Video summary
Recovering from AI Psychosis | TheStandup
Main summary
Key takeaways
Overview
Adam.dev (“Adam”) returns to the stage claiming he’s recovered from “AI psychosis”—a month-long (roughly) burnout spiral where he got trapped in an endless loop of:
- constantly prompting powerful coding models,
- running multiple work branches/agents in parallel,
- obsessively chasing the “perfect” output instead of sleeping and living normally.
Main Plot / Arc
Origin story (the slide into it)
After Claude Code-style agent workflows took off, Adam gradually leaned in more over the year:
- first doing lots of work “by hand,”
- then getting hooked when newer models (he cites the GPT 5.2 era) felt smarter and faster—keeping him engaged.
The psychosis pattern
Adam describes a familiar addiction loop:
- Prompt
- Dopamine hit from results
- Review/refactor endlessly
- Spawn more work and keep going
He also noticed the pattern in others online, saying he could “tell where they were” from behavior like:
- multiple accounts/subscriptions,
- agents running overnight.
Peak damage
The clearest symptom was sleep. He went from being energized by programming to being unable to stop:
- waking up with his laptop near him to start more prompts,
- eventually crashing into severe burnout.
What stopped him
Two forces, basically:
- His wife noticed the sleep disruption and pushed boundaries.
- Adam hit a wall hard enough that he effectively took about a month off the internet (around five weeks).
He credits the forced break with giving him perspective.
Offline “recovery montage”
Adam describes rebuilding his life outside the model loop:
- Netflix (including the first time he watched Breaking Bad)
- getting back into golf
- spending time with his wife
- playing tennis while his kids were in school
Where he is now
He says he feels amazing and is back working on Open Code, but with a healthier approach:
- slower, outcome-based goals
- less parallel “churn”
- more review
- using AI to drive toward real product outcomes instead of generating huge diffs nonstop
Highlights / Jokes / Notable Moments
-
Running gag: “Becoming coffee / order correctly” A host interrupts with a sponsor-style bit contrasting “HTTP” getting blocked with “SSH terminal.shop coffee,” leaning into the show’s tech-obsessed humor.
-
AI model debate → “last success wins” The panel discusses how people pick “favorite” models based on the most recent big win rather than rigorous comparison. They also mention differences in prompt style and “flavor,” e.g.:
- Claude adding tests
- GPT being useful but sometimes very “UI boilerplate”-heavy
-
Gary Tan / token burn + chargers joke Adam jokes that his “AI psychosis” was literal—his USB-C cords allegedly burned out, and one adapter failed / imposed charging limits. The group riffs that maybe he only reached the “first circle” of psychosis.
-
“AI psychosis detector” idea They joke about adding an opt-in Open Code warning like the old Wii “go outside” pop-up:
- tracking if you have too many agents running too long,
- even “turning it on for your friends” via social graph/handles.
-
Addiction comparisons Adam compares the prompt/reward loop to slot-machine addiction. The hosts connect it to other compulsions like:
- trading,
- WoW-style compulsive play,
- “AI vampires” destroying sleep by waking to check agents.
-
Stage persona humor about language/word choice Humor about using “big words” vs. plain language (e.g., “Ozarks” / “we don’t use big words”) repeatedly contrasts with Silicon Valley tech intensity.
Key Takeaway / Message
Adam’s recovery isn’t “AI is bad,” but that the process can become self-destructive when it turns into:
- chasing dopamine,
- churning parallel work,
- ignoring sleep and craft.
He’s aiming for slower throughput, fewer pointless refactors, and measurable outcomes, while staying vigilant so the hype cycle doesn’t pull him back in.
Personalities Appearing
- Adam.dev (“Adam”) (guest; Open Code / recovering from “AI psychosis”)
- TJ (host / interviewer)
- Casey (host / commentator)
- Prime (host / commentator; frequent tech/golf analogies)
- Other recurring voices referenced in jokes (e.g., “Trash,” “Gary Tan” mentioned; not necessarily physically present)
Rate this summary
Your feedback will help improve summaries.
Improve this summary
Reprocess with a stronger model when the summary feels incomplete or inaccurate.
Translate summary in another language
Ask questions to this video
Chat for follow-up questions, clarifications, and source-backed answers.