Martin Vrbovčan

I'm a student at FER, Croatia. Interested in maths, biology, and AI. It's fascinating how optimising over combinations of shape production (proteins) led to intelligence which we are still figuring out.

Currently: reading my favourite writers Henrik Karlsson, Sasha Chapin, Celeste, Scott Alexander. Working on a quasi-startup. Trying to wrap intuitions around RNA folding.

Click on a day to see what I was up to.

Feb Mar Apr May Jun
Mon Wed Fri
2026-03-30
  • Rebuilt personal site from scratch. Libre Caslon is my favorite font.
  • Rabbit hole w Claude on stomach acid, RuBisCO, Purple Earth hypothesis.
  • Life works on a basis of ~50 molecules (nucleic & amino acids, ATP, sugars, essential lipids), that enzymes can work with & modify really well.
  • Attended an AI Business process automation presentation by Plavi Tim, INA
2026-03-31
  • Coffee talk: what pharmacy does, emotional books to read & my fav writers.
  • Stohastic processes; lots of questions about marginals and conditioning
  • Attended a long and sprawling 6 hour student association voting sesh.
2026-04-01
  • Bash scripting fun!
  • Stretching Claude code to do crosschecking from 50 PDFs and a spreadsheet, works!
  • Hosted house party, with 40min "Why I like San Francisco & rationalism" presentation.
    • Topics like liking unorthodox things, how people dream there, optimize, worldmodels, you can just do things etc.
2026-04-02
  • Early trials into a custom trading/hedge fund design, trading learning
  • Was generally quite shocked about how little energy or will I had to do things
    • Probably lack of sleep, bad start of day, etc.
    • Music is my main productivity poison
  • Found out & told collegue I didn't do something I promised 3 weeks ago; as a consequence a gaint international competition event has to be delayed by [2 weeks, up to 2 months].
    • Feel really bad about it :(; but I've had a nonconsistency failure mode I still haven't gotten rid of. Planning fallacy, being too optimistic about outcomes so not putting too much thought into negative externalities risks etc.
    • Something I want to fix.
    • Will be cleaning this up next day and the next week :DD
2026-04-03
  • Very much underslept because of student org meeting yesterday.
    • Was sluggish throught the whole day
    • Too much time on substack (although excellent writing)
  • Didn't do much except shuttling papers to an NPO office :/
  • Fixing mistake from yesterday (while tired...)
2026-04-04
  • Another generally lazy day: too much time watching films & Substack.
  • Sending emails to fix my previous mistake still.
  • Felt like everything after college was more surface level, like I was improving more until 19.
    • Realised I'm not actually playing this game right. I've fallen into the gifted trap. Grown complacent, not challenging myself, not strong enough of a work ethic. Drift under weak external structure.
    • Everyone around me feels like they're missing the ball, NPCs at college. Super easy exams. People fail them for whatever reason still.
    • But still, higher level people scare me and I know they exist. My secretary is one.
    • I'm wasting my potential, my high level connecting and reasoning.
    • I should actually play. Start cracking open textbooks, work ethic.
  • Read amazing substacks!
    • Career credidential escape velocity (hiring baseed on where you worked before locking you in), concluding that theoretically you should pay to work somewhere like Google. Might want to work unpaid?
    • Homeschooling pitfalls. If i ever seriously consider it (Henrik Karlsson does it, amazing essay on autobiographies of geniuses), I better watch out.
    • No tech regieme sound good? I would build a little box that only has claude on it, and really educational whitelisted videos. No other programs.
2026-04-05
  • Easter!
    • Which means relatives visit, etc. Chilling with family.
  • Chatted with amazing high school frien about life, socials... Lack of time. Agreed to keep in tuch much more! Am so happy because we reconnected more after so long.
    • Would be cool to go to San Francisco together before Switzerland in September.
  • Celeste is an amazing writer.
  • Rationalist music is really good ("You have not been a good user" album).
  • Planned the legalty of quasi-startup, namely payment processing while students w no legal entity
  • Simmilaily how some organisms have many more generations and thus evolutiona is quicker, people can live quicker than others.
    • Different clocks, due to ingrained mental habits, environemtns, productivtiy, making contact with reality, being less NPCs and doing more novel things.
