Home

Uses

The toolchain behind the work.

What I use every day to build, ship, and operate AI systems, and what each tool is for.

01

Coding agents

Most of my code is written with agents and reviewed by me. The skill is in the specs, the tests, and the review.

Codex
Parallel branches and reviews. Hundreds of codex/* branches across the DeepChamp repositories.
Claude Code
Long-running build, debug, and release sessions, steered by repository rules files.
Cursor
Editing and quick refactors inside the IDE.
02

MCP servers

Agents operate real systems through MCP instead of screenshots and copy-paste.

App Store Connect
Builds, versions, review notes, and subscriptions.
Firebase
Auth, Firestore, and project configuration.
Stripe
Products, prices, and webhook debugging.
AppsFlyer and Superwall docs
Attribution and paywall reference while building.
OpenAI developer docs
API reference inside the agent's context.
03

Models

Each model has one job, and code owns every threshold.

OpenAI
Real-time voice over WebRTC, conversation and tool orchestration, evidence-gated expert reasoning.
Gemini
Grounded research, structured output, and Live narration.
Amazon Bedrock
An opt-in reasoning path with a fallback chain.
TypeSafe Jev
Typed, calibrated judgments: scores, choices, and yes-or-no probabilities with confidence.
04

Cloud and data

Boring, observable infrastructure, so the interesting parts can move fast.

Google Cloud
Cloud Run, Cloud Functions, Cloud Scheduler, and Spot VMs.
Firebase
Auth, App Check, Firestore, and Crashlytics.
BigQuery and Dataform
The warehouse, with assertions and freshness targets.
Vercel and Next.js
Web apps, dashboards, and this site.
05

Apps and media

Native where it matters, and media pipelines that run unattended.

SwiftUI and StoreKit 2
Native iOS apps, widgets, Live Activities, and subscriptions.
Superwall
Paywall experiments with holdouts.
ffmpeg, Demucs, Essentia
Audio analysis, rendering, and five-hour film encodes.
Remotion and HyperFrames
Lyric videos and product films rendered from code.
06

Quality

Nothing ships on vibes.

pytest, XCTest, node:test
More than 17,000 automated tests across app, services, and warehouse.
Eval harnesses
A voice autograder and a text eval catalog with a grader.
Deploy gates
Deploy scripts that stop on the first failing test.