Learning AI From the People Building It

I have written PHP for over twenty years. Laravel is my home. For a big stretch of that I worked as a Chief Technology Officer (CTO) in the credit monitoring space, running a platform that answered millions of queries a month.

Now I spend part of every week learning from AI builders. Not the hype accounts. The people shipping models, chips, and products. I put the ten I learn from most into a list on LinkedIn this morning.

Here is the funny part. The crossing-over is smaller than I expected.

My rules did not change

When I led a team of twelve-plus developers, we rolled AI into legacy work slowly. The rules were simple:

1. AI writes the missing tests first. Legacy code with no tests gets characterization tests before anyone touches it.
2. AI never touches auth, credit, or secure code. Not once. Not with a good prompt.
3. Ship small. Measure. Then decide if the tool earned a bigger job.

Those rules still run my work today. I am building Bin X, a bin-rental SaaS, and the same lines hold: tests first, hands off the sensitive parts, verify everything. My whole career is trust but verify. AI did not change that. It just made the verify part faster.

What the builders taught me

The AI builders worth following share one habit. They show their work in public and they ship before they feel ready. I was the opposite for years. I waited until things were polished. Nobody learns from polished. They learn from shipped.

So I ship smaller now. Bin X goes out in pieces. My posts go out rough. The feedback loop beats the planning loop almost every time.

Start here

If you are a veteran engineer crossing into AI, simply steal the workflow, not the fear:

• Pick one boring task.
• Have AI draft it.
• Write the test that proves it works.
• Keep the keys to the secure code yourself.

That is it. Start with the tests. The rest follows.