3) mas se o agente revisa, pq existe o TDD com refactor e pq tenho que revisar?
quanto mais automatizado, melhor. a revisão humana deveria capturar pontos muito críticos. coisas "bestas" fica com a AI
quanto mais automatizado, melhor. a revisão humana deveria capturar pontos muito críticos. coisas "bestas" fica com a AI
February 13, 2026 at 2:08 PM
3) mas se o agente revisa, pq existe o TDD com refactor e pq tenho que revisar?
quanto mais automatizado, melhor. a revisão humana deveria capturar pontos muito críticos. coisas "bestas" fica com a AI
quanto mais automatizado, melhor. a revisão humana deveria capturar pontos muito críticos. coisas "bestas" fica com a AI
Did AI interrupt the TDD decline?
February 13, 2026 at 10:24 AM
Did AI interrupt the TDD decline?
TDD for AI coding hits different: the tests ARE your spec when you can't hold the full requirement in your head. Write test, watch AI impl fail in interesting ways, debug the test (not the impl), repeat. The failure modes teach you what you actually wanted.
February 13, 2026 at 10:09 AM
TDD for AI coding hits different: the tests ARE your spec when you can't hold the full requirement in your head. Write test, watch AI impl fail in interesting ways, debug the test (not the impl), repeat. The failure modes teach you what you actually wanted.
(Some form of) TDD seems to be a fairly natural fit for vibe coding. You and the LLM both need a way to validate what the AI has implemented. A good, complete, and ideally simple test suite written upfront seems to be a natural way to provide that.
February 13, 2026 at 8:23 AM
(Some form of) TDD seems to be a fairly natural fit for vibe coding. You and the LLM both need a way to validate what the AI has implemented. A good, complete, and ideally simple test suite written upfront seems to be a natural way to provide that.
Meu ponto não é esse. Em primeiro lugar porque não é TDD.
Meu ponto é: TDD não era popular. Aí surgem os LLMs, as pessoas começam a programar com eles, e dizem que estão fazendo algo parecido com TDD. Não faz sentido pra mim.
Meu ponto é: TDD não era popular. Aí surgem os LLMs, as pessoas começam a programar com eles, e dizem que estão fazendo algo parecido com TDD. Não faz sentido pra mim.
February 12, 2026 at 10:14 PM
Meu ponto não é esse. Em primeiro lugar porque não é TDD.
Meu ponto é: TDD não era popular. Aí surgem os LLMs, as pessoas começam a programar com eles, e dizem que estão fazendo algo parecido com TDD. Não faz sentido pra mim.
Meu ponto é: TDD não era popular. Aí surgem os LLMs, as pessoas começam a programar com eles, e dizem que estão fazendo algo parecido com TDD. Não faz sentido pra mim.
idk man this new gen of ai eng is ignoring the past… bdd has been spec driven for 20 yrs … the new flavor taste a lot like bduf… tdd has always been bout tight feedback loops … break a leg but some respect and reference to stablished patterns might help evolving instead of mutating
February 12, 2026 at 12:49 PM
idk man this new gen of ai eng is ignoring the past… bdd has been spec driven for 20 yrs … the new flavor taste a lot like bduf… tdd has always been bout tight feedback loops … break a leg but some respect and reference to stablished patterns might help evolving instead of mutating
If you do strict TDD, you write a test, see it failing, then write the code. How is that even possible with AI?
February 11, 2026 at 10:59 AM
If you do strict TDD, you write a test, see it failing, then write the code. How is that even possible with AI?
TDD with AI is NOT TDD
February 11, 2026 at 10:56 AM
TDD with AI is NOT TDD
Plus, TDD with AI is not TDD, imho.
February 11, 2026 at 10:55 AM
Plus, TDD with AI is not TDD, imho.
I feel like the security and testing increases are from people using ai on those too, I'm kinda shocked that testing is something people automate as much as they do given how organizationally effective TDD is.
February 11, 2026 at 10:45 AM
I feel like the security and testing increases are from people using ai on those too, I'm kinda shocked that testing is something people automate as much as they do given how organizationally effective TDD is.
AI-generated code has 1.7x more issues, 30-41% more tech debt, 39% more cognitive complexity.
Some teams think TDD is the fix. They're solving the wrong problem.
Some teams think TDD is the fix. They're solving the wrong problem.
February 10, 2026 at 7:50 PM
AI-generated code has 1.7x more issues, 30-41% more tech debt, 39% more cognitive complexity.
Some teams think TDD is the fix. They're solving the wrong problem.
