Senior Software Engineer with 10+ years of experience building backend systems, internal platforms, workflow automation, and applied AI products.
My recent work includes production LLM pipelines, MCP-based tools, and Kubernetes services.
Every token my AI coding agents burn is metered — Claude Code, Codex CLI, and Antigravity session logs sync to Postgres via a small usage-collector script, and a nightly Action rebuilds this section from the database.
7.53B tokens processed · 45,969 requests · 17 models · since March 2026
| Agent | Requests | Input | Output | Cache reads |
|---|---|---|---|---|
| 🟠 Claude Code (Anthropic) | 42,692 | 3.4M | 56.1M | 6.86B |
| ⚪ Codex CLI (OpenAI) | 3,105 | 10.4M | 1.2M | 319.8M |
| 🔵 Antigravity (Google) | 172 | 1.1M | 183K | 15.4M |
Top models by output tokens
| Model | Output | Requests |
|---|---|---|
| claude-fable-5-1 | 21.3M | 6,175 |
| claude-fable-5 | 21.2M | 15,124 |
| claude-opus-5 | 6.1M | 12,708 |
| claude-opus-4-8 | 4.9M | 4,721 |
| claude-sonnet-4-6 | 1.3M | 1,866 |
| gpt-5.6-sol | 762K | 2,058 |
| claude-opus-4-6 | 639K | 874 |
| claude-sonnet-5 | 479K | 659 |
| gpt-5.5 | 308K | 742 |
| gemini-3-flash-a | 166K | 160 |
Last 30 days — tokens/day, peak 8.2M on Sep 9
█
█
█ █
█ █ █
█ █ █ █ █ █ █ █ █ █
█ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █ █
Updated 2026-09-23 · input/output exclude cache reads