brendan lo
chicago, ilhi, i'm brendan lo, 18, cs + math @ uchicago. i enjoy ai/ml research, hanging out with my friends, and building really cool things.
highlights & accomplishments
3x icml, yc sus ’26, eagle scout, 2nd @ nyc hackathon, 1x journal of cell biology paper, frc robotics world’s engineering inspiration award, and a bunch of projects totalling 10k users.
my takes
- philosophy is important; but uncontrolled philosophical thinking -> arrogance.
- distribution is the most important thing not at scale; but the product is the most important thing at scale.
- agi will be reached when we are able to make something capable of all human senses.
- for ex, neural networks and CoT reasoning are partially based on how the brain works.
- world models, better synthetic data generation (digital twins or better?), llms, and some new architecture (still unknown) will get us there.
- frontend is more important than ever. true, human originality and art is vital to ui/ux.
- llm’s will keep improving, but not to the point of understanding ai slop. all the frontier models will produce new and newer outputs, which will eventually develop a consensus of ai slop.
- full-stack is even more important now than ever with ai coding.
- swe’s won’t be replaced by ai; instead, they will be expected to get more work done. more code gets written, so more of the job becomes architecting it, reviewing it, and knowing why it breaks.
- there’s a massive backlog of ideas from people that finally gets built by an ai-pilled engineer.
- security matters more than effort now. when code is cheap to write, the scarce thing isn’t how hard you worked on it — it’s whether anyone checked what it exposes.
- it’s just like the compiler. people called the compiler cheating, and talked about how they preferred assembly. then they called python inefficient. and so on…