Build log · · 1 min read · shipped
Build log: shipping Unseen UI to npm
How the component library behind this site went from a private workspace to two published packages, including the three things the first consumer broke.
Maps, decisions and failures from enterprise systems
Real systems,
revealed.
How enterprise systems fit together, as maps you can open: what each part does, what I decided and what broke. AI first, because that is where I build now, on the data, integration, cloud and security platforms it depends on.
entries121
maps6
series published18
last entryOct 8, 2026
How to read this
Each box is a role. Pick a lens to see the system as a networking, data, security or identity problem. Pick a stack, or mix stacks per component, to see the product that fills each role.
- Request
- Async or batch
- Telemetry and audit
- Trust boundary
- Cross-platform seam
- Seam not drawn as a line
Adds identity carried end to end, secrets in a vault, an audit record for every call, and real systems of record behind the tools.
Components
All disciplines. Roles only.
The whole estate
Six maps, one company
AI is the top layer, not the whole system. It runs on the platforms below and answers to the controls down the side. 1 of 6 maps are drawn in full, 2 are documented and 3 are being drawn.
- Drawn · interactiveAI Application From idea to production · 19 components
- Being drawnDeveloper Platforms Tools, workflows and developer experience · 4 components
AI plugs in atAgentObservability
- DocumentedEnterprise Integrations Connect systems and people · 4 components
AI plugs in atEnterprise APIsToolsAPI Gateway
- Being drawnData Platforms Pipelines, storage and analytics · 5 components
AI plugs in atSourcesIngestionVector IndexMemory
- Being drawnCloud & Infrastructure Compute, networking and scale · 4 components
AI plugs in atEdgeLLM GatewayModel
Read in order
Pick a track
8 five-day tracks of five short parts each, from the first decision to running it in production. Or follow a path: Agents in production, without a shared robot, Enterprise RAG, from problem to production, New to RAG, in order.
- Agent InfrastructureThe runtime underneath agents: autonomy, isolation, scaling and failure.5-day track · 5 parts
- MCP for Production AIWhat the Model Context Protocol standardises, where servers run, and when MCP earns its overhead.5-day track · 5 parts
- Multi-Agent System DesignWhen several agents actually help, and what it takes to run them: role contracts and handoffs, who controls the run, budgets and stop conditions, traces and recovery.5-day track · 5 parts
- AI EvaluationEvaluation that can block a release: what an eval may stop, golden sets maintained like products, calibrated judges, regression gates and evaluation after launch.5-day track · 5 parts
- AI Safety EngineeringSafety as architecture rather than moderation: threat models, agent identity and scope, untrusted input, blast radius and data exfiltration.5-day track · 5 parts
- AI Operations and ObservabilityRunning AI in production: traces and spans, cost, latency and capacity, production evaluation signals, incidents and rollbacks, drift and an operating cadence.5-day track · 5 parts
- AI Architecture PatternsThe building blocks of a production AI system: the model gateway, choosing the right pattern, control and data planes, approvals and tenant boundaries, and a reference architecture.5-day track · 5 parts
- Production RAGRetrieval as a system you operate: trusted sources and freshness, ingestion as a data product, retrieval as its own subsystem, chunking for the question, and citations you can evaluate.5-day track · 5 parts
Read by discipline
Pick the question you actually have.
- AI & ModelsWhat does the model actually do, and where does it need the system to help?60 entries
- Data & DatabasesWhere does data live, how does it move and how fresh is it?43 entries
- NetworkingWhich paths does traffic take, and where does it leave your network?5 entries
- SecurityWhere are the trust boundaries, and what can cross them?28 entries
- Identity & AccessWho is acting, on whose behalf, with which permissions?7 entries
- IntegrationHow does this connect to the systems the business already runs?23 entries
- Compute & InfraWhat runs where, and how does it scale and fail?21 entries
- Observability & OpsHow do you know it is working, and what do you do when it is not?22 entries
- Cost & GovernanceWhat does it cost, who approved it, and what is on the record?14 entries
Recently written
The journal
Comparison · · 1 min read
LangGraph vs the OpenAI Agents SDK
Two ways to write the same supervisor. Compared on control flow, tracing, provider coupling, testing and what each makes hard.
Deep dive · · 6 min read
Every question your AI readiness review asks was answered months ago
The last six days of a sixty-day series, and the pattern is that operability gets bought early or it does not get bought at all.
Checklist · · 12 checks
RAG production-readiness checklist
Twelve checks to pass before a retrieval system answers a real user. Tick them locally; progress stays in your browser.