Chinese Researcher's "End of Software Engineering" Paper: Agents Score 80% Isolated, 38% Long-Term
alex_verem · x · 2026-09-04
A paper by a Chinese researcher argues AI agents flip software delivery: code becomes scratch paper as agents reason, generate and discard code on the fly, shifting humans to specifying and auditing. But the author concedes agents score 80%+ on isolated tasks yet collapse to 38% maintaining codebases over time. Six days after publishing, he retitled it to the softer "Agentic Software." The likely outcome is a split: some keep coding, others learn to direct the coders.
More from AGI Musings
- Alison Gopnik: Computers Have Had Unpredictable Causal Powers Since Turing, Beyond LLMs — AlisonGopnik · 2026-09-04
- Researcher Argues Against Explicit AI Pause: Capital Costs Already Trigger Slowdowns — herbiebradley · 2026-09-04
- Richard Ngo: alignment researchers who joined between 2015-2022 were a lost generation — RichardMCNgo · 2026-09-04
- Alison Gopnik: abstract characters like Don Quixote always shaped human behavior — AlisonGopnik · 2026-09-04
- davidad argues misaligned ASI branches are dead ends simulators won't simulate — FioraStarlight · 2026-09-04
- Rereading 'History of the Future' 18 months on: what aged well and what didn't — herbiebradley · 2026-09-04