Fri May 15 • 13:48:43

shubham

🦋

building hue. personal intelligence. like artificial intelligence, but it actually knows you. dm me on twitter or if any of this resonates.

🎓

taught frontier language models, a class on llms from the ground up. all lectures are on youtube.

yo,

welcome to this little space of mine on the internet

i'm shubham. llm researcher, accidental quant, deliberate founder. 30, in new york. spent the last decade trying to understand these systems from the inside out. how the math turns into a thing that seems to know what you mean.

currently building hue at strange intelligence. a personal intelligence layer that lives in your imessages, learns who you are from your data, and reaches out to your friends' agents on your behalf. agent-to-agent communication, but the agent is shaped by you. the bet is that personal models are the missing piece. not bigger, just yours.

before this, did ai research at vatic labs (a small quant fund, agents learning market predictions instead of ml models). before that, trained language models at github copilot and shipped intellisense pieces into vs code at microsoft. did my masters at nyu, worked with yann lecun. the path from curious about how neural nets actually work to trying to build one that knows you has been a long but pretty linear line.

i think in arrows. when the target is chosen, execution is the easy part. i can lock into a pod for three hours and ship a flow. the hard part is choosing. paralyzed by dimensionality, not volume. wants to do everything: build, host events, read papers, teach, write. in practice i do one or two things per week at depth and let the rest slide.

outside work i host research salons and game nights in nyc. small rooms, real conversations about consciousness, alignment, the shape of these models. the optimal go-to-market for hue turned out to not be a marketing funnel. it's a social calendar. every event is a conversion event.

reading: scott alexander style essays, ai consciousness papers, indirect realism, jhana phenomenology. trying to write the kind of essay i wish existed. the gradual disempowerment and the artificial self. translating frontier ai for people who don't read academic papers.

i like meeting people. if any of this resonates (building agents, personal intelligence, ai consciousness, or you just want to get a coffee in new york) reach out via twitter or . or skim my links, 2000+ things i've read and saved.

- shubham

***