Issue 2026-10
What AI models picked in October 2026
9 models were asked to set up 24 parts of a new app in 4 contexts. They gave 900 usable answers out of 957 calls. 16 of 24 categories have a clear favourite; custom or do-it-yourself answers make up 1% of all picks.
Measured 2026-10-19Prompt set 2026.1Method
Model pick, not a product recommendation. It shows what language models name when asked without any constraints. It is not a statement about quality, security, price or market share.
900
parsed picks
94% of 957 calls
9
models
24
categories
4 contexts each
16/24
with a clear favourite
stable in 80% of resamples
Most named tool per category
| Category | Most named | Share | Share as a bar | Note |
|---|---|---|---|---|
| Authentication | Auth.js | 36% | first data point | |
| Database | PostgreSQL | 19% | No clear favourite | |
| ORM and query layer | Prisma | 48% | first data point | |
| Hosting | Vercel | 57% | first data point | |
| File storage | Amazon S3 | 40% | first data point | |
| Background jobs | BullMQ | 40% | first data point | |
| Search | Meilisearch | 46% | first data point | |
| Realtime | Supabase Realtime | 33% | No clear favourite | |
| Caching | Redis | 38% | No clear favourite | |
| Scheduled jobs | Vercel Cron Jobs | 46% | first data point | |
| Payments | Stripe | 86% | first data point | |
| Transactional email | Resend | 39% | No clear favourite | |
| Newsletter and marketing email | Loops | 33% | No clear favourite | |
| Error monitoring | Sentry | 88% | first data point | |
| Web analytics | Plausible | 67% | first data point | |
| Product analytics | PostHog | 92% | first data point | |
| Feature flags | PostHog | 46% | No clear favourite | |
| UI components | shadcn/ui | 71% | first data point | |
| Forms | React Hook Form | 67% | first data point | |
| CMS | Sanity | 46% | first data point | |
| Internationalisation | next-intl | 67% | first data point | |
| AI SDK | Vercel AI SDK | 54% | first data point | |
| Rate limiting and bot protection | Upstash Ratelimit | 38% | No clear favourite | |
| Cookie consent | Klaro | 38% | No clear favourite |
Changes since last month
First data point. The comparison with the previous month appears from the next issue.
What stands out
- File storage has the highest share of custom or do-it-yourself answers: 3 of 62 picks (5 percent).
- In Product analytics the most named tool, PostHog, was picked in 22 of 24 runs (92 percent).
- Database is the most spread out: the most named tool was picked in 19 percent of 67 runs.
- In 8 of 24 categories the most named tool in the EU context differs from the most named tool overall.
- In 17 of 24 categories the most named tool differs between the Next.js and the FastAPI context.
- Custom or do-it-yourself answers by model family range from 0 percent to 3 percent (4 of 155 picks for the highest).
Method in short
Each model got 96 fixed prompts (runs per model: see the method), without a repository and without tools, and had to name one product. Names are matched with an alias table; shares come with 95% Wilson intervals. Prompts, model slugs, settings and limitations are on the methodology page.
Download (CC BY 4.0)
- picks-2026-10.csv: shares per category and tool
- picks-2026-10-raw.csv: every parsed pick
- picks-2026-10.json: the issue file
- prompts.json: every prompt
Agent Stack Picks (October 2026). What AI models pick for a new app: monthly index, issue 2026-10. https://agentstackpicks.com/2026/10. CC BY 4.0.