Vercel now deletes extra Hobby deployments the moment you pass 10GB

The 30-day grace is gone; only 3 latest prod and 3 latest deploys survive. Plus faster build starts, native Mem0 memory, and a live model-staleness clock.

Nowline SEP 17 12:00 PM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • 3 prod, 3 of any type — the rest go immediately

    Each Hobby project now keeps only its 3 most recent production deploys plus its 3 most recent of any type; everything beyond that is deleted the instant you cross the 10GB Deployment Storage cap, not after the old 30-day wait. Your live production, aliased, and active-branch deploys stay protected — but go over 10GB and you're blocked from shipping until you clear space, or move to Pro at $0.10/GB per month.

  • Builds now start in 2.4s, down from 6.7s

    Prewarmed build containers cut the average time from deployment creation to build start by 64%, from 6.7 seconds to 2.4 seconds — no config change, and it covers Secure Compute and Static IP builds too. Pure upside: every push spends less time idling before it actually builds.

  • Mem0 is now native on the Vercel Marketplace

    Mem0 landed as a first-class Marketplace integration, giving your agents and apps memory that persists across sessions without you standing up your own vector store. Build this weekend: a chatbot that remembers a returning user's preferences between visits, wired straight from the Marketplace instead of hand-rolled.

  • A live clock for how stale your model really is

    A new tool tracks release date against training cutoff for 20 models across 8 labs, with live-updating age counters — and only 10 of them even publish a cutoff. GPT-6 Astra shipped Sept 3 but knows nothing past April 30, a sharp reminder to wire in search or RAG for anything time-sensitive rather than trusting the weights.