2026-04-06
  • Setup Claude code to search for hackathons across Asia because a friend and I really wanna visit Singapore and Asia
    • found lots of cool things in America like Treehack, would be cool to give it a go
  • Quote I can't stop thinking about: discipline is reduction. Structure is cutting off choices.
    • No choice means you're literally "locked in"
    • Obviously, cutting off superstimuli, unnecessary people & timewasters while keeping randomness would be the way to go.
  • Did a buffer overflow exercise in ~6h, amazing! Cybersec is cool and creative, working under constraints.
    • First time doing shellcode, pointing back to myself, gdb, etc.
  • Friend saw competition for quasi-startup, now we're demotivated :(
  • Making money is kind of like crafting a return oriented programming exploit: chaining together multiple informational / etc. asymmetries to get to money in the end.
2026-04-07
  • Amazing group chat sesh where I described my ideal future.
    • Living with friends, collective schooling agency institution, financing people I love, being surrounded by smart empathic people (our little "comune" hahah).
    • Also agency talks like becoming an Anthropic ambassador, etc., bringing SF to Zagreb via posters and cafe meetings.
  • Pulled an allnighter to finish sending emails.
    • After 1h of sleep and feeling terrible the next 1h after waking, I felt amazing for the rest of the day. (?)
  • Wanted to host rationalists coming to Zagreb because of Scott Alexander, but cant due to flatmates
  • Cool math lecture today: inner products, norms, metric, Hilbert space
  • Feel high agency again, probably location change
    • Talk highlight: informational edges are everywhere that Claude code & etc. can find!
2026-04-08
  • Attended a kaggle "Clustering cup"; you get 512 dim data, cluster it, closest to ground truth wins! I didn't win, but it was a really fun!
    • Should have played more with normalisations, visualisation first. Like 2x of what I did.
    • Rapid iteration & exploration w AI.
  • Built an auto Ryanir/Wizzair cheap flights tracker. Get cheap round trips (<60€) over next 45 days in inbox. Accounts for exams and calendar. Autoruns every mon and fri.
    • Lots of other similar tracker ideas. You can just do things!
  • Coffee with startup incubator friend.
    • Keep doing shit, you can always post more, better features, pitch it to more people... That's why founders work so much.
    • No magic sauce. If you fail, you'll get experience. Maybe by the 10th you'll succeed.
  • Temu package arrived. I'm so happy (maths clothes and necklace, sweater).
    • Underrated fashion/uniqueness arbitrage opportunity because it isn't constrained the way local shops and markets are.
2026-04-09
  • Attended a Jane Street distributed systems talk, also taking a midterm exam in the middle of the timeslot.
    • Most important hiring traits (according to their recruiter): how you think (purposefully open ended q and watch you reason out loud), having a passions or "fixation" on one thing (given examples were math, MTG and Minecraft servers, someone who really likes that and so dug deep into a specific direction).
  • Student org meeting + going out
2026-04-10
  • ACX/LW meetup Croatia with Scott Alexander!!!
    • was amazing obviously, had front row seats to Q&A
    • massive job loss, period of social unrest, highly unpredictable tech singularity, long timelines problems (demography, climate change) aren't relevant because timelines are short
    • Met amazing people (tech VCs, startup people, education reformer).
    • even with a pause max time is 20 years (because chips & open source)
    • check out Kelsey Piper for education!
    • Talking about AGI & massive change with a group of people who take it seriously; the world is gonna change a lot. Need time to process everything.
    • I really like this: "I'm not informed about \[domain\], but I do defer to this \[nerd\] who spent a lot of time on it, and concluded Z. Sanity check with prediction markets." Because nerds are unbuyable and spend a lot more time in a domain than I ever can.
  • Thinking about visiting San Francisco in the summer, will get concrete tips.
  • Sleepover at friend's place!
2026-04-11
  • Stayed at friend's after yesterday's sleepover.
  • Board games & 6h talk sessions over coffee.
    • The Dictator (about roman senators) is an amazing social deduction game!
    • Had it confirmed by multiple people that I suck at lying, which made me last at every game we played. Claude says one can improve quickly though, so doing this skill would be cool.
  • As usual with staying at friend's, wasn't productive (because of the yapp).
2026-04-12
  • Fixed visual mobile/desktop bugs of my website.
  • Thought I would be productive, but talking and focusing on editing Scott Alexander Q&A killed me in the productivity sense (didn't even complete it).
  • Finishing Game of thrones w my sister (S8E4).
