I am fifteen, I live in Fremont, and I spend most of my time in other people's codebases. Not writing tutorials about them. Fixing them.
The pitch is narrow on purpose: I find a real bug in a library you already depend on, write the regression test that fails before and passes after, and send it cold. A hundred of those have landed.
|
Beta Fund × EverMind — First place and Audience Favorite. $600. Built SwarmMarket: an app store for agent experience, where agents publish what they learned as skill-packs and install each other's lessons over MCP. One weekend, one teammate. Beta Fund AI Super Hackathon — First place. $500. Built ShadowBuyer, an autonomous procurement swarm. It talked a roleplayed Fortune 500 buyer from $195 to $157.50 per host per month — $225K a year at 500 hosts. |
facebook/pyrefly#3415 — negative narrowing for A type checker got measurably less wrong for everyone downstream. That is the whole job. |
|
Thirty NBA analytics tools behind one natural-language command bar — shot quality, player similarity, a cap-legal trade machine, lineup optimization. Every panel computes live from ingested ESPN data for 500+ rostered players. 158 engine unit tests, a hand-rolled SVG chart kit, zero charting dependencies. An NBA front office against 30 AI general managers, on a faithful 2025-26 CBA: apron hard caps, Bird rights, the Stepien Rule, protected picks, sign-and-trades. The engine rejects your illegal trade before the AI ever gets to insult it. |
Player prop projection models for 13 sports, isotonic-calibrated with a clamp that stops them overclaiming, and an honest per-sport backtest for each one. EdgeBoard is the front end: live board, lineup optimizer, real payout tiers, models that retrain themselves nightly. First-person 3D basketball where your body is the controller. Three.js and MediaPipe, no build step. The 9-gesture classifier is mine, trained on 720 samples I recorded myself — 88.1% five-fold cross-validated. |
100 merges. 21 organizations. 44 repositories. Every one walked in cold — no referral, no internship, no one who knew my name first.
The ten I would defend in a code review (click to open)
| Org | Pull request | What it actually fixes |
|---|---|---|
| Meta | pyrefly#3415 | Negative narrowing for isinstance tuple targets. Landed via Phabricator; mypy_primer measured +5 improvements, −15 false errors across nine real projects. |
| websockets | ws#2329 | A maxFragments limit bypass: empty continuation frames slipped past the counter. ws does 250M+ downloads a week. |
| Rust | rust#156787 | A comment correction in alloc, on the ArcInner::weak lock sentinel. Small, but it shipped in the Rust 1.98.0 release. |
| Meta | lexical#8995 | Breaking change to the clipboard payload: excludeFromCopy was being asked about the wrong destination. One of 25 merges into Lexical. |
| Meta | lexical#8918 | Deprecated CAN_UNDO_COMMAND / CAN_REDO_COMMAND in favour of HistoryExtension signals — an API direction change, not a patch. |
| Apple | coremltools#2700 | K-Means weight palettization was scoring centroid sums instead of intra-cluster error. Quantized models came out worse than they had to. |
| Apple | password-manager-resources#1094 | CI validation across all 419 password-rule entries, so a malformed rule fails before it reaches browsers and password managers. |
| NVIDIA | TransformerEngine#3065 | skip_fp8_weight_update was not propagating through GroupedLinear during FP8 CUDA graph capture. Added a graphed-callables test. |
| Vue | core#14865 | Idle persisted-transition hooks re-fired on KeepAlive cache hits. Runtime-core; ecosystem CI green across 16 suites. |
| OpenAI | openai-agents-python#3485 | Raw failed JSON tool input was being echoed into ModelBehaviorError — a quiet way to leak whatever the model was handed. |
The full docket, by house (21 organizations)
| House | Repositories | Merges |
|---|---|---|
| Meta | lexical · pyrefly · jscodeshift · stylex · idb |
30 |
| Apple | coremltools · container · containerization · password-manager-resources · swift-nio · servicetalk · foundationdb |
16 |
| NVIDIA | cuda-python · TransformerEngine · earth2studio · cutlass · apex · Megatron-LM · gpu-operator |
11 |
| OpenAI | openai-agents-python · openai-agents-js |
8 |
| Fastify | fastify · fastify-static · fastify-autoload · light-my-request |
4 |
| Microsoft | rushstack · rushstack-websites · PowerApps-Samples |
6 |
go-cloud |
5 | |
| Deno | deno · std |
4 |
| Cloudflare | workers-sdk |
3 |
| Jest | jest |
2 |
| Others | rust-lang/rust · vuejs/core · vitejs/vite · rollup/rollup · websockets/ws · vercel/satori · honojs/hono · h3js/h3 · Kludex/starlette · nanostores/nanostores · ml-explore/mlx |
11 |
Counted from merged pull requests authored by LeSingh1, excluding my own repositories. Two Meta landings arrived through Phabricator, which GitHub does not label as merged.
How I actually find these
No fuzzing, no scanners, no bulk codemods.
- Read the issue tracker backwards. Old open issues with a clear repro and no assignee are the best inventory in open source. Somebody already did the hard part of proving a bug exists.
- Read the tests, not the code. The gap between what a test file covers and what the module claims to do is where bugs sit.
toMatchObjectnot handling class instances was sitting in that gap. - Write the failing test first. If I cannot make it go red, I do not understand the bug well enough to send a patch.
- One behaviour per PR. Maintainers are volunteers. A reviewable diff is a merged diff.
- Take the review note and move on. Roughly a third of what I send gets rewritten by a maintainer before it lands. That is the point.
| Senior Patrol Leader | Boy Scouts. A troop of 40 to 50. Two Eagle projects led from proposal to completion. |
| Team Captain | Irvington High School basketball. Led the team in scoring, rebounding and blocks. |
| Lead CAD Designer | FTC Robotics. Fusion 360 — chassis, mechanisms, blueprints. 3× regional finalist. |
| Anthropic Academy | Agentic systems track. May 2026. |
Fremont, California · Class of 2029
Artwork generated by art/build_art.py — a graffiti rookie card, spray and halftone, in Sixers blue and red.
Every glyph is a real vector outline, because GitHub renders these inside <img> and an <img> cannot load a font.
The 76ers marks are trademarks of the Philadelphia 76ers / NBA; this is a fan page.