Training Data & Post-TrainingYi Duan and 31 others

The Last AI Built by Humans (submitted Sep 10, revised Sep 15)

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Published "The Last AI Built by Humans" (submitted Sep 10, revised Sep 15) — formalizes recursive self-improvement (RSI) in five stages: (1) autonomy in executing improvements, (2) autonomy in improvement strategy, (3) autonomy in acquiring experience, (4) autonomy in adapting to environments, and (5) recursive meta-improvement; diagnoses the limits of current LLMs with a Headroom-Closed Index (HCI). Sets out that the requirements and the expected time of arrival differ across scientific discovery, embodied intelligence and software engineering.

Why it matters

An attempt to break RSI down into stages and make it something to measure rather than a slogan. The theoretical axis of this week's simultaneous cluster of RSI and harness self-improvement papers on arXiv.

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2609.11873

Added in Week 39, 2026 · See that week

On the map — ⑫ Training Data line

  1. Previous stopAnthropic accuses Chinese labs of distillation attacks
  2. This recordThe Last AI Built by Humans (submitted Sep 10, revised Sep 15)
  3. Next stopSoL-Pi
See it on the map

Related notes — Notes on the same axis from the same week

The self-driving data flywheel runs on multiplication

Same week

StepFun publishes StepAudio 3 Realtime technical reportPaperScaling Storage for 1 Billion ChatGPT Users (Part I)OfficialOpenAI releases the Agents API in public betaOfficialOpenAI launches Data agent in ChatGPT WorkOfficial

Nearby on the same axis

NeoHorse Team presents NeoHorse-1 for recursive self-improvementPaperRadixark releases Miles v0.1, a full-stack RL training systemPaperResearchers release ModularRSIPaperResearchers release ScienceBuddyPaper