees. They crash, restart, recover state — standard OTP → fetch() goes through :httpc. WebSocket through :gun. crypto.subtle through :crypto. BroadcastChannel through :pg — works across a cluster → The DOM is lexbor (C library). JS renders into it, Elixir reads it directly — no serialization, no re-parsing → Workers are BEAM processes. They get preemptive scheduling for free → TypeScript toolchain (OXC) and npm client built in — no Node.js on the machine at all Full control over the JS layer: parse ASTs, bundle imports, transform TypeScript, minify — all from Elixir via OXC NIFs. Use cases: — SSR with Preact/React into native DOM, Elixir reads the tree — Sandboxed user-defined business rules with memory limits, timeouts, and a controlled API surface — Parallel Workers that compute and broadcast via distributed process groups — Evaluating or bundling TypeScript without any external toolchain — Running npm packages inside the BEAM Still a research project in early beta. Covered with tests including Web Platform Tests ports, but expect rough edges. github.com/elixir-volt/quick…
17
68
8
508
36,035
Thomas Millar retweeted
We just shipped Distributed Python on top of the Erlang distribution, with full Elixir and Livebook integration: dashbit.co/blog/distributed-… And much more: intellisense, zero-copy Apache Arrow, and more. Read the article for all the details. A huge thank you to NLnet Foundation for sponsoring our work. They are always looking for new ideas and you have until April 1st to join the next batch: nlnet.nl
14
67
5
346
20,928
I sit next to this team, they work on the coolest stuff my old CTF friends should consider applying, very transferrable skillset
Now is a good time to say I'm hiring @anthropicai for the Frontier Red Team. We need Research Scientists on the biggest issues in model safety, like cyber, autonomy, and agent risks. 2026 is the year. I can promise you your life's work and the most meaningful mission.
2
5
1,529
note to self, never leave @DNAutics alone with your claude
1
2
902
Thomas Millar retweeted
We’re aggressively scaling up the Societal Impacts (SI) team at Anthropic as our models are beginning to have non-trivial impacts on the world.
68
86
34
1,334
161,461
Join Societal Impacts! We work on important problems, we care a lot about our work, and we are “extremely productive” (thanks Jack). DMs open if you have questions. Read the recruiting blog: docs.google.com/document/d/1…
We’re aggressively scaling up the Societal Impacts (SI) team at Anthropic as our models are beginning to have non-trivial impacts on the world.
4
7
3
119
20,952
check out some of the research I worked on at Anthropic We're starting to see autonomy increase dramatically with agents out in the wild. quite revealing to see how much people have shifted from a permissions based to an interrupt based workflow with Claude code @MilesMcCain did an incredible job leading this project.
New Anthropic research: Measuring AI agent autonomy in practice. We analyzed millions of interactions across Claude Code and our API to understand how much autonomy people grant to agents, where they’re deployed, and what risks they may pose. Read more: anthropic.com/research/measu…
2
14
1,603
He did it, he made the tweet. Now do the demo
non elixir minds cannot comprehend just rpc'ing prod and introspecting state/debug inside the running vm. Now attach that to an AI agent (w/ human in the loop) and you have insane capability. "xyz seems bugged, use fly ssh -C ... to poke the release to debug it" No MCP required.
2
11
1,799
You can take 3 layovers and still be in America This continent is endless
1
8
898
The fact you can Tailscale from 40k feet and it works well will never cease to amaze me. Can't wait for more Starlink
4
761
Thomas Millar retweeted
Our teams have been building with a 2.5x-faster version of Claude Opus 4.6. We’re now making it available as an early experiment via Claude Code and our API.
783
729
791
14,248
6,518,171
Thomas Millar retweeted
Very suspicious that OpenAI and Anthropic both dropped major announcements within hours of me publishing this: dashbit.co/blog/why-elixir-b… They want to silence it. Don't let them. Read it twice. Share it with your colleagues and friends.
40
80
14
629
78,020