A member was building the control logic for his first product, and he did what almost everyone does now. He asked AI to design the circuit.
It gave him a confident answer: a specific latch circuit, laid out in detail. A latch is just a small circuit that holds a signal on or off until you tell it to change.
He built it, and it didn't work.
So he spent days tweaking values, trying to force it to behave, and none of it made a difference.
Eventually he went back and asked the AI to rework the design.
This time the answer came back even more polished: a full writeup with pin tables and little checkmarks next to every step.
Underneath, it was still the same wrong circuit. The answer had gotten prettier without getting any more correct.
So he brought it into my Hardware Academy, and one of the engineers did the two things a chatbot never did.
First, he rebuilt the AI's design in a simulator, and it failed on the screen the same way it had been failing on the bench. That settled it with a test instead of an opinion.
Second, he worked out what the member's product actually needed, and it called for a different type of latch entirely. In his words, "what you need is a D latch, not an SR latch."
Then he drew up a working schematic around parts the member already had and ran it through the simulator to prove it. The design file was attached right there in his reply, ready to build.
All told, the member had put days of building and tweaking into an answer that was never going to work.
So there are two lessons here.
First, a beautifully formatted wrong answer is still wrong, and it will be wrong the same way twice. AI doesn't notice its own mistakes, no matter how many times you ask.
Second, the difference between an answer and engineering is verification. Get ideas from wherever you like, but before you build one, put it in front of someone who tests it and stands behind it.
That's exactly what the engineers inside the Hardware Academy do, before you spend your time and money building it.
Talk soon,
John Teel
Predictable Designs
|
|
|
|
|
|
|
