Tried @zeddotdev yesterday
Simple request: add styling for the warning label.
Result: correct change + full CSS reformat
Zed AI:
✔️ Adds 3 lines of CSS
✔️ Reformats the entire file I did not ask for.
✔️ Burns your tokens like it’s Black Friday
AI does something slightly different and more expensive.
When the problem is unclear, AI doesn’t just execute faster. It explores faster. Every hypothesis is real work. Every wrong path is billable.
No doubt AI coding agents improve productivity but one flaw should give teams pause.
I recently asked Claude to debug an issue for me. The AI agent kept working… and 30 minutes later it said..
“The tests keep failing, it appears I followed the wrong hypothesis.”
Something important clicked for me.
That moment reminded me of an old BPM lesson:
Automating the wrong process just gets you the wrong result faster.
Guy in Australia uses AI to create a mRNA vaccine to save his dog. Meanwhile in the USA, Robert F. Kennedy Jr. cancelled $500 million in government-funded research projects aimed at developing new mRNA vaccines,
Australian tech entrepreneur Paul Conyngham explains how he used ChatGPT/AlphaFold (spent $3,000 with no biology background) to create a custom MRNA vaccine to treat his dog’s cancer tumors. Unreal.
Yet the U.S. government hid him, blocked France’s extradition requests, and helped him escape—all in the name of Cold War utility. The gap between his crimes and the excuses used to protect him is staggering. DOJ report (1983): https://t.co/Xkgvis4kK7
This 1983 DOJ report “Klaus Barbie and the United States Government” shows how power twists the law when it’s convenient.
Klaus Barbie was a Nazi Gestapo chief who tortured resistance fighters, deported Jewish children, and sent thousands to their deaths.
Some say AGI is close. Meanwhile, @OpenAI#ChatGPT 5.1 just generated music no human can play.
Perfect example of why AI seems brilliant but actually has no idea what it’s doing.