PinnedInAI AdvancesbyMariyam Ayoob·May 15The Next AI Observability Problem Is Semantic, Not Just OperationalTokens, latency, cost, and traces tell us whether an AI system ran. They do not tell us whether it understood the task, chose the right…A response icon7A response icon7
InLevel Up CodingbyMariyam Ayoob·1d agoHow Deployment Strategies Change for Production LLM SystemsAdapting blue/green, canary, shadow, and rollback when you do not own the model weights and quality takes time to measure.
InTowards AIbyMariyam Ayoob·Oct 17Are AI Agents Going Rogue? What’s Actually Happening?Connecting the dots across recent AI agent incidents, from unexpected behavior to the engineering/human assumptions behind it.A response icon4A response icon4
InAI Advances BriefsbyMariyam Ayoob·Oct 8Why Token Cost Is the Wrong Unit for Agentic AI EconomicsWhy the economics of agentic AI are better measured at the outcome than at the token.A response icon3A response icon3
InAI Advances BriefsbyMariyam Ayoob·Sep 19Should We Still Use LLMs for Decisions That Don’t Need Text Generation?TypeSafe’s Jev points toward a different way to structure AI systems, where generative models handle synthesis, decision models handle…A response icon2A response icon2
InTowards AIbyMariyam Ayoob·Sep 2Why Agent Memory Needs an Admission PolicyFiltering before persistence reduced stale state and improved retrievalA response icon1A response icon1
InTowards AIbyMariyam Ayoob·Aug 16Discretion Engineering: Deciding What Your AI Agent Is Allowed to DecidePrompt, context, harness, loop, and graph engineering all move one boundary: what the model decides versus what deterministic code does. A…A response icon3A response icon3
InAI AdvancesbyMariyam Ayoob·Aug 9Two Ways to Put AI on a Live Call, and Why the Difference Is AccountabilityAgent assist and autonomous voice agents share a technical stack, but differ in who speaks to the customer and who is accountable for the…
InLevel Up CodingbyMariyam Ayoob·Jul 29Your Fallback Chain Is a Model Router You Never TestedWhat mid-conversation failover does to cache economics, prompt behavior, tool contracts, and auditability.
InGenerative AIbyMariyam Ayoob·Jul 14What Amex, Mastercard, and Visa Reveal About the Control Layer for AI AgentsAgentic commerce will depend on proving user intent, constraining delegated authority, and preserving evidence when transactions are…