jim
@jimfund.com
(Ironic): I won't read writing that hasn't been looked over by an LLM. If the writer hasn't taken the time to run his work by an LLM, then why should I take the time to read it?
February 14, 2026 at 2:12 PM
(Ironic): I won't read writing that hasn't been looked over by an LLM. If the writer hasn't taken the time to run his work by an LLM, then why should I take the time to read it?
That when model 50% time-horizons are at a certain point their 2% time-horizon is much further out has implications about autonomous research.
February 12, 2026 at 7:48 PM
That when model 50% time-horizons are at a certain point their 2% time-horizon is much further out has implications about autonomous research.
AI forecasters should stop making deliberately conservative assumptions.
February 12, 2026 at 9:56 AM
AI forecasters should stop making deliberately conservative assumptions.
We will likely have 50% time-horizons of over 50 hours by October this year.
February 11, 2026 at 12:51 PM
We will likely have 50% time-horizons of over 50 hours by October this year.
The impact of each time-horizon doubling exceeds that of the last. Going from 3 -> 6 hours was big. From 6 -> 12 will be massive. This pattern will hold. And doubling-times will decrease.
February 9, 2026 at 2:49 PM
The impact of each time-horizon doubling exceeds that of the last. Going from 3 -> 6 hours was big. From 6 -> 12 will be massive. This pattern will hold. And doubling-times will decrease.
It's been a while since I thought seriously about which AI lab is in the lead and which has the best near-term outlook.
February 6, 2026 at 1:21 PM
It's been a while since I thought seriously about which AI lab is in the lead and which has the best near-term outlook.
"[someone suggested] that Dario talks about extremely short timelines and existential risk in order to raise funds. It’s very much the opposite."
January 30, 2026 at 4:30 AM
"[someone suggested] that Dario talks about extremely short timelines and existential risk in order to raise funds. It’s very much the opposite."
METR's graph has been updated with revised values calculated using an expanded task set.
metr.org/blog/2026-1-...
metr.org/blog/2026-1-...
January 29, 2026 at 9:15 PM
METR's graph has been updated with revised values calculated using an expanded task set.
metr.org/blog/2026-1-...
metr.org/blog/2026-1-...
interesting post from Thomas Kwa www.lesswrong.com/posts/Zr37dY...
January 27, 2026 at 6:44 AM
interesting post from Thomas Kwa www.lesswrong.com/posts/Zr37dY...
I think at some point this year the best-performing LLMs will begin thinking in pure neuralese.
January 24, 2026 at 4:01 AM
I think at some point this year the best-performing LLMs will begin thinking in pure neuralese.
reading 'Claude's Constitution'
www.anthropic.com/constitution
www.anthropic.com/constitution
January 23, 2026 at 9:18 AM
reading 'Claude's Constitution'
www.anthropic.com/constitution
www.anthropic.com/constitution
let's think about this market
January 22, 2026 at 3:23 AM
let's think about this market
let's think about this market
January 17, 2026 at 11:20 PM
let's think about this market
the idea of random employees writing reports, doing analysis will soon be of the past
December 9, 2025 at 5:03 PM
the idea of random employees writing reports, doing analysis will soon be of the past
figuring out the most useful markets
- something useful to know
- that market participants can predict well
- and we cannot predict better ourselves
an example is "which lab will lead at a certain tech"
or generally things which are (a) important but which (b) we don't have the time to monitor
- something useful to know
- that market participants can predict well
- and we cannot predict better ourselves
an example is "which lab will lead at a certain tech"
or generally things which are (a) important but which (b) we don't have the time to monitor
September 28, 2025 at 4:10 PM
figuring out the most useful markets
- something useful to know
- that market participants can predict well
- and we cannot predict better ourselves
an example is "which lab will lead at a certain tech"
or generally things which are (a) important but which (b) we don't have the time to monitor
- something useful to know
- that market participants can predict well
- and we cannot predict better ourselves
an example is "which lab will lead at a certain tech"
or generally things which are (a) important but which (b) we don't have the time to monitor
AI beats Minecraft (RSG) in under 10 minutes before 2027?
23% chance. RSG = Random Seed Glitchless
Must be Java version
CAN:
Can use scaffolds or similar.
CAN'T:
Cannot slow down game time or stuff like that.
Can't get human help.
Can't reverse engine...
manifold.markets
September 27, 2025 at 11:24 PM
