I just cancelled my $200 Perplexity Max subscription. Perplexity is buggy, unreliable, and has not meaningfully improved in months. Computer use crashes constantly. Search results feel worse than they did 6 months ago. Codex computer use is genuinely better at taking control of your machine and getting things done. That was not true 3 months ago. It is now. The downfall of Perplexity needs to be studied.

There's a lot of noise around how "Ai generates buggy software". The truth is writing the code was always 20% of the time. Testing and refining was always 80%. But now as that 20% compresses 10x, we want to compress the 80% as well... That's the mistake.

Playing with MSP430 to run some specific TI firmware for tests. MSP Flasher is quite buggy and successfully bricked the programmer board while trying to update the firmware. Fortunately, I was able to restore it.

good prompt for the lorge models > can you find places where we are fighting the frameworks we are using, reinventing things unnecessarily (while still keeping our strict supply chain posture), relying on deprecated behavior, buggy, or needlessly complex? any tests that don't pull their weight which if removed could significantly simplify the code?

Nope, the latest Qwen3.8-Max-0902 slid back in my Cosmic Dodge game test. It took 3 attempts to get the prompt correct and a final one to smooth out the buggy gameplay. The original Qwen3.8-Max one-shotted this game perfectly. Not a step up in this sample of 1 test.