Senior Laravel Developer
Joined August 2012
- Tweets 1,224
- Following 409
- Followers 335
- Likes 1,807
55 Photos and videos
Apr 27
If you feel smarter every time you use Claude, you might be getting dumber.
I was.
The brain optimizes for efficiency. Remove friction, and it stops thinking.
The AI that saves you mental effort makes you dumber. The AI that creates mental effort for you makes you sharper.
guetteluis.com/writing/is-cl…
47
Apr 24
46 issues after using my version of /grill-me /write-prd and /prd-to-issues. 7 PRDs, 39 slice issues. Now, implementing using Ralph loop and let's see how it goes.
59
Apr 23
Spent the last few weeks getting Claude Code to work on my Laravel backlog while I'm AFK. Docker sandbox, GitHub issues as the queue, a dumb while loop that exits when there's nothing left tagged afk. Wrote down what I learned: guetteluis.com/writing/shipp…
1
144
Apr 23
Huge thanks to @mattpocockuk for the inspiration. The Claude Code for Real Engineers workshop was 👌
1
31
Luis Güette retweeted
Apr 18
Morning bathrobe rant: Rule files.
110
143
2,609
708,443
Apr 10
You're not bad at prompting. You're missing vocabulary.
"Make the app show tasks from the database" vs. "Add a route to a TaskController that uses Eloquent to query incomplete tasks and returns a Blade view."
Same request. The second one gives Claude constraints instead of guesswork. The output improves immediately, not because the AI got smarter, but because you told it where to look.
You don't need to learn to code. You need about 16 words.
1
2
45
Apr 10
I wrote a walkthrough of the 16 Laravel terms that close the gap between "make it work" and "build it right":
buff.ly/cfZdJhx
#Laravel #VibeCoding
1
31
Apr 8
My favorite way to explain "background jobs" in Laravel is a restaurant kitchen.
A waiter takes your order and walks it to the kitchen. The waiter doesn't stand at the stove and cook your meal. They drop the ticket and go take the next table's order. The kitchen works through tickets in the background.
That's exactly what dispatch(new ProcessUploadedImage($image)) does. Your app drops the ticket and responds to the user immediately. A background worker handles the heavy lifting.
One line of code. The user never waits.
What's a concept that only clicked for you after someone explained it with a real-world analogy?
1
42
Apr 8
I wrote a full walkthrough of the 16 Laravel terms you need to read AI-generated plans, including the restaurant kitchen analogy for queues:
buff.ly/cfZdJhx
16
Apr 6
Claude writes you a plan.
"Create a migration for the tasks table. Add a route to a TaskController. Use Eloquent to query incomplete tasks."
You don't know if that's a good plan or a terrible one. So you say "looks good" and hope for the best.
1
1
64
Apr 6
You don't need 200 terms. You need about 16.
Route, controller, model, migration, Eloquent, middleware, queue, scheduler, Artisan, .env, MVC, Blade, validation, relationships, Composer, starter kits.
That's enough to read an AI-generated Laravel plan and know whether it makes sense.
1
37
Apr 6
Which of these tripped you up the first time you saw it in AI output?
15
Luis Güette retweeted
Apr 4
Currently having fun on this Onboarding screen for Phant! Just plug and play!
ALT Phant Onboarding Screen
1
1
56
Luis Güette retweeted
Estoy construyendo esto:
🌍 Una plataforma GRATIS para conectar emprendedores
Objetivo: 100 usuarios
Actualmente: 4
Si estás construyendo algo, entra.
👉 world.bypupila.com/
Vamos.
2
4
67
Apr 3
Researchers compared expert and novice problem-solvers across chess, medicine, and physics. They found the same pattern in every field.
Novices see a problem and jump straight to solving it. They start executing immediately.
Experts do something different. They pause. They spend significantly more time just understanding the problem before they touch anything. What do I actually know? What am I trying to find? What's missing?
The difference between experts and novices isn't intelligence. It's that experts have learned to resist the urge to start.
This applies to everything from debugging code to writing a business plan to asking AI for help. The quality of your result is almost always determined by how much time you spend before you act.
Next time you catch yourself jumping straight to execution, try pausing for 60 seconds and just asking: what am I actually trying to solve here?
How do you stop yourself from jumping in too fast?
20
Luis Güette retweeted
Mar 31
Bun is awesome. You can also do this in NPM with
$ npm config set ignore-scripts=true
Mar 31
`bun install` blocks postinstall scripts by default.
This helps protect from supply-chain attacks, like today’s axios incident.
![trustedDependencies
Instead of executing arbitrary scripts, Bun uses a “default-secure” approach. You can add certain packages to an allow list, and Bun will execute lifecycle scripts for those packages. To tell Bun to allow lifecycle scripts for a particular package, add the package name to trustedDependencies array in your package.json.
package.json
{
"name": "my-app",
"version": "1.0.0",
"trustedDependencies": ["node-sass"]
}
Once added to trustedDependencies, install/re-install the package. Bun will read this field and run lifecycle scripts for my-trusted-package.](https://venexa.site/media/HEuXqF3acAAM9PH.jpg)
ALT trustedDependencies Instead of executing arbitrary scripts, Bun uses a “default-secure” approach. You can add certain packages to an allow list, and Bun will execute lifecycle scripts for those packages. To tell Bun to allow lifecycle scripts for a particular package, add the package name to trustedDependencies array in your package.json. package.json { "name": "my-app", "version": "1.0.0", "trustedDependencies": ["node-sass"] } Once added to trustedDependencies, install/re-install the package. Bun will read this field and run lifecycle scripts for my-trusted-package.
3
3
347
Mar 30
My girlfriend tried Claude to tailor her CV to a job post. She uploaded the PDF, pasted the listing, and asked for a new version.
The result was generic. Flat. Could have belonged to anyone.
Her reaction: "AI isn't that useful."
1
71
Mar 30
The difference wasn't a better prompt. It was giving the AI the same context she had in her head.
What process are you repeating from scratch that could be a system like this?
1
29
Mar 30
I wrote about why the skill AI actually demands isn't prompting, it's learning to see systems:
vibemastery.ai/blog/stop-and…
#AI #SystemsThinking
1
27