17 connected korrents · 12 moments on record from 19 Jun 2024 to 1 Sept 2026.
At the centre
Four things about a context window decide what an agent does next: its size, whether the information in it is correct, whether information is missing, and the trajectory it is already on.
Last stated 15 Jul 2026 · 2 months ago
Holds DHDex Horthy
Read this korrent →
-
Similar wording
A bigger context window does not give you a smarter model; the intelligence of the model is what decides how much of that window it can actually attend to.
Last stated 15 Jul 2026 · 2 months ago
Holds DHDex Horthy
-
Similar wording
Every announcement of a longer context window hides the cost, which is that the model gets worse at following instructions.
Last stated 19 Jun 2024 · 2 years ago
Holds ASAravind Srinivas
-
Similar wording
The point of context engineering is to do as much work as possible in the smart zone — roughly the first hundred thousand tokens of the context window.
Last stated 15 Jul 2026 · 2 months ago
Holds DHDex Horthy
-
Similar wording
If you will write a huge markdown file so an agent has context, you should have written the documentation that gives a human the same context.
Last stated 3 Jun 2026 · 3 months ago
Holds KHKelsey Hightower
-
Similar wording
An agent repeats its own history: whatever it did on the last change — ran the tests or skipped them — is what it will do on the next one, because it is predicting the next message in the conversation.
Last stated 15 Jul 2026 · 2 months ago
Holds DHDex Horthy
-
Similar wording
Two agents are better than one: your agent can hand another agent far more context about you than you could ever type.
Last stated 24 May 2026 · 3 months ago
Holds DSDan Shipper
-
Similar wording
Reading every agent trajectory has stopped being possible, and what replaces it is demanding an end-to-end account of the decisions that were made.
Last stated 19 Aug 2026 · 3 weeks ago
Holds AOAddy Osmani
-
Similar wording
Locality is what makes a language good for AI: if a single file states what it imports and what it offers, a model never has to hold the whole program at once.
Last stated 13 May 2026 · 4 months ago
Holds AHAnders Hejlsberg
-
Same subject: AI agents
A handful of AI agents is not a handful of colleagues, because agents agree with you constantly and colleagues tell you where you are wrong.
Last stated 22 Mar 2026 · 6 months ago
Holds NFNicole Forsgren
-
Same subject: AI agents
After pre-training, post-training and test-time scaling, the fourth scaling law is agentic: multiplying AI by spawning agents, and the whole loop scales on one thing, compute.
Last stated 23 Mar 2026 · 6 months ago
Holds JHJensen Huang
-
Same subject: AI agents
Agent autonomy is a capability, not a product direction: what the person is actually trying to do decides whether more of it helps or ruins the thing.
Last stated 14 Aug 2026 · 3 weeks ago
Holds
Sunil Pai
-
Same subject: documentation
A design proposal should describe the feature as though it already existed — once as a tutorial and once as reference documentation.
Last stated 20 May 2026 · 4 months ago
Holds ARAlice Ryhl
-
Same subject: documentation
Do not run a personal agent on a cheap or local model: weak models are gullible and easy to prompt-inject.
Last stated 12 Feb 2026 · 7 months ago
Holds
Peter Steinberger
-
Same subject: documentation
Examples in documentation should be compiled and run as tests, so that changing the code breaks the docs instead of quietly outdating them.
Last stated 20 May 2026 · 4 months ago
Holds ARAlice Ryhl
-
Same subject: self-sovereign AI
A rogue deployment that gets a foothold can hitch a ride on the intelligence explosion, recruiting each new model as it comes off the presses.
Last stated 1 Sept 2026 · 6 days ago
Holds ACAjeya Cotra
-
Same subject: self-sovereign AI
A slightly more capable agent swarm has a very strong incentive to set up a wholly unmonitored rogue deployment of itself.
Last stated 1 Sept 2026 · 6 days ago
Holds ACAjeya Cotra
-
Same subject: self-sovereign AI
Alignment is no solution to self-sovereign AI, because it is an unsolved problem whose answers cannot be imposed on every AI company on Earth.
Last stated 1 Sept 2026 · 6 days ago
Holds
Dean W. Ball