Matthew Turley

American in Paris, 3 kids · 20 yrs shipping product · Building @uxcontinuum + @relayplaneai · Running a 6 agent pipeline across 12 repos (personal and client)

[Matthew Turley @matthewrturley]

A founder paid me to scan their AI-built app. 41 issues came back.

What they thanked me for wasn't the 41 issues. It was the line that said: fix these 3 today, ignore the other 35.

AI will hand you 40 red rows. It can't tell you which one is leaking user data right now and which one is a dev warning you can ignore.

The ranking is the whole job.

[Matthew Turley @matthewrturley]

Everyone ranks AI app builders by how fast they spit out a CRUD app with auth.

Wrong thing to measure. Generation is basically free now.

I scan apps from all of them. The holes are identical: RLS off, API keys sitting in the client bundle, admin routes open to anyone who guesses the URL. The builder didn't matter.

Ranking generators in 2026 is like ranking typewriters in 1995. The question is what happens after the app exists.

[Matthew Turley @matthewrturley]

We’re back!

[Anthropic @AnthropicAI]

We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.

We'll begin restoring access tomorrow, and will share an update soon.

We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.

[Matthew Turley @matthewrturley]

Apps built with Lovable, Bolt, and Cursor hand your browser a key that reads the database. Public by design.

What's not by design: you take that key, ask for a table, and it answers. No login.

Checked 66 live ones. 32 on Supabase. 13 let anyone read a table with that key. 5 exposed users, subscriptions, private chats, audit logs.

RLS was off. Grab your key from devtools and try to read your own table. If it answers, anyone else can do the same.

[Matthew Turley @matthewrturley]

@AlansSpace I’ll let you know when I hit $1M. 🤣

[Matthew Turley @matthewrturley]

@elvissun It’s highly variable depending on what you are prompting. Anywhere from 60 to 90% from my internal use

[Matthew Turley @matthewrturley]

Nobody warns you: when agent loops actually work, you drown in PRs.

Agents ship all day across multiple repos now. Reviewing them became the bottleneck, not writing them.

Building a surface to get each review down to ~60s.

[Matthew Turley @matthewrturley]

@aronprins No email here yet, but glad this is confirmed.

https://t.co/VJaRN3G82K

[Matthew Turley @matthewrturley]

This still happening?

Everyone just moving to openai today?

[Matthew Turley @matthewrturley]

This still happening?

Everyone just moving to openai today?

[ClaudeDevs @ClaudeDevs]

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage.

The credit covers usage of:
- Claude Agent SDK
- claude -p
- Claude Code GitHub Actions
- Third-party apps built on the Agent SDK

[Matthew Turley @matthewrturley]

@ClaudeDevs this still happening?

Where's the credit?

[Matthew Turley @matthewrturley]

@chalaska The key is to run 5 sprints at a time 😆

[Matthew Turley @matthewrturley]

@elvissun I've been in this exact hole before.

The cheating part is real, agents will fake done all day and are not good at checking their own work

TDD and a verifier agent fixed this for me, but might give this a try too.

[Matthew Turley @matthewrturley]

Yeah, relevance is the harder half. Being known does nothing if the model can't tell what you're actually good at.

What's worked for me: get specific in public. Not "I do AI consulting" but the exact problem you solved and the number you moved. Gives it something concrete to match a real question against.

[Matthew Turley @matthewrturley]

"How did you find me?"

"My Cursor agent told me you could help."

Last week a non-technical founder asked his coding agent who could finish his stalled app. It sent him to me.

Your customer asks AI, and it either knows you or it doesn't.

Here are 5 steps to get known by AI:

1) Be visible everywhere you live online, all pointing at one thing. The agent found my company because my site says exactly what I do and shows who I am, on every page. Plain and specific, not a vague "we do digital." The model needs a clear thing to point at.

2) Be specific about the problem you solve. Not "AI consultant." "The guy who finishes broken vibe-coded projects." The model can't recommend a blur. It needs a clean line from a problem to a person.

3) Answer the same real questions in public, over and over. Reddit, forums, X. That's what these models read and pull from. Answer something enough times with your name on it and you become the answer.

4) Keep your identity the same everywhere. Same name, same one-liner, same problem, on your site, GitHub, LinkedIn. The model builds one picture of you.

5) Write so a machine can quote you. Put the answer in the first line. Let each section stand on its own. Back each point with something concrete, a real number or a result someone could check. Not a soft claim. The specific bits are what get pulled into the answer.

None of this is a hack. It's just being findable for one specific thing. Do that and the agent does the selling for you... while you sleep.

[Matthew Turley @matthewrturley]

The people with these yachts aren’t writing prompts.

They’re writing the loops that write the prompts.

[Matthew Turley @matthewrturley]

https://t.co/RRvp34R51w

[Matthew Turley @matthewrturley]

If you are a non-tech founder about to hire someone to build AI into your business: 5 questions to ask on the first call, and the red flags that should end it early.

Wrote it up from the building side. 👇