▶ YouTube Summaries

• https://www.youtube.com/watch?v=n2YtM-MecZk
• Original title: Cloudflare bought Vite to destroy Vercel

Cloudflare acquired Void Zero — Evan You's company behind Vite, Vitest, Rolldown, OXC and vite-plus. Theo argues this is a strategic play to fix Cloudflare's notoriously bad DX and build an end-to-end cloud where agents, not humans, deploy apps.

Background: Void Zero was building "Void," a platform to make deployment as code-driven as building — `import { db } from 'void/db'` provisions a real database, with KV/storage/queues/AI similar; it deploys to Cloudflare Workers. Theo's framing: building apps got trivial with AI (~30 min), but deploying still takes hours of stitching together Vercel, Convex, Clerk, dashboards and Terraform — Void closes that gap.

Cloudflare's weakness was never infrastructure (Workers, D1, durable objects are powerful) but developer experience — deploying requires Cloudflare-specific YAML, Wrangler, and fighting contradictory docs (Theo recounts the pain of porting T3 Chat). That's why Cloudflare keeps poaching Vercel people and acquiring Astro and now Void Zero — they know they can't build good DX themselves. He plots a spectrum (CSS → framework → bundler → CDN → compute → database): Vercel owns framework→compute well; Cloudflare is strong CDN→compute and reaching into databases; Void Zero brings the left side (bundler/framework/build).

He connects this to his own unreleased project Lakebed: an end-to-end "cloud as an SDK" where an agent can `npx lakebed deploy` and get real DB, auth and sync with no dashboards — because agents are far better at writing code than navigating dashboards, and AI makes previously "boil the ocean" projects worth building. After this acquisition, the company he fears for Lakebed isn't Vercel (no good DB story) but Cloudflare — especially if they also buy Convex. Cloudflare is also committing $1M to a Vite ecosystem fund; everything stays open source.