Matt Matheson

Co-Founder @trunkio. Democratizing excellent DevEx.

[Matt Matheson @MatthewMathes11]

The more you push into layers 1 and 2, the faster your humans get through layer 3.
That's how you ship at AI pace without cutting corners on quality.
Are you using AI reviewers yet, or still relying on humans for everything?

[Matt Matheson @MatthewMathes11]

The answer isn't to skip reviews or rubber-stamp PRs.
It's to automate everything that can be automated so humans focus on what actually requires a human.
We've been using @claudeai as an AI reviewer, and I now think about code review as three layers:

[Matt Matheson @MatthewMathes11]

Layer 3: Humans for alignment & intent.
Does this change match the product spec? Is this the right architectural direction? Should we even be building this?
Judgment, context, and strategy. The highest-leverage use of a senior engineer's review time.

[Matt Matheson @MatthewMathes11]

AI-generated code has completely changed the rate (and size) of PRs I need to review.
Code review is quickly becoming our team's bottleneck.
Here's the framework we're using at @trunkio to fix it 🧵

[Matt Matheson @MatthewMathes11]

The AI revolution has been amazing for writing code.
Now we need to upgrade every other part of the pipeline to match.
That's what we're building at @trunkio.
What's your team doing to keep shipping fast without cutting corners on quality?

[Matt Matheson @MatthewMathes11]

AI is making code generation 10x faster, but the rest of the dev lifecycle still feels stuck in 2019.
A thread on why our tooling is buckling under AI-paced output 🧵

[Matt Matheson @MatthewMathes11]

Test selection. Intelligent sharding. Flaky test quarantining. Fast merge queues.
These used to be FAANG-only.
Now they're table stakes for everyone.

[Matt Matheson @MatthewMathes11]

@samcook_ Do you have hardware acceleration enabled? What Mac hardware are you using?

[Matt Matheson @MatthewMathes11]

@nbrempel 👋 Hi Nick, it's very effective at mitigating logical merge conflicts (https://t.co/g6gB1uIxhg) for teams. Are you looking for a mergequeue for @strobe_app?

[Matt Matheson @MatthewMathes11]

@housecor These are all really good advice but nothing is free - automated testing that actually is good enough to be a release validator is a lot of effort.
Feature flags introduce process overhead and obscure the answer to “did we ship x”?

[Matt Matheson @MatthewMathes11]

@anthonysheww - monorepos must be able to deploy apps individually.
- teams work better in a monorepo: consistent tooling, build, ci, etc allow people to contribute everywhere. Less “my code” more “our code”
- monorepos require more tooling but imo that always pays off