Gil Zilberfeld (TestinGil)

CTO & owner TestinGil.
SW dev & testing coach, helping agile professionals
Author & International keynote speaker.

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

How is a code review like candy?
The belt speeds up, the chocolates get hidden, and the supervisor yells "Speed it up!"
Don't be the supervisor. Fix the line.
Free webinar, Nov 11. Full post on LI:
https://t.co/wazhESTO7E
https://t.co/8CjctaKeay
#CodeReview #AICoding

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

@kuroshiro01 Nope. The automation is nice, but most of the work is thinking. Every bit that I wanted to automate was either scripted (which I trust more) or left to Mr. Claude.

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

I built an AI alter-ego with Claude Code that runs my business with me. Talk recording: what I planned, built, broke, and still maintain.
It's in Hebrew. Want it in English? Tell me.
https://t.co/wNJkFJSSGN
#ClaudeCode #AIAgents #BuildInPublic

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

It's not easy being green.
Your build dashboard says everything's ok. The customer complaints say otherwise.
Does the dashboard lie, or we just don't understand "green"?
Full post on LI:
https://t.co/eRKd60dl7Q
#ClaudeCode #AIAgents #BuildInPublic #Productivity #SoloFounder

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

In Hell's Kitchen, Gordon Ramsay shows exactly how to make a dish. Then the contestants try. They get it wrong every time.
Explaining is not enough. Same with coding agents.
New blog:
https://t.co/ThX44SnYHZ
#AIGeneratedCode #CodeQuality #AIAgents

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

AI reverses the polarity. (Always wanted to say that.)
Instead of adapting to tools, we build tools that fit how we work. I built a whole system that way.
Talk (in Hebrew), Thu Oct 8, 11:00 Israel time →
https://t.co/2mU5rQXXCZ
#ClaudeCode #AIAgents #BuildInPublic

[BrowserStack @browserstack]

Red, green, and the step everyone quietly skips.

Gil Zilberfeld on why refactoring is the brilliance of TDD: once you have something working with evidence, you can play with the code safely, because you know you are not breaking anything. Refactor for the next test, and for the version of you reading this code in six months.

Full episode of #BrowserStackTalks: https://t.co/mbJj0mN5iJ

@gil_zilberfeld

[Software Testing Weekly 🚀 @QANewsletter]

Hey! 👋
⠀
The 329th issue is out!
https://t.co/zlgZQgnKFw
⠀
Congrats @jamesmarcusbach, @ailuj876, @michlbrmly, @gil_zilberfeld, @o_kwasniewski, @jarbon, @reallymello, @alexusadays and more!
⠀
#SoftwareTesting #QA

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

"You are responsible for our 2.5 hour shutdown in production."
"It wasn't me. Claude wrote the code, the tests, and promised it'll be ok."
"Do you know what the code really does?"
"Sort of."
New video →
https://t.co/q6oD8GbX76
#AITesting #FlakyTests #AIGeneratedCode

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

Ever signed a 28-page legal form after reading every line?
Now imagine reading that ten times a day. In code font. Full of "oopsies" and "I apologize", because code agents love to tell a story.
New blog:
https://t.co/5uNct5zC9M
#AITesting #AIGeneratedCode #CodeQuality #AIAgents

[BrowserStack @browserstack]

"A request goes in, a status code comes back." If that is your mental model of an API, you have blind spots.

Gil Zilberfeld on why APIs deserve more respect: headers, tokens, context, admin vs regular user, soft deletes, retries, queues. Your test said 201. Did it actually add the book? Can someone else delete it?

Full episode on #BrowserStackTalks: https://t.co/mbJj0mN5iJ

@gil_zilberfeld

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

@Mr_Munjal True. The chance of that happening is lower if you actually write the test, rather than generate it.

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

We nagged developers for years to give us tests. Any proof the code works.
Thanks to AI, we got what we asked for. The bugs are still in there.
New blog →
https://t.co/QZzScRrAkX
#AITesting #SoftwareTesting #AIGeneratedCode #CodeQuality #UnitTesting

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

I make coffee with a scale and a moka pot, and I watch the flow. It works because there's a controller paying attention.
With code agents, we gave up our controller role.
That's how slop comes out.
Full post on LI:
https://t.co/kSKTFecvuZ

[BrowserStack @browserstack]

"I can't pass this test, therefore I can delete it."

Gil Zilberfeld on why AI code agents cut both ways: they will happily generate loads of code and loads of tests, and we have quietly outsourced both the writing and the reviewing. TDD's answer is very small steps, so you stay in control of what you are creating.

Full episode on #BrowserStackTalks: https://t.co/mbJj0mN5iJ

@gil_zilberfeld

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

Risks, risks. Anything good to say about AI?
Sure. Intent used to live in your head - docs, specs, sometimes coffee. Now you type it to an agent, so it can be captured, recorded, tracked.
The rest of the risks are real. Webinar recording →
https://t.co/hhSDtI3f7K
#AITesting

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

I warn people about the risks of AI-generated code for a living.
A manager told me: we don't have those problems. The devs have a rule. Turned out to be a good one.
Full post on LI:
https://t.co/XEKO3dScH7

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

Self-healing tests: what they are, how they magically heal, and whether you can build your own.
On the BrowserStack podcast with David Burns. Plus TDD, clean code, and where AI helps - and where it frightens.
https://t.co/3RJMCKUrYL
#AITesting #TDD #CleanCode #SoftwareTesting

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

Automation is only worth something if it's repeatable.
The CI runs our tests, while the ground shifts while the tests pass.
But today's green isn't as green as yesterday's. Or tomorrow's.
Full post on LI:
https://t.co/8IBJQoRqbv

[Gil Zilberfeld (TestinGil) @gil_zilberfeld]

Who's afraid of the big bad AI?
Turns out a lot of you. I asked what worries you about AI-generated code: "Too much unreviewed code." "Unwanted AI additions."
It gets scarier. I'll show you at my webinar.
Tomorrow, 3PM CEST / 9AM EDT →
https://t.co/3BGKwcf4Cv
#AITesting