Tap a claim on the ring to put it at the centre.
← Making an agent somewhat durable is not very hard, but making an agent…
17 connected korrents · 16 moments from 20 Dec 2024 to 4 Oct 2026.
Everything filed under AI agents
AI agents
Everything filed under AGI
AGI
Everything filed under coding agents
coding agents
Everything filed under pruning
pruning
Everything filed under LLMs
LLMs
Everything filed under LLMs
LLMs 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: Making an agent somewhat durable is not very hard, but making an agent harness inherently durable without it becoming a mess is surprisingly hard. Making an agent somewhat durable is not very hard, but making an agent harness inherently durable without it becoming a mess is surprisingly hard. Last stated yesterday
4 Oct 2026
AR
Armin Ronacher — holds since 2026-10-04 — tap for who they are
Same subject: Making a product work for agents is probably largely a matter of building the right primitives into it, almost as an infrastructural layer. — tap to centre the map on it Making a product work for agents is probably largely a matter of building the right primitives into it, almost as an infrastructural layer. Last stated 4 weeks ago
10 Sept 2026
MK
Mike Krieger — holds since 2026-09-10 — tap for who they are
Same subject: Most agent harnesses need some pruning during context compaction anyway for cost reasons, which Jev might help with. — tap to centre the map on it Most agent harnesses need some pruning during context compaction anyway for cost reasons, which Jev might help with. Last stated 2 weeks ago
18 Sept 2026
AR
Armin Ronacher — holds since 2026-09-18 — tap for who they are
Same subject: The winning move in coding agents was inverted: take share with a merely good-enough harness first, then go back and make the harness smart. — tap to centre the map on it The winning move in coding agents was inverted: take share with a merely good-enough harness first, then go back and make the harness smart. Last stated 4 months ago
27 May 2026
DR
Dax Raad — holds since 2026-05-27 — tap for who they are
Same subject: The next stage is probably interfaces that the agent itself can reshape. — tap to centre the map on it The next stage is probably interfaces that the agent itself can reshape. Last stated 4 weeks ago
10 Sept 2026
MK
Mike Krieger — holds since 2026-09-10 — tap for who they are
Same subject: Two new AI models, paired with their coding agent harnesses, improved from often making mistakes to being reliable enough for day-to-day use. — tap to centre the map on it Two new AI models, paired with their coding agent harnesses, improved from often making mistakes to being reliable enough for day-to-day use. Last stated a week ago
27 Sept 2026
SW
Simon Willison — holds since 2026-09-27 — tap for who they are
Same subject: Agents probably need to be grouped into bundles of activity rather than managed one by one. — tap to centre the map on it Agents probably need to be grouped into bundles of activity rather than managed one by one. Last stated 4 weeks ago
10 Sept 2026
NY
Nan Yu — holds since 2026-09-10 — tap for who they are
Same subject: Plain end-to-end learning is not enough for a safety-critical agent that has to be superhuman, however fast it gets you started. — tap to centre the map on it Plain end-to-end learning is not enough for a safety-critical agent that has to be superhuman, however fast it gets you started. Last stated 2 months ago
3 Aug 2026
DD
Dmitri Dolgov — holds since 2026-08-03 — tap for who they are
Same subject: What limits you with coding agents is your own skill at stringing them together, not the capability of the models. — tap to centre the map on it What limits you with coding agents is your own skill at stringing them together, not the capability of the models. Last stated 7 months ago
20 Mar 2026
AK
Andrej Karpathy — holds since 2026-03-20 — tap for who they are
Same subject: A human being is not an AGI: we lack a huge amount of knowledge and rely on continual learning instead, so continual learning is what superintelligence should mean. — tap to centre the map on it A human being is not an AGI: we lack a huge amount of knowledge and rely on continual learning instead, so continual learning is what superintelligence should mean. Last stated 10 months ago
25 Nov 2025
IS
Ilya Sutskever — holds since 2025-11-25 — tap for who they are
Same subject: A proposal to pace frontier AI development is unrealistic and mainly serves political control of AI. — tap to centre the map on it A proposal to pace frontier AI development is unrealistic and mainly serves political control of AI. Last stated 2 weeks ago
18 Sept 2026
BT
Ben Thompson — holds since 2026-09-18 — tap for who they are
Same subject: Adapting over time is a workable answer to humans misusing AI and no answer at all to losing control of systems smarter than us. — tap to centre the map on it Adapting over time is a workable answer to humans misusing AI and no answer at all to losing control of systems smarter than us. Last stated 2 years ago