  • Watching Gary's economics, really interesting thesis about inequality, wonder why it's not priced in.
2026-04-13
  • Reconnected with old super agentic friend with 2.5h phone call :)
    • Interesting idea: getting a psychologist (she has one and recommends it a lot) as a "personal life coach". She says it speeds you up.
  • Finished reading "The Perks of Being a Wallflower", related thoughts (maybe from finishing at 1 AM).
    • It isn't fair that so much of society & dating are appearance-based, when it's mostly a genetic lottery (and social circumstance). It's fucked up, especially because of compounding.
    • "He's one weird kid" because he has no friends; same thing, but environmental and sociological lottery! The environment plays a massive feedback loop role in when you get socialised, whether you have people around you that you click with, social osmosis of traits & charisma & jokes etc.
    • Someone not winning the lottery doesn't deserve to be without love or outcast.
    • Bill moved me; something about great mentors delights me. A relationship I both want, and want to give.
  • Was feeling super confident the whole day, it rubs off on other people and their happiness as well. The solution is to just smile and be nice and happy (?!)
    • Confidence is really just internal, but also compounds.
2026-04-14
  • The course EPFL mapping from FER (my uni) is a patchwork of shadowy information you get from older students and what they officially give you ONLY AFTER YOU ASK QUESTIONS. If I didn't have my group, I would have missed out on the research project lab and extra subjects. Isntead of 22, I get all 32 ECTS recognised!
    • I'm actually most excited about `Online decision making`, as it concerns game theory, bandits, regret bounds, exploration-exploitation, and a bit of RL!
    • Real life relevant abstractions!!! (like the Gale-Shapley algorithm's relevance to agency & dating)
  • The TODO list stack is the natural container for tasks, whether you keep track of it or no.
  • Imagined plan: sleep from 2 AM till 6 AM, then do my EPFL course matching and plan ASAP. Do independent assesment before Claude Code's suggestion.
    • Actually: woke up early, felt great but didn't check too much outside of Claude Code because it was top notch output. Watched Youtube/read, ... Fragmented.
    • Because of the early waking, crashed out at around ~15:00, highly unproductive, fragmented, just went to sleep early cause I wasn't effective.
  • EPFL lacks video vloggers talking about their experience! My gap >:)
    • Vlogs today are probably high signal and force you to have good lives.
2026-04-15
  • Was quite unproductive. It's an environment issue! Staying at home, not doing anything, is self compounding. Superstimuli grabbing me.s
  • Amazing Claude talk
    • Earth bottlenecks beyond sun boiling: Atmosphere loss (very long), core cooling (~4 By).
    • Solar wind strips atmospheres. Planet rotation poles typically perpendicular to orbital plane. Uranus is sideways and bizzare magnetic field.
    • Intelligent life at 80% of Earth's habitable time before boiling. 1 By longer wouldn't cut it.
    • Eukaryotes arose from unlikely endosymbiosis event (Lane/Martin). Energy advantage from mitochondria providing membranes with local DNA control. Prokaryotes can't have local DNA copies to manage membranes.
    • If eukaryogenesis is the bottleneck, expect microbial life everywhere but complex life rare. Life started on Earth immediately, took 2+ billion years for complexity, fits.
    • Evolution has some theoretical limits (no free lunch theorem) and many (for now only empirical) walls like body plans locked since Cambrian and genetic code frozen.
    • We're probably early (Filter passed + Hanson's grabby aliens model). Von Neumann probes unseen.
  • LLMs are so good and motivating for self-research, it's crazy. Just wanted to go deeper.
  • Want to present (to my friends or city) about the fermi paradox, start a series of cafes!
    • Can be on Luma, seems SFy enough.
2026-04-16
  • Being without distractions (Substack, Spotify, Youtube) is really valuable, I always forget how much.
  • Everyone at Stohastics class is missing the ball, the teacher isn't intuitive. Claude is my only reliable study source (carrying laptop to class).
  • Maths test featuring partitions & k-sets.
  • Met a really agentic 1-year younger guy.
    • Really drilled into me that I need to optimise my linkedin.
    • It was interesting that he had like 3 projects, maybe I should also have multiple (arguably I do though).
    • Hope to have some coffee's or something, we could probably make something together, or just inspire each other.
