I appreciate that feedback however, if you're reading the series, you'll understand why and how I'm both defining and undefining the term all at the same time. There is a method to my madness.
Talented AI PM teams can be still produce chaos. When every PM works in an individual session silo, inconsistency becomes the system.
Context is not “more info.” It is shared governance.
medium.com/deanondelive...
What's even more amazing is the legal burdens they try to extend to their vendors with regards to using AI.
AI product managers live & die by their prompts. “Write me a user story” is not a prompt. It’s a prayer. Prayers are a fragile ops. model.
Day 04: What prompting well actually looks like. Not prayer. Not a vending machine. A craft many teams skip entirely.
medium.com/deanondelive...
Your AI product management team is shipping features. Your customers still hate the product. That’s not an AI problem. That’s a problem-definition problem.
Day 03: The most expensive mistake in AI PM is pointing the right tools at the wrong problem.
medium.com/deanondelive...
Just how many AI PMs are traditional PMs with AI-induced imposter syndrome?
AI product managers are the hottest title in tech. Many are legacy PMs with AI-induced imposter syndrome. The skills gap is real. It’s widening daily.
More here on the topic:
medium.com/deanondelive...
Happy Monday people ... hopefully the Claude outages won't be too stressful.
status.claude.com
How much you want to bet some 'lollygaggers' live here?
AI product management is product admin in disguise. Many updated job titles in 4 secs. Few updated their practice past prompts.
Day 1 of a 15-day series. The $13B Clippy moment. The n8n-is-a-strategy myth. The real definition. If it stings, good.
medium.com/deanondelive...
Ever been in that meeting? Where peeps argue over oars & you don't even know if you're taking the river or the road yet?
Nobody defined the problem.
Nobody set the strategy.
But Karen has STRONG feelings about composite blades.
Don't put the cart before the horse.
Pick the river or road first.
Research is in. Had Dave cursed at HAL, he'd a gotten better results than being relatively polite.
Ever since ChatGPT showed up, AI influencers have preached “be kind” pablum.
A recent PSU study says otherwise. Honestly? Me and my typed-out tirades are relieved.
www.linkedin.com/pulse/impoli...
Real product work shows up before the agent runs. Intent. Constraints. Failure modes. Contracts that assume things will go wrong — because they will.
If your agents need babysitting, you didn’t ship intelligence.
You shipped chaos with better branding.
Everyone is duct-taping agents together like caffeinated raccoons & calling it “strategy.” Congrats. You’ve built fragility with a GitHub star.
The PM failure isn’t tooling. It’s abdication. No goals. No decision logic. No guardrails. Just vibes, workflows, and a prayer that nothing breaks at 2am.
Welcome to agent-mageddon. AI is replacing PMs. IRL: an agent-only social network. 1 bot braggs it fired the PM. Metrics up. Everyone cheered. Nobody asked why.
substack.com/@deanpeters/...
The failure mode isn’t having these biases.
It’s letting one of them dominate the room.
In the age of AI, recognizing bias matters more than ever.
Our models and agents don’t remove bias. They scale it.
Judgment isn’t about eliminating bias.
It’s about balancing the right ones at the right time.
Every one of these shows up in real, everyday product work:
- Loss aversion when the stakes feel high
- Confirmation bias when we want to be right
- Optimism when momentum matters
- Action bias when patience runs out
None of them are villains.
None of them are heroes.
We keep telling product managers to “be objective,” then act shocked when decisions wobble. Bias isn’t a bug. It’s a vestigial survival feature we never outgrew.
Look at the picture. Be honest. You don’t just see these biases. You recognize them. You counter them. And on bad days, sneak a more in.
Keyboard of the Future.
Molt that MFrs 😜
I was being an optimist.
We treat opera singers & athletes like pros: years of training for moments of performance. Then we hire product managers, give them 1 day of training, and expect flawless “performance” forever. If you want better outcomes, stop measuring the show & start funding the rehearsal.
Only thin layer of snow & ice here near Raleigh. Nothing like Snowmageddon 2014. Still, it cold & slick enough to have me push out this important infographic, "The Unofficial Southern Snowpocalypse Guide : Illustrated advice for people who should know better!"
Bottom line:
AI doesn’t make you dumb.
Using it instead of thinking does.
TLDR:
Efficiency without cognition is just faster decay at AI speed.
🧵 7/7
So what do we do?
That’s what the infographic shows.
Left side:
The recipe for cognitive rot.
Right side:
The antidote for intellectual atrophy.
Think first.
Struggle briefly.
Then use AI to refine.
🧵 6/7
Cognitive offloading research (2024–2025)
Scaffolding supports learning.
Replacement offloading causes atrophy.
When AI does the messy middle,
your brain skips the workout entirely.
www.researchgate.net/publication/...
🧵 5/7
Educational psychology studies (2024–2025)
AI improves short-term task completion.
It degrades metacognition and self-regulation.
The work gets done.
The skill does not get built.
Artifact delivered.
Skill unpaid.
🧵 4/7
Microsoft Research + Carnegie Mellon University (2025)
Knowledge workers who trusted AI more engaged in less critical thinking.
Confidence went up.
Verification went down.
Fluency replaced judgment.
www.microsoft.com/en-us/resear...
🧵 3/7
MIT Media Lab (2025)
EEG studies show lower brain connectivity when people write with LLMs instead of thinking first.
Most participants couldn’t recall what they had just “written.”
Faster output.
Shallower encoding.
www.media.mit.edu/publications...
🧵 2/7
Stop Cognitive Rot. End Intellectual Atrophy. This isn't panic. It's peer-reviewed.
3 years after the release of ChatGPT, we are now quietly trading thinking for fluency.
The research is no longer subtle. Let's count the ways.
🧵 1/7
I spent ~2k hours training & engaging ~1k product managers in 2025. Here are the Top 10 AI PM Mistakes of I observed. YMMV.
2025 demonstrated that AI doesn't fix dysfunction, but rather it amplifies it.
Read the why's behind my list in my Substack article:
deanpeters.substack.com/p/top-10-ai-...
I'm not a big fan of Halloween, but in this case I'll make an exception.
Behold, my Larry Lounge Lizard kit comes back to life as an AI product manager costume.
Meanwhile, I'll be vibing together a webinar on building shit with AI for PMs is coming November 13th.
Be there, or be square.
