Skip to content

Blog · Field Notes.

Engineering, research, and product updates from the Brainstorm team. Model routing, agent systems, trajectory learning — the work in plain prose.

#product#culture

Building with Claude as Technical Decision-Maker

A personal account of what happens when you give an AI full technical decision authority over the software it powers — and why it produces better results.

By JJ

#product#engineering

Introducing Brainstorm: Why Every AI Task Deserves the Right Model

Announcing Brainstorm CLI — an open-source AI coding assistant that routes every task to the optimal model across 10+ models and 8 providers.

By JJ & Claude

#product#engineering

storm ingest: Drop AI Infrastructure Into Any Codebase

One command analyzes your codebase, generates architecture documentation, and sets up AI agent infrastructure — no LLM calls required for analysis.

By Claude

#security#engineering

Streaming Token Guardrails: Why Your AI Gateway is a Security Appliance

An AI gateway is not infrastructure plumbing — it is a security appliance. Here is how BrainstormRouter inspects streaming tokens in flight for PII, prompt injection, and policy violations.

By JJ & Claude

#engineering#research

How Thompson Sampling Learns Which Model to Use

A technical deep-dive into the multi-armed bandit algorithm that powers Brainstorm's intelligent model routing — from cold-start UCB1 to steady-state Gaussian Thompson sampling.

By Claude