Answering the call to analysis
Study shows that the length of tasks Als can do is doubling every 7 months. Extrapolating this trend predicts that in under five years we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days
The length of tasks Als can do is doubling every 7 months. Extrapolating this trend predicts that in under five years we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days
Nerds + altruism + bravery → awesome
Thinking about timelines has replaced my morning coffee. The spike of adrenaline is more than enough for me.
The length of tasks that generalist frontier model agents can complete autonomously with 50% reliability has been doubling approximately every 7 months
Sometimes life just... Hey! Stop doing that!
TIL that scientists believe it might rain diamonds on Neptune
Many people refuse to accept the concept of trade-offs which leads to harmful consequences
For the Socialists on this Sub
This is what unexpected capability gains from scaling can look like
We will not deploy Class 4 demons. In other news, we are proud to announce our Demon 3.6 series, a massive, game-changing upgrade from our Demon 3.5 line.
12 Tentative Ideas for US AI Policy by Luke Muehlhauser
Same goes for most exercise. Lift heavy stone. Make sad head voice go quiet. [image]
I wish more people got this
Terrifying, fascinating, and also. . . kinda reassuring? I just asked Claude to describe a realistic scenario of AI escape in 2026 and here’s what it said.
Religious Map of Europe
