#avr Logs

Jul 02 2022

09:27 AM rue_bed: I do not suggest using any simulators
09:36 AM rue_bed: the real world is always different
09:36 AM rue_bed: if you need something that only ever works in a simulator, then its fine
09:36 AM rue_bed: if you want it to work in the real world, build it in the real world
09:36 AM rue_bed: its also REALLY important to learn how to not try to write an entire project in one go and then hope it all just works
09:36 AM rue_bed: projects need to be staged with modules and tests that limit the scope of error
09:36 AM josuah: it might be different for FPGAs and ASICs then? simulated, proven wrong, benchtests, and finally loading on a chip
09:58 AM josuah: every time a hardware engineer made a choice: "hmm, it is easier to just move that value directly, would be faster too, but doing so would make the project impossible to maintain", the A.I. would come and offer better performance by burning away 6 years of technical debt
10:02 AM josuah: soo... 5% of "bright geniuses" that burn all the advance in technical debt accumulated by 90% of the "code monkey" who will all have to spend their next 25 years cleaning-up all that mess the smart guys gave to them all over again?
12:10 PM Cracki: >the AI Would come and offer better performance by burning away 6 newborns before they become bad engineers
12:11 PM Cracki: I saw that "AI" programming an fpga, it's decades old. it was a genetic algorithm or sth. it used the specific chip's peculiarities to construct oscillators.
12:12 PM Cracki: the "program wouldn't run" on another chip of the same model.
12:16 PM Cracki: they actually could analyze it... it just did things that no regular engineer would come up with for an FPGA
12:16 PM Cracki: those oscillators were unconnected logic that was affected still by the other stuff, and affected it in turn.
12:17 PM Cracki: just removing "unreachable" circuitry made it fail because that actually _was_ reachable by HF stuff
12:17 PM Cracki: if you've ever seen some RF antenna designs where apparently unconnected copper polygons sit by each other... that's exactly that
12:18 PM Cracki: don't kid yourself. code monkeys cause a lot of tech debt. I'm seeing it at work.
12:19 PM Cracki: they aren't smart enough to make the code clean.