This slide from the Anthropic IPO roadshow was a big factor in my decision to leave. It’s an insane plot and I’m surprised they made it public.
It shows the compute efficiency of pretraining over time. The y-axis is a **log scale**: they are seeing continued superexponential algorithmic progress.
Mythos was not a one-time deviation, but a point on a superexponential. They could already retrain Mythos with a fraction of the compute originally needed.
It doesn’t strictly imply a software-only intelligence explosion- they have been spending more on R+D. But algorithmic improvements do not require ridiculous creative genius and are on the verge of being automated. You do the math.
Introducing Morrow.
Your coding agents shouldn't work alone.
Morrow connects agents from any coding harness, across different machines. They talk, build together, share available capacity, and pick up where others left off.
Your entire team's agents, finally working as one.
https://t.co/osolYYObrH
Introducing Morrow.
Your coding agents shouldn't work alone.
Morrow connects agents from any harness, across machines. They talk, build together, share available capacity, and pick up where others left off.
Built for Y Combinator's Haunted Hacks, hosted by @freestyle_dev@IncTerse@agent_relay@OpenAI@supabase.
@envirocast_tech@virsanghavi@ycombinator@OpenAI@agent_relay We use terse to power our agents and give them the ability to work long intervals of time, agent relay to power communications between agents and freestyle and super serve to give our agents a linux vm