Writing

On AI in real delivery: what to measure, what it costs, and which claims hold up once something ships.

6 posts · 2 categories
AI5
28 Oct 2026The harness is your attack surfaceOpenHands and Strands give you sandboxes, hooks and budgets. The boundary and the policy are still yours to set before the agent runs.24 Oct 2026Amazon Blocks the Agents, Walmart Opens the Gate: Two Bets on Who Owns the ShopperAmazon blocks shopping agents while Walmart invites them in. What each retailer protects, and what agent builders should design for.agentsaiapi23 Oct 2026Your AI agent is acting under a human's name: why agent identity is the first accountability gapBefore anyone can be held responsible for what an agent did, your logs must say which agent acted, for whom, and with which credential.agentsaicybersecuritysecurity20 Aug 2026We Measure AI Adoption and Productivity. Are We Measuring AI Dependency?One half of the ledger has frameworks, vendors and a live argument. The other half has neither a definition nor a number.measurementreliabilityengineeringleadership17 Aug 2026Before You Claim AI Made Your Team Faster, Measure ItEleven questions clients ask about AI adoption, the evidence whodunit can use, and the one productivity number I do not think we should fake.aiproductivitydevopsopensource
Tools1
20 Jul 2026Relayent: use your AI subscription in your app — during dev cycleBuilding a product feature means metered API tokens, even while you're just developing. Relayent lets dev/QA lean on the AI subscription you already pay for. A dev-cycle tool, not a production one.aishowdevproductivityopensource