Live
AI pair-programming
11.5B tokens this month (incl. cached context) · ~$14.5k at API rates
Claude Code · 9.3B · 75 sessions · ~$14.1kCodex · 2.2B · 58 sessions · ~$390.80Kimi · 37.7M · 2 sessions · ~$3.80
14 days agotokens / day, stacked · peak 2.0B on Jul 17 · updated Jul 18today
GitHub — Elliot-Sones
Elliot's GitHub streak stats — total contributions, current streak, longest streak
26 weeks agohover a square for detailtoday
What I'm working on
whisper-flowbuilding

Personal push-to-talk dictation for macOS

probabilistic-pitcher-physicsmodeling

Pitcher physics modeling and validation from public Statcast data

This portfolioshipping

Next.js 16 · MDX · live stats — you're looking at it

What I'm learning
Kernel-level optimization for transformer modelscurrent

Writing and tuning GPU kernels instead of importing them

About

When I'm not coding…

I'm Elliot — a CS student at Toronto Metropolitan University training agents that play games and building tools that use LLMs. Most of my learning happens in public: hackathons, Kaggle, and an unreasonable number of tokens.

When I'm not coding, I play as much soccer as I can — I've played my whole life, including one season in Portugal. When I'm not playing, it's music, anime, and friends.

Favorite anime
One Piece
One Piece
Attack on Titan
Attack on Titan
Jujutsu Kaisen
Jujutsu Kaisen
Solo Leveling
Solo Leveling
AI Agents

Building autonomous agents that can reason, plan, and take actions — from RL-based game agents to tool-using LLM agents that solve real problems.

Deep Learning & MLOps

Training models from scratch (transformers, CNNs, RNNs) and deploying them — with a focus on reinforcement learning and attention mechanisms.