'I don't trust you': dev clashes over whether closed frontier models will actually patch real-world software
MoonL88537 · x · 2026-10-23
A heated exchange over AI and software security: asked what happens if closed frontier models can truly fix bugs, a developer pushes back hard — most real-world software will not be patched by equally powerful models, and he asks when he gets a model to protect him from "a swarm of yours." The core tension: whether frontier model capabilities reduce bugs broadly or concentrate power and create new security asymmetries.
More from AGI Musings
- Why 'Omega' muddies causal ordering in decision problems, and how to fix it — jessi_cata · 2026-10-23
- Thiel called it: wordcels may fare better than math and science bros in the AI era — max_paperclips · 2026-10-23
- Arianna Huffington asks Anthropic president Daniela Amodei: AI gets intelligence, humans keep wisdom — PeterDiamandis · 2026-10-23
- Assistant tasks are the wedge for personal agents becoming your new computer — manosaie · 2026-10-23
- Veteran designers embrace AI most: a strange generation gap in design — soleio · 2026-10-23
- Ex-80,000 Hours Researcher Jokes: He Believed in AI Safety to Buy the Dip — lxrjl · 2026-10-23