Notes from building AI and automation systems in production — retrieval architectures, agentic workflows, and the delivery pipelines that get them shipped.
A production rate limiter is not a counter bolted onto an endpoint. It is a distributed admission-control system that has to be accurate, fast, synchronized and honest with the clients it throttles.
From a local commit to a zero-downtime production release — repos, build artifacts, service connections, gated approvals and slot swaps, in the order you actually build them.
Vector databases became the default answer for retrieval. For a lot of enterprise knowledge work, they are more machinery than the problem deserves.
Autonomy is not the goal. Reliable outcomes are. A look at the guardrails that make multi-step agents safe enough to put in front of a business process.
The gains rarely come from clever phrasing. They come from deciding what the model gets to see, and in what order.