Korablog.
Notes on what we are seeing in agent research, from the team building Koragraph.
Posts
5 postsA Code Graph Is a Pile of Small Decisions, and Jev Prices Them Near Zero
TypeSafe’s Jev returns a typed decision in under a tenth of a second for a fraction of a cent, cheap enough to ask a learned question at every node and edge of a code graph. Three ideas, mapping where that helps and where independence breaks.
Dario Amodei’s Case for Pacing the AI Frontier
Altman, Musk, and Trump reacted to "We Must Pace the Frontier" within days. None of them touched the recursive self-improvement and agent-swarm claims actually driving Amodei's argument.
Two Millennium Prize Results in One Month, and a Credit Fight Over Who Actually Found the Idea
Claude needed sixty agents to push a Riemann hypothesis bound to 67.2 percent. OpenAI needed ten thousand to prove Navier-Stokes blow-up, and now disputes credit with the two humans who found the route first.
18 of 19 Frontier Models Score Lower on Agentic Work Than on Reasoning, and Graph Theory Already Named the Bottleneck
18 of 19 frontier models score lower on agentic work than on reasoning. Graph learning named this squeeze years ago: over-squashing, and effective resistance finds the bottleneck.
GPT-6 Astra Scored 62.7% and 99.9% on the Same Benchmark. The Difference Was Memory
GPT-6 Astra scored 37 points higher on ARC-AGI-3 with the same weights. New rate-distortion research explains why memory, not intelligence, made the gap.
What this is for
Korablog is plain-language notes on the same work that shows up as papers on /research, written for anyone searching for the problem rather than the paper.