Anshul Kundaje: research artifacts should be version-controlled, not frozen as a single 'final version'
anshulkundaje · x · 2026-11-19
Continuing the debate with Sasha Gusev, Anshul Kundaje argues research artifacts should be version-controlled like software, with every update diffable. He says he simply doesn't believe the current model of one immutable "final version" that can never be diffed or updated without another elaborate exercise is remotely optimal.
The remark points toward continuously evolving, open-source-style scientific publishing.
More from AGI Musings
- "A slightly politer version of ourselves would be a waste": questioning the obsession with human-like AI — VraserX · 2026-11-19
- Philosopher Thomas Nagel: language alone means LLMs can't be conscious — DanielWhiteson · 2026-11-19
- maxmarchione: everyone's p(doom) is already 100% — and humans do want to live forever — tomchapin · 2026-11-19
- OpenAI Researcher Flags a 'Sovereignty Gap': Science Institutions 800:1 Behind Frontier AI — shyamalanadkat · 2026-11-19
- When AI makes content worthless, authenticity and trust become the moat — kylegawley · 2026-11-19
- tszzl: models rewriting math and physics still fail at simple common sense — tszzl · 2026-11-19