C-HD algorithm beats Dijkstra only narrowly in an exotic regime, prompting RSI skepticism
teortaxesTex · x · 2026-09-23
A blog post introducing the C-HD pathfinding algorithm claimed it beats Dijkstra, but a follow-up analysis shows the improvement holds only in an exotic regime: most real road networks fall in regimes where the 2026 algorithm or plain Dijkstra remains the relevant benchmark, and the comparison uses theoretical asymptotic bounds rather than actual runtime — C-HD may still be slower in practice.
teortaxesTex draws a broader lesson: if 'better than Dijkstra' is this narrow, Recursive Self-Improvement (RSI) may be less dramatic than imagined — we don't know how close the best human algorithms already are to algorithmic limits. Or maybe they aren't.
More from Research
- Pokemon benchmark Paradigm 3: Astra generalizes to scrambled maps and fan-made games while rivals memorize — gleech · 2026-09-24
- AI Completes Fan-Made Pokemon Brown in 10K Steps: Real Generalization or Whack-a-Mole? — gleech · 2026-09-24
- Astra Beats Fan-Made Pokemon Brown in ~10K Steps Without Memorizing It — gleech · 2026-09-24
- Genome language model Omnii enters research preview, unveiled on Latent Space podcast — exnx · 2026-09-24
- Economist Novosad: AI detector debates trip over FPR vs FNR confusion — paulnovosad · 2026-09-24
- exNX unveils research preview of genome language model Omnii on Latent Space — exnx · 2026-09-24