Tap a claim on the ring to put it at the centre.
← In GRPO, a high standard deviation of the advantages is desirable.
17 connected korrents · 15 moments from 5 Feb 2025 to 28 Oct 2026.
Everything filed under reinforcement learning
reinforcement learning
Everything filed under LLMs
LLMs
Everything filed under AI alignment
AI alignment
Everything filed under scaling laws
scaling laws
Everything filed under startups
startups
Everything filed under SaaS
SaaS
Everything filed under mental models
mental models
Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject
Read this korrent: In GRPO, a high standard deviation of the advantages is desirable.
In GRPO, a high standard deviation of the advantages is desirable.
Last stated a month ago
10 Oct 2026
SR
Sebastian Raschka — holds since 2026-10-10 — tap for who they are
Same subject: In GRPO training, the loss should stay roughly stable around a value, and large loss spikes are what to avoid. — tap to centre the map on it
In GRPO training, the loss should stay roughly stable around a value, and large loss spikes are what to avoid.
Last stated a month ago
10 Oct 2026
SR
Sebastian Raschka — holds since 2026-10-10 — tap for who they are
Same subject: The small GRPO settings are only for efficiency; whatever you can increase, such as rollouts, you should. — tap to centre the map on it
The small GRPO settings are only for efficiency; whatever you can increase, such as rollouts, you should.
Last stated a month ago
10 Oct 2026
SR
Sebastian Raschka — holds since 2026-10-10 — tap for who they are
Same subject: Data is becoming more valuable not only for training models but as a complement to cheaper intelligence. — tap to centre the map on it
Data is becoming more valuable not only for training models but as a complement to cheaper intelligence.
Last stated a month ago
9 Oct 2026
PC
Patrick Collison — holds since 2026-10-09 — tap for who they are
Same subject: The p(abundance) outlook is far more likely than the heavily covered p(doom) discussion and deserves more engagement. — tap to centre the map on it
The p(abundance) outlook is far more likely than the heavily covered p(doom) discussion and deserves more engagement.
Last stated 2 months ago
14 Sept 2026
DH
David Heinemeier Hansson — holds since 2026-09-14 — tap for who they are
Same subject: A rising share of statistically insignificant experiments is the signal that a team has exploited an area too far and should go back to exploring. — tap to centre the map on it
A rising share of statistically insignificant experiments is the signal that a team has exploited an area too far and should go back to exploring.
Last stated a year ago
5 Oct 2025
AC
Albert Cheng — holds since 2025-10-05 — tap for who they are
Same subject: Premium AI products will earn much higher margins, closer to enterprise software's 40%, because client data makes them stickier. — tap to centre the map on it
Premium AI products will earn much higher margins, closer to enterprise software's 40%, because client data makes them stickier.
Last stated 4 weeks ago
28 Oct 2026
AD
Aswath Damodaran — holds since 2026-10-28 — tap for who they are
Same subject: GRPO is a bit nicer and simpler than classic PPO. — tap to centre the map on it
GRPO is a bit nicer and simpler than classic PPO.
Last stated 2 months ago
3 Oct 2026
SR
Sebastian Raschka — holds since 2026-10-03 — tap for who they are
Same subject: Uncovering and encoding information about organization or population-specific quirks can be more valuable than adding more data. — tap to centre the map on it
Uncovering and encoding information about organization or population-specific quirks can be more valuable than adding more data.
Last stated 2 months ago
12 Sept 2026
AK
Adam Kucharski — holds since 2026-09-12 — tap for who they are
Same subject: "Scaling" was powerful because it was one word: naming a research direction is what tells a whole field what to do next. — tap to centre the map on it
"Scaling" was powerful because it was one word: naming a research direction is what tells a whole field what to do next.
Last stated 12 months ago
25 Nov 2025
IS
Ilya Sutskever — holds since 2025-11-25 — tap for who they are
Same subject: A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number. — tap to centre the map on it
A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number.
Last stated 5 months ago
26 Jun 2026
NB
Noam Brown — holds since 2026-06-26 — tap for who they are
Same subject: A CPU and memory crunch is coming. — tap to centre the map on it
A CPU and memory crunch is coming.
Last stated 4 weeks ago
26 Oct 2026
DM
Dillon Mulroy — holds since 2026-10-26 — tap for who they are
Same subject: "Hire smart people and get out of their way" is bad management advice; leaders should stay closely involved, building trust and working through hard problems with their people to create alignment. — tap to centre the map on it
"Hire smart people and get out of their way" is bad management advice; leaders should stay closely involved, building trust and working through hard problems with their people to create alignment.
Last stated a month ago
9 Oct 2026
CR
Carl Rivera — holds since 2026-10-09 — tap for who they are
Same subject: A 5% level of AI risk is crazy and unacceptable; we would not accept a 5% chance of massive fatalities from paint companies. — tap to centre the map on it
A 5% level of AI risk is crazy and unacceptable; we would not accept a 5% chance of massive fatalities from paint companies.
