Yes, when you place an AI system on the EU market or its output is used in the EU. The dated timeline for a UK company selling into Europe, with what moved.
ai · regulation
Probably yes, and two pages covers it. The sections a small company's AI policy needs, a skeleton to adapt, and the agent rules we run on ourselves.
ai · regulation
Most companies owe far less AI paperwork than they fear. The five records worth keeping whatever your risk tier, who asks for each one, and when.
ai · regulation
Most SME software is not high-risk. Annex III lists eight use cases, and if yours is not one of them you can settle the question in an afternoon.
ai · regulation
Providers develop AI systems and put their own name on them. Deployers use them at work. Most UK SMEs are deployers, and the two roles owe different things.
ai · regulation
The architecture we ship for field work: local store as source of truth, an event outbox with dependency edges, foreground sync, one database per engineer.
engineering
Four failure modes: a frozen interface, a silent overwrite, a conflict found weeks later, and a request that stalls instead of failing. With the test for each.
engineering
Offline evidence survives an audit when the record keeps capture time, sync time, provenance and a change history. What an auditor asks, and what to store.
engineering · industry
How offline apps settle conflicting edits: last write wins, field-level merge on per-field timestamps, or CRDTs. What each costs, and when each is enough.
engineering