1. Engineer started taking on more difficult projects and dancing weekly
An engineer I met at a prediction market conference set a bounty on increasing his social and career agency.
One session in July 2024; he kept up the practice on his own.
Within a month, he was:
Dancing weekly
Choosing more difficult projects at work
Embracing social interactions that used to be triggering
Expressing romantic feelings frankly
Previously he’d done:
I’m choosing to work on problems beyond my capabilities, and get excited about situations where my weaknesses are repeatedly on display.
He continued the work.
6 months:
Every week I’ve had some new ridiculous, fun, stupid project that I had others rallied around. It's exactly the stuff I kind of pictured doing but never really did…
One of my new friends said, “You're so adventurous.” That shocked me
18 months: He says he’s very happy and resilient, has lots of friends (and dates) now :)
2. Antoine is no longer incapacitated by corn
Antoine was the first to try. Multiple therapists and modalities had fallen short; “I’ve had anxiety my whole life.”
One session and he kept up the practice on his own. We also had a refresher call in month 9.
11 months:
I used to get triggered by very small things that some of the people surrounding me would say or do. Something as simple… “You told me that you didn’t like corn. Why have you bought corn?” These things would incapacitate me for 48 hours, I would have to shut myself in my room. Now I just laugh. —Video
18 months: He reports all is well; he’s continued unlearning through what he’s learned.
3. Sid asked out women without delay
Seven years of meditation and hundreds of hours of IFS therapy hadn’t gotten this AI researcher the results he wanted. He still didn’t ask out women he was attracted to.
After he started asking women out, we continued unlearning for further goals.
5 months after first unlearning (1 month after conclusion), we recorded this reflection.
At 11 months I asked for updates and he replied:
Update: They’ve been dating for over a year now. Things are good :)
4. Happy guy who had a 20y porn addiction for some reason ?
A man man set a bounty on his 20-year porn addiction. He had tried to quit for most of that time.
One session, a little take-home work.
1 month: “very few urges”
7 months: “I don’t have the urge at all anymore”
13 months:
18 months: he says things are good ya.
Update (2026 September): I’ve been puzzling on why this case was so easy. I think it’s because he had no external bottlenecks, happy life, good job, happy relationship… and just happened to have a porn addiction. His addiction wasn’t load-bearing anesthesia (as addictions usually are), so the unlearning was quick. This was the only time I can recall accepting a bounty on a predetermined outcome, let alone an addiction. Historically I’ve mostly screened out people with addictions, too, though this is no longer necessary.
Noticed that unlearning is faster when their emotions are easily accessible
I noticed that internal bottlenecks get unlearned faster when they have low-latency access to sensations in their body, so in my first year of research I tried to exclusively choose these people. I had been already selecting for people who had tried years of meditation, coaching, therapy, etc. already (so I can know I’m creating counterfactual change), which happened to help with this.
The end of single-session attempts
In my first year of research, I focused on “How much is possible in as few sessions as possible?” In my second year, I focused on “How much is possible in ~8 weeks?” This allowed me to help people who have trouble feeling emotional sensations in their body, which takes some learning (gasp) not just unlearning. This can support even greater change than the results above, especially for my target audience (AI people), e.g.—
—so I no longer attempt to handle everything in one session. It was cool though.
About | Twitter | Finding & removing bottlenecks for people helping ASI go well
I’ve had some “one-off” successes like this too (sometimes in as little as 30 minutes, though this was in an environment where we had a lot of shared context). A lot of your intuitions about this match mine, and I hadn’t actually made the “anxious vs avoidant” heuristic explicit to myself until I read your account of it, so thank you! Also love the “bounty” idea, and will think about how I might be able to implement it in my own practice.
I'm super curious, I've reread a bunch of your pieces. whaty do tou do in these sessiosn?