TURN YOUR FEARS INTO A BUCKET LIST

on leaving a job

“I love struggling, actually. It makes me feel alive.” - Alysa Liu

---

one of my personal quests over the last few months has been to retrain my brain to interpret discomfort as fun. this was motivated in part by Alysa Liu’s winter olympics interviews, and in part by my own experience with driving: in high school i was terrified of driving on highways, whereas nowadays it is a deeply relaxing activity that i look forward to. i wondered what it would feel like to take an experience i was very uncomfortable with and make peace with it, and then repeat that process over and over again

i started by making a list of fears: horror movies, heights, physical contact with strangers, the dentist, cooking for other people, bugs, improv. then i decided to expose myself to all of them repeatedly (so far i have gone through the first five; the last two will happen soon). some of the exposures went pretty much how you’d expect: i cooked for people, it turned out fine, and i was a bit less scared of cooking for people afterwards but didn’t learn too much. some of the other exposures were quite interesting; i’ll focus on my experience with horror, which i found somewhat beautiful:

before this quest started i’d only ever watched a single horror movie (The Conjuring), which i’d spent the majority of hiding under a blanket so i’m not even sure if that should count. then my housemates decided to watch Backrooms, and i was really not looking forward to it, which i decided was exactly why i should go. i spent the first half of the movie under a state of extreme anxiety, convulsing at every jumpscare and dreading whatever was about to come. my chest muscles became sore from all the tension i was holding in, and i desperately wanted to fast-forward through every scene not located in the real world

and then, halfway through the film, an odd sense of peace descended upon me. i realized that whatever happened next in the movie was out of my control and the same plot would play out whether i liked it or not, so why not try to enjoy it? once i let go the theater no longer seemed like such an awful place to be. i began to appreciate the absurd elements of the film, which i’d previously only found unsettling. i continued to jerk uncontrollably every time a sudden motion occurred onscreen, but i was no longer tense or anxious. i spent the next few days bugging my housemates to watch Obsession, and it turned out to be a similarly pleasant experience - terrifying but enjoyable, and i did not have to look away a single time

---

an unintended side effect of this exercise is that i began to notice just how much of my life was constrained by fear in ways that were previously invisible. it was the kind of noticing i would’ve historically arrived at via introspection or dialogue, but direct experience proved to be a much faster path towards new perspectives. there were two related types of constraints: sometimes i would instinctively discard certain options when making decisions because they seemed uncomfortable, and sometimes i could not properly understand my own preferences because they were too deeply concealed by fear

the biggest realization was about what i wanted to spend my time on. if you asked me at the start of 2026 i would’ve said that i was fine with doing interp research (which i found intellectually interesting but thought was most likely a dead end) as my day job while spending my free time on things i cared more about, like writing and trying to improve human-AI interaction. once i could more clearly see the invisible constraints in my life i understood this compromise was borne out of a fear that i was not competent enough to successfully execute on the directions i cared more about. and then i realized that 1) this fear has a decent chance of being incorrect 2) if it does turn out to be correct, i would like to know as soon as possible so that i waste less time chasing the wrong dreams

some relevant images from my pinterest

afterwards i discovered i could no longer agree to the compromises i’d previously been happy with. i wanted to work on what i cared about full-time, and was unwilling to settle for less; it was very easy to forgo money, prestige, and legibility for six months if necessary. existing perceptions of me as indecisive or incapable of making hard decisions were simply not true. i was surprised by the strength and clarity of my own resolve, once all the fear had been stripped away

you can’t become who you want to be by doing what you’ve always done. nowadays i recommend that all my friends turn their fears into bucket lists, and that they watch horror movies, especially if they’re the kind of person who doesn’t want to. at the very least you’ll rediscover how it feels to be alive

---

if you liked this you might enjoy another post i wrote two years ago when i left my previous job, which is spiritually similar but more focused on scarcity vs abundance (and looking back it has aged surprisingly well)

anyway, this is the end of the main body of the post but here’s a FAQ for people i haven’t talked to recently

Q: so what was your job again?

A: i was at Transluce, a research lab focused on measuring and understanding language model behaviors. for instance: when and why are models deceptive to users? how often do models encourage suicide? and so on. the org is very dear to me — in my two years there we’ve done really great work, from our initial launch where we labelled every neuron in a LLM to our upcoming studies quantifying how AI affects mental health. the team is wonderful and my time there made me a much better researcher and collaborator; if you’re interested in understanding language models i strongly encourage you to consider working there!

Q: how can you simultaneously be leaving and encouraging people to work there?

A: i think work around measurement / evals / societal impacts is important and genuinely think Transluce is a great place to do that work; i just also think it may not be the best fit for my skillset and personality. i believe we are far from understanding how to build AI which is good for humans (not “good for humans” as in “productive assistant”, but actually good long-term: supporting agency and growth and self-knowledge, not inducing brainrot or social dependence, etc) and measurements of the most important AI-related harms may not be as helpful as people expect: they tell you a lot about what to avoid but say very little about what to build towards. to figure out what to build i think people should be putting out as many independent and concrete perspectives about positive human-AI futures as possible (prototypes and demos, weird model finetunes, position papers, etc) and i’m excited to see if i have anything to contribute

Q: what are you actually going to do next?

A: the current plan is to spend six months on independent research and then re-evaluate. i will be trying some ideas around: nonstandard narratives and sci-fi, different kinds of collaborative relationships, changes in long-term user preferences. also i want to try working in operations; i think it will unlock new perspectives. reach out if you’d like to chat! the top of my website is an accurate summary of my interests:

Q: why not join a lab?

A: maybe that will happen at some point? but i don’t really agree with what existing labs are working on and don’t have much clout so if i joined a lab i’d basically have to work on whatever direction my manager had already decided; a better path is probably to produce some interesting work first and then convince a lab to continue work in that direction. also i’ve never given my own ideas a fair chance before, so i want to try that out and see how it goes. resource constraints are probably not be an issue as i am focusing on ideas that i think can work without large amounts of compute

Q: why not go to school?

A: i may apply again but academic cycles are frustratingly slow, eg. the earliest i’d be able to go to school is fall 2027. visiting researcher roles are possible though!