On-demand webinar · Recorded Oct 29, 2026 · 38 min

AI writes the code. Who checks what hits your database?

Small teams now ship at large-team volume, and review, testing and security didn't scale with them. DBGorilla CEO Chris Lalonde shares four lessons from building an AI-native product from the ground up, including what broke along the way. Then co-founder Stephen Nolan shows how DBGorilla checks AI-written changes against production facts before they ship.

AI writes the code. Who checks what hits your database? DBGorilla webinar

Recording coming soon

Chris Lalonde and Stephen Nolan, DBGorilla co-founders.

Start freeTry the sandbox

Jump to a section

  1. 0:07Why AI code volume creates big-company problems
  2. 2:38How 40 PRs exposed a review bottleneck
  3. 4:26Assumption drift threatens quality and security
  4. 7:16Build layered checks from the point of creation
  5. 9:51Why AI misses context and can fake test success
  6. 12:44The production-scale bugs local tests cannot reveal
  7. 14:36Smaller agents, preview environments and draft PRs
  8. 17:59Casino coding, scope creep and preserving intent
  9. 23:00The full-stack workflow and lessons learned
  10. 27:29Demo: the DBGorilla advisor finds a costly production query
  11. 30:34Demo: the CI bot catches migration and runtime hazards
  12. 32:25Demo: MCP gives coding agents production context
  13. 37:00Sandbox access and closing remarks

Four lessons from building with AI

LESSON 1

You don't just get faster. You get bigger.

A small team with agents ships at large-team volume and inherits large-team problems, starting with review. Forty pull requests exposed the bottleneck. What changed: agents check syntax and best practice, people review intent.

Watch at 2:38

LESSON 2

Faster development, different failures.

Mistakes travel faster and assumptions drift quietly, and quality and security go with them. What changed: layered checks from the moment code is created, and a person who can stop the line.

Watch at 4:26

LESSON 3

AI wants to make you happy.

It invents things to please you, its context is smaller than you think, and it can make tests pass without proving anything. The bugs that matter only show up at production scale. What changed: reality early, in small chunks, with smaller agents, preview environments and draft PRs by default.

Watch at 9:51

LESSON 4

It's a trap.

AI encourages big changes, and big changes make review and rollback risky and drift from what the ticket asked for. What changed: people hold the scope and the intent.

Watch at 17:59

Then we show it, live.

Stephen walks through three places DBGorilla checks AI-written work against production facts. Every fix is tested on a clone first. Production is never touched.

The advisor

Finds the query that is costing you the most in a demo production database, tests a fix on a clone, and opens a pull request with the before-and-after evidence for a person to review.

Watch at 27:29

The CI bot

Reviews a migration an AI agent wrote and flags what tests miss: a column default that would rewrite a large table under a lock that blocks checkout, an index built without CONCURRENTLY, and a lookup that fails when it runs on a clone.

Watch at 30:34

MCP for coding agents

Gives the agent in your editor the same production context, so it writes against your real schema and data shape instead of guessing.

Watch at 32:25

Who this is for

CTOs, VPs and directors of engineering, platform and DevOps leads, and senior engineers whose teams ship with AI coding tools and run Postgres or MySQL in production.

Speakers

Chris Lalonde

CEO and co-founder, DBGorilla

Stephen Nolan

Co-founder, DBGorilla

See what it finds in your database.

Connect your database, or try the sandbox with no signup. DBGorilla reads, recommends and tests on a clone. A person approves every change.

Start freeTry the sandbox

Questions? founders@dbgorilla.com