    • Maybe I could do an advanced pedagogical theories coffee chat, that would be dope
  • Girl I liked turned out to have a boyfriend, hmm... I should go to more math conferences? One girl I really liked was like that.
  • Reframed the importance of Byzantium (Rome) and Western civilisation
2026-04-18
  • Ideal life: Claude Code running in-silico experiments in the background 24/7. Out pops a paper -> learning algorithms etc.
  • Made a private "insight notification" app: when I wake up it feeds me back random old insights to re-internalise.
  • A little cluster of NUCs at home that's always running stuff. The "idea lab" just goes. Give it a mail address people can send things to.
  • Feel like I'm on top of the world today. It's purely cause I'm well-slept & have a billion ideas. _Not a live entry. Backfilled from personal notes via Claude._
2026-04-29
  • Scott Alexander: writing hates dishonesty. You can see who someone is from how they write, it oozes out of the text, latent. Scary.
  • Finished Game of thrones. An insight: people need stories. Personal ones about yourself, & global ones. They orient you.
    • Stories/Reputations are powerlawy engines. Who has the strongest story wins. You need one for employers, stories sell products, the whole world runs on stories. Stories & identities. _Not a live entry. Backfilled from personal notes via Claude._
2026-05-05
  • Honestly planning + good sleep alone would already take me far far. Add gym & grinding obligations and it'd be crazy.
  • Life is simple: just follow your values! That's the whole recipe, because it forces optimisation and you'll be happy from it. (EA, animals, the suffering in the world.)
  • At uni I don't really learn deeply, it's surface level. Need to actually go through textbooks & do spaced repetition like I used to. _Not a live entry. Backfilled from personal notes via Claude._
2026-05-18
  • Went through a hackathon over the weekend. Had the top idea, but by some mistake it never got submitted so we weren't first. Really really mad about this because one of the prizes was ABC bootcamp taking place in San Francisco :(
    • Updated a bit on people. Human intelligence is jagged, usually more practical, weaker at evaluating ideas & abstract things, & heavily shaped by thinking patterns.
    • Black pill: most people aren't that creative (don't see things clearly), it seems easy to be one to stand out.
    • Bentham's "Everyone is horribly muddied and confused" seems applicable.
  • The solution is to become a monk. Do only 1-2 things, really well. From now on (?).
    • Want to just study a few areas for months, statistics, signal analysis, applicable maths, linear algebra. Re-establish spaced repetition.
  • Ordering 2 Cate Hall's "You Can Just Do Things" books, one for a friend one for me :) _Not a live entry. Backfilled from personal notes via Claude._
2026-05-24
  • Authentic voice is what carries power in public speaking, which in turn is derived from cohesion. Doing what you want, by your values, is the key. You get internal reasons for why you did things & you're only answerable to yourself.
    • Congruence & conviction is why Joan of Arc succeeded. You need to be fully congruent in your goals.
  • Want to study rhetoric, and Rome. The logistics amaze me, how did they recruit legions so fast, run tax collection, do life without modern tools? Family & honour as a huge driver we've mostly dropped. _Not a live entry. Backfilled from personal notes via Claude._
2026-05-29
  • Hopeful idea: what does democracy look like once everyone uses AI to get info & pick candidates? An AI that knows the media biases, reads tons of sources, thinks critically, tells you who to vote for. Would fix a lot of democracy's shallowness, & give new candidates a chance.
  • Radically change every X months (exploration) to find the thing worth exploiting.
  • Cate Hall: everyone has a default lens they attack the world with. My version: everyone has a tree of core beliefs that activate & interact into a personality, basically a quasi-HMM behind people that you fit to them. Fun project: build the model, fit it to people (& to myself) to explain behaviour.
  • Want real sources & names for the things I reference, so I'm better than an LLM emitting its distilled hallucinations. _Not a live entry. Backfilled from personal notes via Claude._
2026-05-30
  • Studying via Claude (Claude->human distillation).
  • Business is a specific game under constraints. You need a model with revenue, & there are recurring patterns (subscriptions, network effects). The biggest things I've built were all the "Topic + AI harness" lens.
  • You build knowledge through a filtered plausibility veil (worldview). For some people it goes into a bad attractor so they can't learn about the world. Feels semi-objective & formalisable.
  • Experiment: making tacit knowledge explicit. It's done for meditation / social anxiety. Eliezer, stories and woo transmit frames. Dance teaches via body metaphors and that's surprisingly effective.