Last stated a year ago
8 Jul 2025
RG
Ryan Greenblatt — holds since 2025-07-08 — tap for who they are
Same subject: A base model learns humanity's collective knowledge, and it then has to be aligned to behave one way or another. — tap to centre the map on it
A base model learns humanity's collective knowledge, and it then has to be aligned to behave one way or another.
Last stated a year ago
10 Sept 2025
SA
Sam Altman — holds since 2025-09-10 — tap for who they are
Same subject: "Return on tokens" has the same measurement problems as hours and capacity: it still hinges on a theory of value. — tap to centre the map on it
"Return on tokens" has the same measurement problems as hours and capacity: it still hinges on a theory of value.
Last stated 3 months ago
17 Aug 2026
JC
John Cutler — holds since 2026-08-17 — tap for who they are
Same subject: A badly written AI outbound email is evidence of a bad vendor, not of a limit of AI. — tap to centre the map on it
A badly written AI outbound email is evidence of a bad vendor, not of a limit of AI.
Last stated 11 months ago
1 Jan 2026
JL
Jason Lemkin — holds since 2026-01-01 — tap for who they are
Same subject: A base model is essentially a very expensive autocomplete that continues text according to the statistics of its training documents. — tap to centre the map on it
A base model is essentially a very expensive autocomplete that continues text according to the statistics of its training documents.
Last stated 2 years ago
5 Feb 2025
AK
Andrej Karpathy — holds since 2025-02-05 — tap for who they are
same subject or similar wording a cloud: claims about one subject, named for it bar: when it was last stated, on a scale from 2015 to today — full is today a face: someone who holds the claim — tap it for who they are
At the centre
In GRPO, a high standard deviation of the advantages is desirable.
Last stated 10 Oct 2026 · a month ago
Holds Sebastian Raschka
Read this korrent →
Similar wording
In GRPO training, the loss should stay roughly stable around a value, and large loss spikes are what to avoid.
Last stated 10 Oct 2026 · a month ago
Holds Sebastian Raschka
Similar wording
The small GRPO settings are only for efficiency; whatever you can increase, such as rollouts, you should.
Last stated 10 Oct 2026 · a month ago
Holds Sebastian Raschka
Similar wording
Data is becoming more valuable not only for training models but as a complement to cheaper intelligence.
Last stated 9 Oct 2026 · a month ago
Holds Patrick Collison
Similar wording
The p(abundance) outlook is far more likely than the heavily covered p(doom) discussion and deserves more engagement.
Last stated 14 Sept 2026 · 2 months ago
Holds David Heinemeier Hansson
Similar wording
A rising share of statistically insignificant experiments is the signal that a team has exploited an area too far and should go back to exploring.
Last stated 5 Oct 2025 · a year ago
Holds Albert Cheng
Similar wording
Premium AI products will earn much higher margins, closer to enterprise software's 40%, because client data makes them stickier.
Last stated 28 Oct 2026 · 4 weeks ago
Holds Aswath Damodaran
Similar wording
GRPO is a bit nicer and simpler than classic PPO.
Last stated 3 Oct 2026 · 2 months ago
Holds Sebastian Raschka
Similar wording
Uncovering and encoding information about organization or population-specific quirks can be more valuable than adding more data.
Last stated 12 Sept 2026 · 2 months ago
Holds Adam Kucharski
Same subject: scaling laws
"Scaling" was powerful because it was one word: naming a research direction is what tells a whole field what to do next.
Last stated 25 Nov 2025 · 12 months ago
Holds Ilya Sutskever
Same subject: scaling laws
A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number.
Last stated 26 Jun 2026 · 5 months ago
Holds Noam Brown
Same subject: scaling laws
A CPU and memory crunch is coming.
Last stated 26 Oct 2026 · 4 weeks ago
Holds Dillon Mulroy
Same subject: AI alignment
"Hire smart people and get out of their way" is bad management advice; leaders should stay closely involved, building trust and working through hard problems with their people to create alignment.
Last stated 9 Oct 2026 · a month ago
Holds Carl Rivera
Same subject: AI alignment
A 5% level of AI risk is crazy and unacceptable; we would not accept a 5% chance of massive fatalities from paint companies.
Last stated 8 Jul 2025 · a year ago
Holds Ryan Greenblatt
Same subject: AI alignment
A base model learns humanity's collective knowledge, and it then has to be aligned to behave one way or another.
Last stated 10 Sept 2025 · a year ago
Holds Sam Altman
Same subject: LLMs
"Return on tokens" has the same measurement problems as hours and capacity: it still hinges on a theory of value.
Last stated 17 Aug 2026 · 3 months ago
Holds John Cutler
Same subject: LLMs
A badly written AI outbound email is evidence of a bad vendor, not of a limit of AI.
Last stated 1 Jan 2026 · 11 months ago
Holds Jason Lemkin
Same subject: LLMs
A base model is essentially a very expensive autocomplete that continues text according to the statistics of its training documents.
Last stated 5 Feb 2025 · 2 years ago
Holds Andrej Karpathy