Across 115 verified production architectures approved in the STACK IT FAST directory over the past 30 days, TypeScript is the common denominator: 76% (87 of 115) of stacks are built with it, and Docker ships 63% (73) of them. Below the application layer the picture is more varied than in August, with Rust now in 26% (30) of stacks and Go in 23% (26).
1. Database & Persistence Layer: PostgreSQL First, SQLite Rising
PostgreSQL is still the default system of record, but no longer the near-universal one it looked like in a smaller August sample (71% of 48):
- PostgreSQL: 65 projects (57%)
- Redis: 42 projects (37%)
- SQLite: 26 projects (23%)
- MySQL: 12 projects (10%)
- Prisma: 12 projects (10%)
- ClickHouse: 10 projects (9%)
- MongoDB: 6 projects (5%)
SQLite’s 23% share reflects single-binary tools and self-hosted apps that embed their database instead of requiring a server.
2. Frontend Frameworks & Meta-Runtimes: React Leads, Vite Beats Next.js
- React: 54 projects (47%)
- Tailwind CSS: 30 projects (26%)
- Vite: 29 projects (25%)
- Next.js: 19 projects (17%)
- Vue: 15 projects (13%)
- Angular, Svelte and Astro: 4, 3 and 4 projects
Vite (29) now appears in more production stacks than Next.js (19), consistent with dashboards and admin UIs being built as SPAs served by a separate backend. Build orchestration is common too: Turborepo appears in 17 stacks (15%). Model Context Protocol shows up in 9 stacks (8%), usually as a server that exposes the product to AI agents.
3. Development Modalities: Classic vs. AI-Agent Assisted
- Classic: 62 projects (54%)
- Hybrid (AI generation with human review): 36 projects (31%)
- AI-Agent heavy: 17 projects (15%)
- Vibecoded flag: 10 projects (9%)
Combined, 46% (53 of 115) of production architectures ship through an AI-assisted pipeline.
4. Category Distribution: Developer Tools and AI Lead
- Developer tools: 24 projects (21%)
- AI / LLM apps: 17 projects (15%)
- Analytics: 15 projects (13%)
- Data infrastructure: 14 projects (12%)
- SaaS: 11 projects (10%)
Content, collaboration and internal tools follow with 7, 6 and 6 projects.
Sample Size & Methodology Transparency
This report covers 115 architectures approved in the directory in the 30 days before September 25, 2026. Technology names are canonicalized before counting (for example “Postgres” and “PostgreSQL” count once), and a project counts once per technology regardless of how many services use it. Percentages are rounded to whole numbers. The August report used a 48-project window, so month-over-month changes partly reflect the larger and more varied sample rather than projects switching technologies.