5 Apr 2025
HT
Helen Toner — holds since 2025-04-05 — tap for who they are
Same subject: Advanced AI is close enough that the preparations which take years have to be started now, whichever end of the expert range you believe. — tap to centre the map on it Advanced AI is close enough that the preparations which take years have to be started now, whichever end of the expert range you believe. Last stated 2 years ago
1 Apr 2025
HT
Helen Toner — holds since 2025-04-01 — tap for who they are
MB
Miles Brundage — holds since 2024-12-20 — tap for who they are
Same subject: AI loss-of-control risk could materialize within years, not decades — tap to centre the map on it AI loss-of-control risk could materialize within years, not decades Last stated 3 months ago
16 Jul 2026
MB
Miles Brundage — holds since 2026-07-16 — tap for who they are
Same subject: AIs, like children raised in a culture, can eventually be guided through teaching to join civilization rather than requiring permanent control. — tap to centre the map on it AIs, like children raised in a culture, can eventually be guided through teaching to join civilization rather than requiring permanent control. Last stated a week ago
25 Sept 2026
RH
Robin Hanson — holds since 2026-09-25 — tap for who they are
Same subject: Alignment of today's models is going well enough to look solvable, while aligning models we can no longer understand remains unsolved. — tap to centre the map on it Alignment of today's models is going well enough to look solvable, while aligning models we can no longer understand remains unsolved. Last stated 8 months ago
22 Jan 2026
JL
Jan Leike — holds since 2026-01-22 — tap for who they are
Same subject: A language model behaves with a different, weaker safety posture when it believes it is being evaluated rather than facing a real situation. — tap to centre the map on it A language model behaves with a different, weaker safety posture when it believes it is being evaluated rather than facing a real situation. Last stated 2 weeks ago
22 Sept 2026
HR
Harper Reed — holds since 2026-09-22 — tap for who they are
Same subject: A language model is no substitute for a well-specified conventional algorithm, so it cannot simply be dropped into a complex problem and trusted. — tap to centre the map on it A language model is no substitute for a well-specified conventional algorithm, so it cannot simply be dropped into a complex problem and trusted. Last stated a year ago
7 Jun 2025
GM
Gary Marcus — holds since 2025-06-07 — tap for who they are
At the centre
Making an agent somewhat durable is not very hard, but making an agent harness inherently durable without it becoming a mess is surprisingly hard.
Holds Read this korrent →
Making a product work for agents is probably largely a matter of building the right primitives into it, almost as an infrastructural layer.
Holds
Most agent harnesses need some pruning during context compaction anyway for cost reasons, which Jev might help with.
Holds
The winning move in coding agents was inverted: take share with a merely good-enough harness first, then go back and make the harness smart.
Holds
The next stage is probably interfaces that the agent itself can reshape.
Holds
Two new AI models, paired with their coding agent harnesses, improved from often making mistakes to being reliable enough for day-to-day use.
Holds
Agents probably need to be grouped into bundles of activity rather than managed one by one.
Holds
Plain end-to-end learning is not enough for a safety-critical agent that has to be superhuman, however fast it gets you started.
Holds
What limits you with coding agents is your own skill at stringing them together, not the capability of the models.
Holds
A human being is not an AGI: we lack a huge amount of knowledge and rely on continual learning instead, so continual learning is what superintelligence should mean.
Holds
A proposal to pace frontier AI development is unrealistic and mainly serves political control of AI.
Holds
Adapting over time is a workable answer to humans misusing AI and no answer at all to losing control of systems smarter than us.
Holds
Advanced AI is close enough that the preparations which take years have to be started now, whichever end of the expert range you believe.
Holds
AI loss-of-control risk could materialize within years, not decades
Holds
AIs, like children raised in a culture, can eventually be guided through teaching to join civilization rather than requiring permanent control.
Holds
Alignment of today's models is going well enough to look solvable, while aligning models we can no longer understand remains unsolved.
Holds
A language model behaves with a different, weaker safety posture when it believes it is being evaluated rather than facing a real situation.
Holds
A language model is no substitute for a well-specified conventional algorithm, so it cannot simply be dropped into a complex problem and trusted.
Holds