I replaced my entire tech stack with Postgres...

Overview

This video argues that modern web development can become overcomplicated and tool-heavy, causing teams to pay for many separate “shovels” (services like caches, cron, auth, monitoring, analytics, etc.).

The premise is that you can replace much of the stack with PostgreSQL alone, leveraging its extensibility and built-in features, and then augmenting it with select PostgreSQL extensions.

Key caution: “Just because you can do something doesn’t mean you should”—choose the right tool for the job.

What Makes PostgreSQL “Special” (as stated)

• Advanced data types out of the box
  • Includes binary, JSON/JSONB, arrays, key-value-like patterns, and geometric types.
• Extensible architecture
  • A large ecosystem of extensions can add new capabilities—likened to “modding” a game.
• Pragmatic guidance
  • The creator warns against using capabilities just because they’re available.

“11 Useful Unorthodox and Weird Ways” to Use Postgres

1. Handle semi-structured/unstructured data with JSONB

   • Store raw JSON in a JSONB column.
   • Query inside JSON using SQL filters/access operators (claim: built-in support for structured querying of JSON data).

2. Spin up hosted Postgres using Neon (sponsor)

   • Neon provides Postgres on a serverless platform with:
     • free tier
     • developer-friendly guides for major frameworks
     • branching for dev/test/prod workflows
     • automatic scaling with “zero effort”

3. Cron jobs without separate scheduling infrastructure

   • Use the PG_cron extension.
   • Schedule SQL statements (e.g., daily delete/aggregation jobs) directly via SQL.

4. “Poor man’s Redis” using tables/cache tactics

   • Use unlogged tables as a cache-like mechanism.
   • Rationale:
     • unlogged tables avoid WAL durability cost → better performance
     • trade durability for speed
   • Mentions tuning Postgres config (e.g., shared buffers) and using auto-vacuum, optionally with TTL cleanup via cron.

5. AI / RAG-style vector storage and processing

   • PGVector: store vector embeddings and run similarity search (nearest neighbors via distance such as L2).
   • PGAI: handle embeddings/vectorization workflows inside SQL (load/vectorize datasets via SQL).

6. Full-text search with typo-tolerant style querying

   • Use TSVector plus inverted indexes (described as generalized inverted indexing).
   • Use search operators to rank results (double-app/operator mentioned).

7. Turn Postgres into a GraphQL API

   • PG GraphQL extension:
     • exposes the database as a GraphQL API
     • “write GraphQL resolvers in SQL”
     • claims no extra server, libraries, or paid middleware required

8. Real-time data sync without WebSockets: ElectricSQL (library, not a Postgres extension)

   • ElectricSQL is described as a sync layer between the database and frontend.
   • Goal: keep the UI updated with fresh data automatically, avoiding manual websocket/re-fetch logic.
   • Mention: works well with Neon.

9. Roll your own “auth” by combining extensions (crypto + JWT + RLS)

   • PGCrypto:
     • hash passwords
     • generate verification values
   • PGJWT:
     • cryptographically sign JSON Web Tokens on the server
   • Row-Level Security (RLS):
     • enforce that users can only read/write rows they own
     • JWT validated during query execution

10. Analytics / time-series needs using a Postgres extension

    • PGmoonCake (spelled “PG moon cake” in subtitles):
      • turns Postgres into a time series database
      • adds column store tables
      • uses DuckDB execution
      • supports dropping/visualizing with tools like Grafana

11. Expose your DB as an API and serve UI assets from the DB

    • PostREST:
      • automatically turns Postgres into a REST API
      • browse endpoints like localhost/<table> to get JSON
      • supports filtering/pagination/auth (mentioned generally)
    • “Full stack” idea: store HTML/CSS/JS inside the database.
    • Mentions experimentation such as React Server Components running in/through Postgres (subtitles suggest extreme co-location; phrasing is unclear).

Overall Message / Conclusion

• The creator’s thesis: by leaning on Postgres features and extensions, you can avoid many separate backend services and reduce stack sprawl.
• Sponsor note: Neon appears as a callout, even though the creator claims Neon “doesn’t approve” of all the “crazy stuff,” while still sponsoring the segment.

Main Speakers / Sources

• Speaker: The video creator (Fireship-style narration; exact name not given in subtitles)
• Sponsored source: Neon (PostgreSQL hosting platform)