SOSP 2026 panel takeaways: what systems research looks like in the AGI era
tianyin_xu · x · 2026-10-08
Vaastav Anand summarized the PACMI 2026 workshop panel (co-located with SOSP 2026 in Prague) on systems research in the AGI era, featuring Laurent Bindschaedler (MPI-SWS), Ana Klimovic and Shivaram Venkatraman (ETH Zürich), moderated by Kostis Kaffes (Columbia).
Key points:
- Are we in the AGI era? Panelists agreed no: AGI lacks an agreed definition; today's models are "capable of being so good but also stupid," strong at easily verifiable tasks with uncertain futures.
- What's changed: Early AI-in-systems work put small models on the execution critical path, requiring surgical usage; now models generate code and are usable beyond the critical path.
- Open questions remain in model design vs. harness design, with the field likely to oscillate between them; one clear trend is model selection — choosing which model for which task.
- The post also covers audience participation and the author's personal takeaways; blog link in the original tweet.
More from AGI Musings
- Ethereum researcher Justin Drake warns AI could crack Bitcoin private keys in months — jamestagg · 2026-10-08
- AI safety circles spar over whether AI exploits stay within the convex hull — QuintinPope5 · 2026-10-08
- Dev compares vibe coding to buying books you never read: 15-min PoCs, zero learning — MarcJSchmidt · 2026-10-08
- Refugees in Kenya's Kakuma camp power AI tasks for dwindling, uncertain pay — nordicinst · 2026-10-08
- Developer claims the application layer will be gone in two years — _AustinCalvert_ · 2026-10-08
- Blockchain instructor of 5 years: don't learn to code, learn vibe coding and product design — mfckr_eth · 2026-10-08