AI-generated code needs verification before it reaches a PR. Erica Hughberg and @Dpbrinkm discuss QA sub-agents, workflow-based tests, and using screenshots to catch broken behavior before human review.

AI coding ROI isn't measured in lines of code. Pepijn Senders shares how to track token spend against product value and why accepting code you don't understand can create risks in production.

AI agents can test thousands of attack paths in parallel. A small network vulnerability in a supposedly isolated sandbox can become an escape route. Erica Hughberg and @Dpbrinkm discuss agent security and bots with buying power.

Browser agents scraping DOMs and screenshots are slow, token-intensive, and unpredictable. Sarah Drasner of @Google explains how WebMCP exposes defined actions, flags consequential operations, and helps developers guide agents safely.

Can a narrowly scoped agent outperform a general-purpose setup? Justin Schroeder, co-founder of Standard Agents, explains how prompts, tools, models, and runtimes shape agent effectiveness - and why standards matter for sharing reliable agents.

What if an MMORPG launched already full of agent players instead of scripted NPCs? Issac Lee of NEXUS sees it as a way past a new game’s chicken-and-egg player problem.

A prompt bloat flag at C.H. Robinson traced back to a code error that kept a service in a retry loop, using tokens again on every pass. Jason Ward on tracking AI spend with observability metrics.

James Barney (MetLife) on tying token spend to cost per business outcome, and why a cheaper model you have to run twice can double the bill.

AWS has ~16,000 APIs, so its MCP server runs in code mode. @JamesWard (@AWS) on why agent-written Python is hard to review, and the restricted, schema-validated DSL he's testing instead.

A server that lists its skills and serves them as resources, so tools ship with instructions for using them. Ola Hungerford of @Nordstrom on Skills over MCP, dependency gaps, and the context bloat that returns at hundreds of skills.