Notes from
the team.
Product, engineering, and research updates from the people building Brainmox. Written for operators, not analysts.
AI assistant evaluation scorecard: a practical worksheet
Score an AI assistant on one real workflow, with evidence for outcomes, memory, permissions, recovery, privacy and review—not feature-page promises.
All
posts.
Why Brainmox costs $299 per month
What the $299 Brainmox plan includes, why we chose a flat price, how to evaluate the cost, and who should not buy it.
Building Brainmox: Why We Made Privacy Non-Negotiable
An AI agent that reads your files and works your inbox needs a higher privacy bar than a chatbot. The five rules we build Brainmox around, and why.
What a stable identity adds to an AI agent
A stable agent identity can make multi-channel work easier to recognize and review—but it does not replace memory, permissions or human accountability.
Why an AI says “I don't know”—and how to diagnose it
An AI may lack the source, fail to retrieve it, face a permission boundary, or be uncertain. Here is a practical way to find the actual cause.
How to keep an AI agent loop from running up costs
A practical look at repeat-call detection, idempotency, call budgets and audit logs—and the limits of each control.
AI agent vs copilot vs chatbot: a practical buying guide
Agent, copilot and chatbot features increasingly overlap. Compare autonomy, scope, memory, tools, approvals and evidence instead of trusting the label.
What persistent AI memory means—and how to test it
AI products use the word memory for different systems. Learn the practical distinctions and test recall, provenance, correction and deletion yourself.
The AI assistant test: 12 capabilities to verify before you buy
A vendor-neutral test for memory, scheduling, tools, permissions and auditability—based on a real workflow rather than a feature-page checklist.