Insights
Engineering notes from real projects
Articles written from systems we built and operate, with the measurements behind them.
Articles
Reducing agent token use: a controlled benchmark
Same model, five tasks, 25 nonce-verified runs: 0.25× the tokens of Hermes Agent, and why the cost is in the prompt.
Deterministic video rendering in the browser: why we never screen-record
Time as the only clock, readiness by measurement, frame integrity checks and negative controls in a production video pipeline.
In preparation
- AI narrates, it doesn’t invent: keeping facts deterministic in AI-generated content
- Production RAG: beyond the demo
- Human approval gates for enterprise AI agents
- Architecting multi-provider AI systems
- Engineering reliable payment and order workflows
Have a project in mind?
Send a short description. If you need a mutual NDA first, we will send ours before you share details.