One accuracy number hides opposite failures. Astra fails 91% of bad runs but passes only 51% of good ones. Gemini 3.1 Pro is the mirror: 83% of good runs accepted, 25% of bad runs caught. High reasoning effort moved Astra one way only: 71 to 89 caught, 27 to 27 accepted.

A judge that is wrong is usually wrong alone. Claude Fable 5.1 scores 36% on its own misses and 90% on GPT-6 Astra's. Astra: 38% on its own, 93% on Fable's. Opus 5 selected none of the items and still drops to 44% on Fable's misses. Blind spots run in families.

Who grades the graders? LLM-as-judge on short answers is well studied. Agent judges are not: they open the files, run the code, and investigate before ruling on another agent's work. Introducing Arbiter-Bench: 71 real agent runs that frontier agent judges get wrong.

apparently doomscrolling improves my sleep quality. do I trust my whoop or @bryan_johnson?

Built with Shuyu van Kerkwijk, Harry Cheng, and Isabel Lew. Thank you to the @CalHacks organizers for this opportunity!

Recent research (Ibrahim et al., Nature 2024) shows targeted muscle stimulation can eliminate FOG. We used a wearable sensor (MPU-6050) to collect leg motion data. Using our model's prediction, a robotic arm delivers a localized “poke” to end the freezing episode.

This weekend, we built a robotic wearable that helps Parkinson’s patients walk independently. A 1D CNN + LSTM model predicts freezing of gait in real-time and triggers mechanical stimulation to restore movement. 1st Place + Hacker’s Choice @ Berkeley AI Hackathon

intern vs founder for 10% equity