Is there something you’ve been trying to do that keeps getting stuck?
I research internal bottlenecks for ASI people and offer both evals and unlearning:
Daniel Tan, coauthor of Emergent Misalignment, was trying to accelerate alignment research and enjoy life but kept losing weeks/months to burnout. Willpower would help, but only briefly. Since our unlearning in summer 2025, he reports long periods of burnout no longer happen. He also spends far less time distracting himself every day, instead pursuing personal projects and hosting events for friends. Interview + video.
An AI researcher wanted to speak his mind but kept holding back. Seven years of meditation, 150 hours of IFS therapy, and multiple retreats had not resolved this. After the first afternoon of unlearning he began asking women out. After a few more he started disagreeing with colleagues, being direct with his mom, and texting friends “I miss you.” Video interview. At our last follow-up (18mo), he confirmed these changes have lasted. He also entered a happy long-term relationship, which freed his attention to join a lab. He stood up for himself when someone took credit for his work.
Everyone’s bottlenecks are different. To assess if internal bottlenecks are limiting what you’re trying to do, I offer evals. In a week you’ll learn how to feel internal incentives so you can assess your bottlenecks firsthand, and decide whether unlearning would help unblock you:
Contact: Twitter DM or chris@chrislakin.com. I’m in SF.
Long-term data
Update: They’ve been together ~1y
In the first year, my research focused on what can be done with ~1–4 afternoons. It was very experimental; payment was forbidden until their genuine satisfaction with results. Of 8 AI people who tried, 6 paid bounties and 2 turned out not to be a fit. A few examples:
In the second year, I started to filter for readiness and commitment. The incentive structure changed to half upfront, half on results. After 2 mis-starts (one of whom I refunded, the other says they liked it and might still pay(?)), I learned how to select better and the engagement converged on 8 weeks and required take-home exercises for the first time. Of 6 AI people selected after this, 5 paid bounties. 1 had to pause due to unrelated personal crisis.
Now in the third year, I’ve created a one-week expensable eval to lower the barrier for people who don’t know their bottleneck while still maintaining selection. The incentive alignment for unlearning has also evolved to align with impact: Refundable commitment bond + Payment proportional to post-hoc outcomes, forbidden until your genuine satisfaction.
I’ve learned that attempting to remove internal bottlenecks while external bottlenecks are your primary bottlenecks isn’t a fit. This includes severe sleep or health issues, financial or material security, and active personal crisis. If the eval reveals significant external bottlenecks, I require that you address them before attempting unlearning.
Finally, due to growing concerns about AI timelines, I no longer have capacity to work with people outside of AI.
I got exactly the results I intended to.
—MTS @ OpenAI, reference available on request
Further references available upon request.
Not medical treatment, not therapy. Substances aren’t used. Stories were shared with permission. Client matters are confidential by default.