Tim A.
banner
flipperpa.bsky.social
Tim A.
@flipperpa.bsky.social
Open-source maintainer, objective "A.I." skeptic, speaker, and organizer in tech (mainly Python). 🖖 Guitar and serenity geek 🖖 Principal Engineer, WRDS @ Wharton @ Penn 🖖 he/him 🖖 https://github.com/flipperpa 🖖 https://youtube.com/flipperpa
The problem is, what if the human is distracted or gets tired of watching the assembly line? What if their skills degrade over time? The brain is a muscle after all, and we need our reps at the gym.

I don't have answers. There are handy tools here. I also see an incredibly strong addiction vector.
February 14, 2026 at 3:11 AM
My friend Randy's "Are You Sure?" paper underlines the issue. www.randalolson.com/2026/02/07/t...
The "Are You Sure?" Problem: Why Your AI Keeps Changing Its Mind
Ask your AI 'are you sure?' and watch it flip. Models fold 60% of the time because we trained them to please, not push back. The fix isn't better prompts.
www.randalolson.com
February 13, 2026 at 4:24 PM
I've been hearing the "feature" argument for years. In some contexts, sure, that's true.

For the vast majority of use cases for which these LLMs are being sold, they're a bug.

And I agree, they are fundamental and will never be eliminated. That's a problem in most contexts.
February 13, 2026 at 4:23 PM
Yes, a pretty hilarious mistake. $70 million spent in a domain, if they’d spent $1000 on a developer review…
February 9, 2026 at 7:12 PM
I only ponder, “why the fuck does anyone think PDF is an acceptable format for anything?”
February 9, 2026 at 3:19 AM