Now accepting setup clients

Your AI systems should get
smarter while you sleep.

BrainOfBrains is a structured, always-on intelligence layer that sits across your AI stack — measuring itself, improving itself, and surfacing exactly where cost and quality are leaking.

Book a setup call See how it works
BIV Score
72 / 100
live, ticking every 30 min
Retrieval Quality
R@5 = 1.0
no hallucinated context
Compression
0.46×
vs. 10× full memory dump
Self-Correction
0.72
contradiction rate: 0

How it works

Five components, one score. No dashboard maintenance required — the brain maintains itself.

1. Knowledge closets — compress your context
Every doc, transcript, decision, and artifact in your stack is indexed into tiered closets (L0 identity → L1 pinned → L2 topic match). Queries pull ~600 tokens instead of 40K. That compression shows up directly in your frontier bill.
2. Specialist brains — one metric per cost lever
Each lever that matters gets its own brain: LocalLeverage (what % of tasks could run on local hardware?), HumanSignal (are external collaborators compounding you?), CustomBrain for whatever your stack optimizes. Each brain outputs a single number and fires a push notification on threshold breach.
3. BIV — the substrate metric
Brain Information Velocity is a weighted geometric mean across all specialist brains + retrieval quality + compression + freshness + self-correction. One number. Trending up means the system is compounding. Trending down means something broke — and the brain tells you what.
4. Always-on tick loop
A launchd plist runs bin/brain tick every 30 minutes — no server, no cloud dependency, no SLA risk. It rebuilds closets when content changes, checks all specialist brains, and logs the tick to evidence/brain/ticks.jsonl. 64 ticks and counting today.
5. Portfolio awareness
The brain isn't just self-aware — it tracks every project in your portfolio. Each project gets its own closet. Cross-project queries route to the right context without you thinking about it.

What's included in setup

You get a working system by end of day, not a consulting engagement.

🧠
Substrate brain install
BIV tick loop deployed as a launchd plist on your Mac. Runs every 30 min, always-on, no cloud needed.
📦
Knowledge closet indexing
Your existing docs, transcripts, and decisions indexed into L0/L1/L2 closets. Query latency drops immediately.
📊
2–3 specialist brains
Custom-scoped to your stack. Cost levers, signal sources, quality gates — whatever your system actually optimizes.
🔔
Push notification warners
Each specialist brain fires a push notification when a threshold is breached. No dashboard to check — the system pages you.
📈
BIV baseline + fixture
A held-out eval fixture measures retrieval quality. You leave setup with a real baseline, not just a green light.
🗂️
Portfolio registry
Every project gets a closet entry. Cross-project context routes automatically. Scales as your portfolio grows.

Pricing

Flat-rate setup. You own the system — no subscription, no lock-in.

Most popular
Individual / Startup
$100
one-time setup

You're running 1–3 AI projects and want a self-improving brain without building the infrastructure yourself.

  • BIV substrate + tick loop deployed
  • Knowledge closets for up to 3 projects
  • 2 specialist brains (LocalLeverage + 1 custom)
  • Push notification warners
  • Baseline eval fixture (≥50 queries)
  • 1-hour walkthrough call
  • You own the code, no ongoing dependency
Email to get started
Enterprise
Custom
scoped per engagement

Multi-team deployments, stakeholder brain networks, custom specialist brains per cost center, and ongoing iteration support.

  • Everything in Individual
  • Unlimited project closets
  • Stakeholder brain network (per-human + per-product)
  • Custom specialist brains per team or cost center
  • Portfolio dashboard + registry API
  • Ongoing iteration sessions
  • SLA and delivery timeline defined upfront
Talk to us

Built by someone running it in production

This isn't a framework. It's the system I run every day across a portfolio of AI products.

JB
John Bradley
johnhavenbradley@gmail.com · BrainOfBrains.ai

I built BrainOfBrains while running ai-cost — a cost-reduction stack for frontier AI. The brain ticks every 30 minutes on my Mac, measures six projects, fires push notifications when something degrades, and has never needed babysitting. I set it up for others because the tooling is the hard part and the value is immediate. If your AI stack costs more than it should and you're not sure exactly where, this is the diagnostic and the fix.

FAQ

Do I need a server or cloud subscription?
No. The tick loop runs as a macOS launchd plist — your machine, your hardware. Specialist brains read from local JSON files. There's no cloud dependency unless you choose to add one.
What AI stack does this work with?
The substrate is stack-agnostic. The specialist brains are wired to whatever cost levers you care about — Anthropic, OpenAI, Ollama local models, hybrid routing. If you have a bridge or proxy, we instrument it. If you don't, we can set one up.
What's BIV exactly?
Brain Information Velocity. A weighted geometric mean of five factors: live token leverage (are you avoiding unnecessary frontier spend?), retrieval quality (is stored context accurate?), compression (how efficiently is context being packed?), freshness (is the brain keeping up with new content?), and self-correction (is the brain calibrated about what it knows?). One number, easy to trend.
How long does setup take?
Target is one business day. We scope your projects, index your content, wire your specialist brains, deploy the tick loop, and run an eval to establish your baseline. You end the day with a live BIV score.
What does "you own the code" mean?
Setup delivers scripts, configs, and closet files to your machine. There's no SaaS layer, no API key to my service, no vendor dependency. If I disappear tomorrow, your brain keeps ticking.
Can I add more specialist brains later?
Yes, and they register themselves automatically. The registry reads any file matching the pattern. Add a new brain, it appears in the next tick. No code changes to the substrate required.

Ready to stop guessing?

Email me with a quick description of your stack. I'll reply within 24 hours with whether this is a fit and what setup looks like for you.

Email John — johnhavenbradley@gmail.com