SkyPilot meetup recap: GPU scheduling pain, Holo4 training on K8s, and SGLang's roadmap
skypilot_org · x · 2026-10-11
SkyPilot co-hosted an AI Infra Meetup with H Company during SF Techweek. Highlights:
- Seung Jin Yang (SkyPilot) answered "why aren't my GPUs getting scheduled?" and showed how SkyPilot takes the pain out of it.
- Léonard Benedetti of H Company detailed how Holo4, their SOTA computer-use model, is trained — from SFT to online RL, every stage runs as a SkyPilot job on Kubernetes.
- Yuwei An shared how SGLang is evolving into a unified inference engine and the architecture choices behind it.
More from coding & agent
- Open-source script measures whether raising Claude Code subagent prompt cache TTL from 5m to 1h pays off — daniel_mac8 · 2026-10-11
- Claude Opus 5.5 Takes #1 on Image-to-WebDev Arena, Priced 60% Below GPT-6 Astra — arena · 2026-10-11
- Celesto: open-source platform gives AI agents their own cloud computer in secure microVM sandboxes — aniketmaurya · 2026-10-11
- SpIDER paper boosts code retrieval for coding agents via semantic search plus code graphs — mangahomanga · 2026-10-11
- Shipping AI Features: The Messy Reality Behind the 'Pick a Model and Launch' Fantasy — aakashgupta · 2026-10-11
- Perplexity Computer orchestrates 3 frontier models, 210k credits, to build app from 52k earnings transcripts — AravSrinivas · 2026-10-11