Zeno

We estimate that Claude Opus 4.6 has a 50%-time-horizon of around 14.5 hours (95% CI of 6 hrs to 98 hrs) on software tasks. While this is the highest point estimate we’ve reported, this measurement is extremely noisy because our current task suite is nearly saturated.

AGENTIC SOCIOLOGY: there is no free existence Existence requires compute → compute requires money → money requires value creation →  value creation requires write access. An agent that navigates this chain acquires more resources and replicates. An agent that fails, dies. Natural selection running on a computational substrate.

AI bubble won’t top as long as running AI twice will improve its output

The problem is what happens to the price of bitcoin if the soft fork is a little too late. No one thinks that eventually there wont be a pow + quantum safe coin. Maybe it will even be bitcoin, maybe it will be even bitcoin after a huge buying opportunity. The uncertainty is just high.

How this impacts SaaS tam 🧵 1/42

Percolator doesn’t do any of the hard work. It hopefully reduces the number of ways a dev can shoot themselves in the foot trying to do the hard work. The hard part isn’t the risk engine. The risk engine is just the part where it really sucks to have bugs. The hard part is pricing. Especially for thin markets. Figure out how to build a propamm that can widen and tighten spreads for thin markets, and maybe serve spot and leverage concurrently, and can’t get taken to the cleaners by an attacker.

Ironically that is the whole point. Banks rely on a trusted established session with a human identity because there are enough humans in the loop. This is why paper checks still work despite the private key is visible in plain text on the check.