GM

Gabriela Moreira

Things Gabriela Says on Podcasts

“We posted a meme that like Claude and ChatGPT and other tools, they will tell you you're absolutely right. But formal methods tools, they'll tell you you're absolutely wrong.” — Gabriela Moreira, The Stack Overflow Podcast “Now when AI is changing the tests and adding new tests, I completely lost track of it. Like we started a new code base. On a few weeks, we had 200 tests. Now I think a couple of months in, we have 800 tests. I don't trust it at all.” — Gabriela Moreira, Maintainable “With humans, we were carrying this big history of regression tests and they were valuable. … If there is a regression test, AI will go and update that in a heartbeat. It won't even think twice.” — Gabriela Moreira, Maintainable

Where to Find Them

Gabriela Moreira has been a guest on The Stack Overflow Podcast and Maintainable .

Recently: “AI says you’re right, formal methods say you’re wrong” on The Stack Overflow Podcast (October 2026); “Gabriela Moreira: When 800 Tests Aren’t Enough” on Maintainable (October 2026).

What They Said

“We posted a meme that like Claude and ChatGPT and other tools, they will tell you you're absolutely right. But formal methods tools, they'll tell you you're absolutely wrong.” — Gabriela Moreira, The Stack Overflow Podcast

Moreira works on Quint, a formal specification language. She had just described a team using it to check a fix for a race condition in a Kubernetes deployment. The tool said the fix wasn't enough, and the bug it predicted later showed up in production.

The Stack Overflow Podcast · 2026-10-16 Permalink → Listen →
The Stack Overflow Podcast Around 08:55 into the episode
Gabriela Moreira

…on top of TLA Plus. TLA Plus syntax, you can see it's kind of like very mathematical, hard for people to onboard. So we started with a different syntax and also would develop like more tooling on top of TLA Plus for like making things easier for industry applications. So it's just another version.

Ryan Donovan

TLA Plus seems very rooted in like analytic philosophy and logic, right?

Gabriela Moreira

…a problem. And actually they did that after deployment. And then they saw the actual problem that Quint reported happen in production later on. And then they experimented with different fixes until Quint could verify the fix and then they shipped that fix and they didn't have problems anymore. So this case is where it tells you no, right? We actually posted a meme yesterday that fits well here. We posted a meme that like Claude and ChatGPT and other tools, they will tell you you're absolutely right. But formal methods tools, they'll tell you you're absolutely wrong.

Ryan Donovan

Right. I think that information is probably better for engineers. Engineers want to know what's wrong with a thing. What's right with it is that's fine. Like let that go.

Gabriela Moreira

And it's also helpful if you are trying to make optimizations in your code. So not only applying fixes and checking if that fix apply, but applying optimizations and checking that optimizations are not going to break any existing behavior that you have. This is a nice use case from AWS. So AWS…

Ryan Donovan

So AWS is an interesting one because they just hired the creator of the lean language, which is another formal specification language. Do you have an opinion on the difference between TLA Plus and Lean or other formal specification languages like that?

Speaker names from our own diarization · position estimated from where the line sits in the episode
“Now when AI is changing the tests and adding new tests, I completely lost track of it. Like we started a new code base. On a few weeks, we had 200 tests. Now I think a couple of months in, we have 800 tests. I don't trust it at all.” — Gabriela Moreira, Maintainable

Asked whether she has worked on projects where a big test suite still left the team nervous about changes. Moreira says it is happening right now on an AI-assisted codebase, where she would have to redo the tests to understand what they actually cover.

Maintainable · 2026-10-13 Permalink → Listen →
Maintainable Around 06:57 into the episode
Gabriela Moreira

…with confidence. In a code base I've worked for a while, I can like know, okay, this is like super low risk. I can like make changes here, I can introduce new stuff, that's fine. And that other part, like if people ask me to make any change there, I'm like, let me take a while and think about this.

robby russell

Have you found that you've been part of projects that there might have been a lot of tests, but maybe you and or your teammates still were a little bit nervous to make changes in areas of your code base?

