Sholto Douglas
@_sholtodouglas
Scaling RL @AnthropicAI, ex @DeepMind - working towards intelligence too cheap to meter
Very excited to announce that I joined Anthropic at the start of the month! Everything I'm seeing says that we are on trend for AGI in 2027. If the trend lines continue - then we have a shot at building a radically better world, but the path there is fraught with very real…
Taking bets on eval introduction -> eval saturated time? It’s a great eval - time for us all to solve it
back in May, @_sholtodouglas said on the @dwarkesh_sp pod: "there's no SWE-bench for tax" ...well, we were already working on it! and now there is!
It’s not every city where you can bump into a friend at a coffee shop, and they say “So we think we cracked alchemy recently…” Kudos @kylecschiller. We’ll make golems, while you transmute gold.
marathonfusion.com/alchemy.pdf
Fantastic post. Useful intuitions on both why we discovered the major drugs of the past in the order we did, and why we should be quite hopeful for the next decade.
why don’t we have more medicines? inventing medicines has grown less efficient, limited by our knowledge of biological targets for the 1st time, AI & genomics have the potential to make this knowledge abundant – one of the largest potential impacts of the intelligence age.
Also notable that both the peptides and the robots are manufactured in China - we’ve forgotten how to build things!
“Where did you meet the bootleg peptide dealer?” “At the underground robot boxing ring” Visceral future-shock
“Where did you meet the bootleg peptide dealer?” “At the underground robot boxing ring” Visceral future-shock
Over 500M Artifacts have been created since we launched! And today, your Artifacts can now embed Claude's intelligence in them; we call it "Claude in Claude" internally and it's been a blast to see what people create with it.
Introducing two new ways to create with Claude: A dedicated space for building, hosting, and sharing artifacts, and the ability to embed AI capabilities directly into your creations.
My job? I'm a rare token hunter. I track down dead languages in Tibetan monasteries, decrypt Tesla's private journals, chase whispers of pre-contact Amazonian dialects. The AIs pay top credit for tokens they've never tasted, you know. Work is work, even if it's for the machines.
Anthropic purchased millions of physical print books to digitally scan them for Claude
Thrilled to be sharing an important milestone towards bringing humanoid robots to the home: 1X AI has trained Redwood, a VLA capable of end-to-end mobile manipulation tasks like retrieving objects for users, opening doors, and navigating around the home. Here’s some of its…
Redwood NEO’s AI
Claude Code now integrates more deeply with VS Code and JetBrains IDEs. Claude Code can see your open files, LSP diagnostics, and any text you highlight. When Claude Code proposes changes, it'll show you the edits in the editor, allowing you to work seamlessly in your favorite…
🎁 Weekend Special We picked some of the best moments on @tbpn this week, put them in a single recap episode to catch you up to speed. (00:00) Week in review (02:57) Ukraine's historic drone attacks on Russia (03:21) Commentary from @soren_ma, @ConnorLoveCA, and…
Extremely fun joining @tbpn yesterday!
"Building X would take human months, but if you have fast and good enough AIs, you could do it in several hours." - @_sholtodouglas, research @AnthropicAI "True creativity is much harder to replicate than analytical reasoning." "The AI models are actually pretty good at making…
Morning. Here are our guest call-ins today: - @shaunmmaguire (Sequoia) - @jackgwhitaker (Stanford) - @AarushSelvan (Google) - @olivercameron (Odyssey) - @mignano (Lightspeed) - @markchen90 (OpenAI) - @_sholtodouglas (Anthropic) - @eshear (Softmax) - @jeff_weinstein (Stripe) -…
TBPN | Thursday, June 5th x.com/i/broadcasts/1…
"I look at [DeepSeek's] work, and I'm like, 'Ah, a kindred soul'." Anthropic researcher @_sholtodouglas and @TrentonBricken explain how DeepSeek papers reveal good research taste. "They very clearly understand this dance between hardware systems and algorithms." More technical…
"We finally have proof of an algorithm that can give us expert human reliability and performance, given the right feedback loop. We haven't yet demonstrated long running agentic performance. But you're seeing like the first stumbling steps of that now. And you should see much…
situational-awareness.ai/from-gpt-4-to-…
AGI timelines are very bimodal. It's either by 2030 or bust. AI progress over the last decade has been driven by scaling training compute of frontier systems (3.55x a year, 160x over 4 years). This simply cannot continue beyond this decade, whether you look at chips, power,…
dwarkesh's most recent episode with sholto + trenton is one of the best resources for vibe-checking your takes on LLMs/AI and thinking on the near-future, since broadly everything they say here is both correct and good: youtube.com/watch?v=64lXQP…