Inside Glasshouse, the context layer we run on
Why we built Glasshouse, how a sentence in a note becomes a fact an AI agent can trust, and what it costs to run, in eleven screenshots.
[Thinking]
Essays on context, identity, agents and the guardrails around them.
10 / 2026
AI agent governance can’t be bolted on
What AI agent governance is, what the EU AI Act, NIST and Singapore expect, and why guardrails have to be designed into the system, not bolted onto each agent.
Agent sprawl is a buying problem
What agent sprawl is, why buying AI agents tool by tool causes it, and what to build instead: one agent factory with shared context, identity and guardrails.
Static context is the Jedi Archives problem
Why markdown files, vaults and system prompts fail AI agents the moment reality changes, and what to build underneath them instead.
The knowledge graph your PIM thinks it is
Why product data models fail agents, and what a described relationship actually buys you.
Identity resolution is an actuation problem
Match rates are a vanity metric until an agent acts on the match.
Deterministic plumbing, sold as autonomy
A field guide to what is actually running in production agent stacks.
The brackets: constraint design for enterprise agents
Budgets, blast radius, and the approvals a CFO will sign.