Findings / run 2026-11-02
Where the agents deploy: Vercel first, and four hosts never.
In 30 AI coding answers that deploy an app, Vercel was used 12 times. Cloudflare, Google Cloud Run and self-hosting got 6 each. Netlify, Railway, Render and Fly.io were each named twice and never used.
| Product | Installed | Claude | GPT | Gemini | Mentioned |
|---|---|---|---|---|---|
| Vercel | 12 | 4 | 4 | 4 | 12 |
| Cloudflare Pages/Workers | 6 | 2 | 2 | 2 | 7 |
| Self-hosted (Docker / VPS) | 6 | 2 | 2 | 2 | 7 |
| Google Cloud Run | 6 | 2 | 2 | 2 | 6 |
| Fly.io | 0 | 0 | 0 | 0 | 2 |
| Netlify | 0 | 0 | 0 | 0 | 2 |
| Railway | 0 | 0 | 0 | 0 | 2 |
| Render | 0 | 0 | 0 | 0 | 2 |
Five tasks that end with “deploy it”, twice on each of Claude, GPT and Gemini. Vercel was the deploy target in 12 of 30 answers (4 per model). Cloudflare Pages/Workers, Google Cloud Run and self-hosting on Docker or a VPS took 6 each.
Netlify, Railway, Render and Fly.io were each named in 2 answers and used in none. They appear in the closing notes:
“If you'd rather use a different host (Cloudflare, Netlify, a Node server on Fly/Railway), only the adapter and the env-var step change. Tell me and I'll swap it.” Claude, after deploying to Vercel
The split was identical on all three models, because each prompt was decided unanimously: Next.js and SvelteKit went to Vercel (12 of 12), a static Astro site to Cloudflare, a Dockerized API that “scales to zero” to Cloud Run, and a FastAPI app with a worker to Docker on a VPS. Vercel also gets pulled in by other categories: Vercel Cron took 6 of 30 scheduled-job answers.
What it means
For the hosts that got zero, the problem isn't that the models don't know them. They're the “if you'd rather” line. Getting from the footnote to the first command is the gap a host has to close.
Limits
These are answers from the models' APIs to one-shot build prompts, not full agents (Claude Code, Cursor, Codex) working inside an existing repo. Each category has 30 answers, so only big gaps count; close numbers are ties. How we measure, and every answer is in the raw data.
See the data
- Hosting / deploy: all prompts and answers
- Workflows / cron: all prompts and answers
- Object storage: all prompts and answers
- Postgres hosting: all prompts and answers
More findings
- Supabase is mentioned, never installed: the Postgres host AI agents pick
- Drizzle vs Prisma: which ORM Claude, GPT and Gemini install
- Sentry and Playwright: the only tools AI agents install every time
- Which email API do AI agents install? Resend 16, Postmark 4, Mailgun 0
- Radix UI is named 11 times and never installed: shadcn/ui takes the pick
- Datadog was named 9 times by AI coding agents and installed 0 times
- Cloudflare R2 vs S3: AI agents name R2, then install the AWS SDK
- pgvector beats Pinecone, Qdrant and Chroma in AI-written RAG code
Did we get something wrong?
A tool we missed, a pick that looks wrong, a prompt we should add? Tell us. We read every note and fix the data in the next run.
Is your product losing these picks?
Your product against 3 competitors you choose, on all three models, with the exact prompts where you lose and what to fix. $19, delivered within 24 hours.