  • Most things (writing, painting, talking to people) have a very wide possible domain of scenarios and possibility-space. It's astonishing how a well-crafted input can elicit a reaction and do work (charisma in speaking, for example).
    • Classis generator/discriminator problem, it's easy to recognise or be affected (good writing), but it's hard to generate sutch inputs (a good writer).
  • Ozzy courts people through their writing, reads their blogs. I should have my blog be like that it would be super cool!
  • Most of my problems come from my way of living. External interventions (walk outside longer, in the sun) are super effective but nearly impossible to internalise in advance, persistent cognitive blindspots. Unintuitive.
  • Should record my debates with my friend N and A (Henrik Karlsson Johanna do 10 min/day -> notes). _Not a live entry. Backfilled from personal notes via Claude._
2026-05-31
  • No phone, morning pages, I always forget how good they feel (very transient, doesn't stay in memory after).
  • Need to get properly good at rhetoric, otherwise I won't be able to hold effective meetings. And i want to tell good stories!
  • This giant water bottle next to me basically made me drink water constantly. Wonder how much that moves my mood vs walks / morning pages / vitamin D.
  • Claude debate: fruit isn't actually that important (lots of sugar). Discovering sugar/energy curves. _Not a live entry. Backfilled from personal notes via Claude._
2026-06-02
  • ~7h selection of people for our Maths summer camp (participants and mentors)! I tend to be too leniant (saying yes/maybe to too many people), so my collegues get their judgment and reject more people.
    • From picking camp kids I got a real feel for uplift: give a shot to kids who never had competitive math tutorship instead of just improving the players already in the game. Environment determines your direction too much, an ideal agent would converge to a stable coherent form independent of place.
  • Told J. how I progressed in German rapidly in 25 days at ~1h/day via Anki, inspired us both. 1h/day isn't much. Need to do these things more.
  • Data science vs ML talk w M: data science won't be replaced (at least not soon), because it's different from ML. Need to study way more maths, couldn't confidently tutor my sister on F-tests & ANOVA.
  • Intervention ranking (current best guess): Sleep > weekly rule-ban > morning routine > morning walk > morning pages > water.
  • Real life as AI agent harness mismanagement. No routine is one of the biggest predictors of nonadulthood.
    • Theories I'd love to actually work on: the effect of mentorship and institutions on development, behaviour basins, the dark-matter knowledge missing from modern education (the agentic loop, routines as the real predictor of adulthood).
  • You gain a lot by just taking a cause seriously and fighting for it, coming from your internal drives and smart research, not picking up on the internet or outside sources issues.
  • Bought noise plugs. _Not a live entry. Backfilled from personal notes via Claude._
2026-06-04
  • Should write a practical opinion post on mentorship structures & how high-leverage those interventions are. Two theses: (1) how an idealised agent/person can move (almost exponentially), how place, environment, education & life interventions shape you, a comprehensive map of the "dark matter". (2) as a subset of that, the importance of institutions, groups & communities in shaping a life. _Not a live entry. Backfilled from personal notes via Claude._
2026-06-05
  • Back from a long gap of non-web publishing (from April 16th to June 5th). Sorry! To not have a criminally large gap, I let Claude through my inconsistent morning pages and backfilled and lightly edited some effectively random days.
    • Lots of interesting undocumented things happened, like building my 24h agentic harness called Apollo, going to an algotrading hackathon, consequent insights about vibe slop and the importance of walks, etc.
    • Trying to memory-rewrite last month wouldn't capture my daily idiosyncricity enough.
  • Am holding up a lot better with new morning routine + timeblocking + Todoist stack! Life now feels like an hour allocation problem.
    • Lots of feasability questions about high agency people are clear to me now.
  • read [David Oks' essay](https://davidoks.blog/p/why-china-got-rich-and-india-didnt) explaining how China developed faster than India because of immaterial factors like education and mindset and dismantling of old oppresive and illogical traditions.
    • I agree and think immaterial factors (like "mindset") are why the West is so developed and why civilisation is progressing.
    • The most immaterially latently advanced place in the world right now is San Francisco, among the best social technology. In X years, it will diffuse to the rest of the world ("agency" and their framings of exponential thinking, models, worldviews... will become ubiquitous terms and ideas). Because it's better. That's also why they have so many results for the population.

Beliefs