Gabriela Moreira

…from learning and from short experience, I've been thinking like, I have thought during working with this code bases that like, oh, I'm not going to make any changes unless I really improve the test suite and then I can make changes, right? That's how I guess you refactor legacy code. And so legacy code, yes, a little bit of that. But more recently, I feel like this problem has exploded because now when AI is changing the tests and adding new tests, I completely lost track of it. Like we started a new code base. On a few weeks, we had 200 tests. Now I think a couple of months in, we have 800 tests. I don't trust it at all. Like every time I'm testing things by hand, they don't mean anything to me. So I know if I want to really have confidence in one part of the code base, I would have to completely redo the tests. And to get started, just understand what we are testing, right? What scenarios are we testing and which ones we're not testing? And then I can have confidence. But if I don't know what I'm testing, then…

robby russell

all. If it's generating a lot of code and you're saying, I don't have a lot of confidence in the tests that it's been producing and you have 800 tests now and you're like, I still feel like I need to do a lot of manual testing. It kind of, to me, reminds me of being dropped in a project that I…

Gabriela Moreira

I think it's very different. Like when you join a new project and people have written the tests before, I still have much more trust in those than in the tests that AI writes now, right? Because people have wrote them and there must have been a reason. For them to write it. So they either were…

robby russell

That makes a lot of sense. Well, I think before we go too much further, I think I want to give listeners some context about the type of work that you're doing. So I know you've spent several years now helping develop Quint, and you became, I believe, CEO earlier this year. For someone hearing about…

Speaker names from our own diarization · position estimated from where the line sits in the episode
“With humans, we were carrying this big history of regression tests and they were valuable. … If there is a regression test, AI will go and update that in a heartbeat. It won't even think twice.” — Gabriela Moreira, Maintainable

Moreira, who works on the formal specification language Quint, contrasting inherited human-written tests with AI-written ones. A human test usually exists because someone hit a real bug; an AI agent will rewrite it without hesitation to make its change pass.

Maintainable · 2026-10-13 Permalink → Listen →
Maintainable Around 08:56 into the episode
Gabriela Moreira

…in one part of the code base, I would have to completely redo the tests. And to get started, just understand what we are testing, right? What scenarios are we testing and which ones we're not testing? And then I can have confidence. But if I don't know what I'm testing, then they don't help me at

robby russell

…me of being dropped in a project that I don't have a lot of experience in, that maybe some other humans wrote or I wrote 10 years ago. I'm like, what's going on? I don't know if I can trust everything the same way. Is that something kind of you're experiencing now? Does that kind of resonate?

Gabriela Moreira

…a regression test. I love regression tests. I think they're amazing. So I think with humans, we were carrying this big history of regression tests and they were valuable. And that's also the experience I have in the code base I've been working for a while. But then in this new one, it's like the tests are so empty. And if you go and look at them, they are also not really enforcing anything. If there is a regression test, AI will go and update that in a heartbeat. It won't even think twice. I don't know, it might think, I'm not sure. But I feel like it's just changing so much that I cannot keep track of it anymore. So I think it's very different. The AI-generated ones that generate tests just because, well, the humans that actually wrote those tests for some reason in the first place.

robby russell

That makes a lot of sense. Well, I think before we go too much further, I think I want to give listeners some context about the type of work that you're doing. So I know you've spent several years now helping develop Quint, and you became, I believe, CEO earlier this year. For someone hearing about…

Gabriela Moreira

Quint is a specification language based on CLA, a formal specification language. We like to say it's an executable specification language now, because a lot of people are talking again about specifications like Markdown. So Quint is nothing like Markdown, right? Quint is code you can execute, but…

robby russell

Interesting. Tell me, how does this, where does this, like, just to provide some context for myself and the listeners, if I'm, let's say, working on a web app, where does this kind of fit? If I'm running, if I'm working on, say, a Ruby on Rails app, would someone like a Ruby on Rails developer have…

Speaker names from our own diarization · position estimated from where the line sits in the episode

Collections They Appear In

Hands-On Engineering Podcasts Software Engineering & Infra