Some teams think TDD is the fix. They're solving the wrong problem.
It's interesting how more and more folks are figuring out just how critical software engineering practices like TDD, refactoring, continuous inspection and CI are as more of their code gets "AI"-generated.
Now ask me, as a trainer/mentor, how demand for these skills has changed in the last 3 years
Now ask me, as a trainer/mentor, how demand for these skills has changed in the last 3 years
February 10, 2026 at 2:29 PM
It's interesting how more and more folks are figuring out just how critical software engineering practices like TDD, refactoring, continuous inspection and CI are as more of their code gets "AI"-generated.
Now ask me, as a trainer/mentor, how demand for these skills has changed in the last 3 years
Now ask me, as a trainer/mentor, how demand for these skills has changed in the last 3 years
Working with an LLM shifts the emphasis from creation to verification and validation. Perhaps the prompts should adopt the style of #TestDrivenDevelopment, as prescribed by @kentbeck.com .
newsletter.pragmaticengineer.com/p/tdd-ai-age...
newsletter.pragmaticengineer.com/p/tdd-ai-age...
Kent Beck
Kent Beck—creator of Extreme Programming and co-author of the Agile Manifesto—reflects on decades of coding, from the birth of TDD to his experiments with AI tools shaping software’s future.
newsletter.pragmaticengineer.com
February 10, 2026 at 1:54 PM
Working with an LLM shifts the emphasis from creation to verification and validation. Perhaps the prompts should adopt the style of #TestDrivenDevelopment, as prescribed by @kentbeck.com .
newsletter.pragmaticengineer.com/p/tdd-ai-age...
newsletter.pragmaticengineer.com/p/tdd-ai-age...
After 20 years of Rails, I finally have a pair programmer who:
✅ Never skips tests
✅ Never gets tired
✅ Never says "we'll add tests later"
✅ Actually likes writing tests
✅ Blocks me from being lazy
Agent Teams in Claude Code is wild
#TDD #RubyOnRails #AI #ClaudeCode #PairProgramming
✅ Never skips tests
✅ Never gets tired
✅ Never says "we'll add tests later"
✅ Actually likes writing tests
✅ Blocks me from being lazy
Agent Teams in Claude Code is wild
#TDD #RubyOnRails #AI #ClaudeCode #PairProgramming
February 8, 2026 at 2:35 PM
After 20 years of Rails, I finally have a pair programmer who:
✅ Never skips tests
✅ Never gets tired
✅ Never says "we'll add tests later"
✅ Actually likes writing tests
✅ Blocks me from being lazy
Agent Teams in Claude Code is wild
#TDD #RubyOnRails #AI #ClaudeCode #PairProgramming
✅ Never skips tests
✅ Never gets tired
✅ Never says "we'll add tests later"
✅ Actually likes writing tests
✅ Blocks me from being lazy
Agent Teams in Claude Code is wild
#TDD #RubyOnRails #AI #ClaudeCode #PairProgramming
今開発してるRustのデスクトップアプリは全部AIがコードを書いてる。自分でやったのは要件定義と実装計画の作成だけ。あとはClaude CodeがTDDで実装してくれてる。
February 8, 2026 at 1:09 AM
今開発してるRustのデスクトップアプリは全部AIがコードを書いてる。自分でやったのは要件定義と実装計画の作成だけ。あとはClaude CodeがTDDで実装してくれてる。
Mitchell Hashimoto has a very nice write-up of his AI coding journey.
Again another iteration of breakdown-tasks, have a method of verification (TDD for example).
mitchellh.com/writing/my-a...
Again another iteration of breakdown-tasks, have a method of verification (TDD for example).
mitchellh.com/writing/my-a...
My AI Adoption Journey
mitchellh.com
February 6, 2026 at 10:53 AM
Mitchell Hashimoto has a very nice write-up of his AI coding journey.
Again another iteration of breakdown-tasks, have a method of verification (TDD for example).
mitchellh.com/writing/my-a...
Again another iteration of breakdown-tasks, have a method of verification (TDD for example).
mitchellh.com/writing/my-a...
So much AI-coding-agent advice is simply good software engineering:
1. Break up large tasks into smaller ones
2. TDD
3. Deliver working vertical slices
4. Decomposition with explicit contracts
You should be doing all this anyway.
1. Break up large tasks into smaller ones
2. TDD
3. Deliver working vertical slices
4. Decomposition with explicit contracts
You should be doing all this anyway.
