Write Tool Contracts That Survive Ambiguous Requests
Reliable agent tools need clear meanings for identity, filters, missing fields, result completeness and errors, beyond valid structured input.
Agentic Systems 3 min read
Notes from building and operating AI systems, commerce platforms and the measurement around them.
Reliable agent tools need clear meanings for identity, filters, missing fields, result completeness and errors, beyond valid structured input.
Agentic Systems 3 min read
Separate fluent generation from evidence checks for prices, product identifiers, links, and policy claims before releasing an AI response.
AI Engineering 3 min read
A practical way to decide which steps need model judgment, which belong in code, and where an agent should be allowed to choose its next action.
Agentic Systems 3 min read
Combine semantic retrieval with lexical search while preserving identifiers, access filters, and a measurable baseline for ranking quality.
AI Engineering 3 min read
A retrieval system needs evidence for answering and a deliberate path for missing evidence. Negative cases make that boundary measurable.
AI Engineering 3 min read
Build webhook receivers that verify input, accept work durably, handle duplicate deliveries, and reconcile state when events arrive late.
Backend Architecture 3 min read
Why tenant isolation must extend through database queries, caches, background jobs, retrieval, and support tooling in a SaaS application.
Backend Architecture 3 min read
Lessons from owning Sanexo’s Magento platform, where a Hyvä storefront and Luma checkout shared one customer journey.
Commerce Engineering 3 min read
A workflow can fail after its external action succeeds. Design retries around business operations, checkpoints, and explicit failure states.
Backend Architecture 3 min read