guanni (jenny) qu

i'm jenny. pebblebed gave me money to build AI that finds zero days first and i am locked in.

math/physics/cs at uc davis. rl for math agi at caltech.

RESEARCH

---------

- reinforcement learning: distributional rl, rl with dynamic action space

- qft

- math: allen hatcher, 3-manifolds, serre's a course in arithmetic

- ctfs: team squid proxy lover, crypto & pwn

- security: smartkuang

- ai control

HOBBIES

--------

- skiing: palisades tahoe

- piano: chopin ballade no.1, mendelssohn andante and rondo capriccioso,

beethoven sonata op.109, chopin fantasie impromptu

- guannigolf: 12k+ follower wechat channel (golf slang, ai safety, ea, qft, math)

- golf

- art

ON-POLICY LEARNING

-------------------

- 80/20 by default; stop at diminishing returns

- optimize for expected impact, not vibes

- meditation, yoga, gym, sleep, water: basic maintenance of hardware

- try not to overfit on any single datapoint, person, paper, job, or experience

- default reaction to chaos: contain, sleep, lift, then think

- choose actions that make my five-years-from-now self maximally happy

- start at the north star, plan backwards, then walk forwards

- slope of trajectory over current value of position

COOL TO THINK ABOUT

--------------------

- visualizing 3-manifolds

- what it would be like to live on the klein bottle

- the hat problem from elliot glazer

- gpu parallelization

- visualizing p-adic field and p-adic ring

- how a golf ball compresses at impact

- visualizing the vector space of all continuous functions on [0, 1]

- visualizing the vector space of all real valued sequences

- visualizing how em waves (ASK, FSK, PSK, QAM) fill the "empty" space around us

- what it feels like to be a plant/boulder/pebble/squirrel

// if you're working on rl, qft, ai security, or love ctfs, let's talk.

pebblebed