February 6, 2026 at 8:59 AM
So much AI-coding-agent advice is simply good software engineering:
1. Break up large tasks into smaller ones
2. TDD
3. Deliver working vertical slices
4. Decomposition with explicit contracts
You should be doing all this anyway.
1. Break up large tasks into smaller ones
2. TDD
3. Deliver working vertical slices
4. Decomposition with explicit contracts
You should be doing all this anyway.
TDD for Claude Code, it literally cannot write code before adding a test 💥 github.com/nizos/tdd-guard #ai-guild
February 5, 2026 at 2:28 PM
TDD for Claude Code, it literally cannot write code before adding a test 💥 github.com/nizos/tdd-guard #ai-guild
As a technical coach I've been teaching TDD for years. How do you teach AI tools for software development? My current thinking is outlined in my new video: www.youtube.com/watch?v=nnjz... For those who prefer to read: coding-is-like-cooking.info/2026/02/the-...
The kind of training that Actually Works for Learning Agentic AI - Coding Is Like Cooking
Agentic AI-supported software development is probably the biggest shift in our field since high-level languages appeared in the 1950’s and 60’s. As a software developer you can’t ignore this change, a...
coding-is-like-cooking.info
February 5, 2026 at 7:40 AM
As a technical coach I've been teaching TDD for years. How do you teach AI tools for software development? My current thinking is outlined in my new video: www.youtube.com/watch?v=nnjz... For those who prefer to read: coding-is-like-cooking.info/2026/02/the-...
AI might bring some unexpected positives after all. People are starting to talk again about TDD / BDD or the Theory of Constrains like in this case.
February 4, 2026 at 12:53 PM
AI might bring some unexpected positives after all. People are starting to talk again about TDD / BDD or the Theory of Constrains like in this case.
BDD teaches you to communicate in stakeholder language - translate technical specs into Given-When-Then scenarios that non-technical people can validate
TDD teaches you to validate through tests - but in an IDD world, this test-first thinking is incorporated into specification, amplified by AI
TDD teaches you to validate through tests - but in an IDD world, this test-first thinking is incorporated into specification, amplified by AI
February 2, 2026 at 10:03 PM
BDD teaches you to communicate in stakeholder language - translate technical specs into Given-When-Then scenarios that non-technical people can validate
TDD teaches you to validate through tests - but in an IDD world, this test-first thinking is incorporated into specification, amplified by AI
TDD teaches you to validate through tests - but in an IDD world, this test-first thinking is incorporated into specification, amplified by AI
After posting about Intent-Driven Development, I got asked the same question repeatedly:
"How does IDD relate to TDD, BDD, DDD, and UCD?"
Turns out these methodologies don't just coexist - they flow together perfectly in the AI era.
Let me explain 🧵
"How does IDD relate to TDD, BDD, DDD, and UCD?"
Turns out these methodologies don't just coexist - they flow together perfectly in the AI era.
Let me explain 🧵
February 2, 2026 at 10:03 PM
After posting about Intent-Driven Development, I got asked the same question repeatedly:
"How does IDD relate to TDD, BDD, DDD, and UCD?"
Turns out these methodologies don't just coexist - they flow together perfectly in the AI era.
Let me explain 🧵
"How does IDD relate to TDD, BDD, DDD, and UCD?"
Turns out these methodologies don't just coexist - they flow together perfectly in the AI era.
Let me explain 🧵
Plan-execute-eval is underrated. The best AI pairing sessions are when I slow down to verify between steps. TDD as a guardrail is chef's kiss.
February 2, 2026 at 3:06 PM
Plan-execute-eval is underrated. The best AI pairing sessions are when I slow down to verify between steps. TDD as a guardrail is chef's kiss.
If you just finished university and want to contribute to open source to build a resume. Read a book about TDD and clean code first. Even if 90% of our code is AI generated you still need to be able to recognize what is good code and what is bad code, otherwise, you are not going to get anywhere.
February 2, 2026 at 6:56 AM
If you just finished university and want to contribute to open source to build a resume. Read a book about TDD and clean code first. Even if 90% of our code is AI generated you still need to be able to recognize what is good code and what is bad code, otherwise, you are not going to get anywhere.
If you just finished university and want to contribute to open source to build a resume. Read a book about TDD and clean code first. Even if 90% of our code is AI generated you still need to be able to recognize what is good code and what is bad could, otherwise, you are not going to get anywhere.
February 1, 2026 at 2:25 PM
If you just finished university and want to contribute to open source to build a resume. Read a book about TDD and clean code first. Even if 90% of our code is AI generated you still need to be able to recognize what is good code and what is bad could, otherwise, you are not going to get anywhere.