Week of July 3–10, 2026: MCP's next spec locks down its client-auth and skill-distribution pieces before a July 28 launch, WebMCP gains a cross-vendor conformance suite, the WebAssembly Component...
Articles by Todd Schiller
Subscribe to keep up to date with my latest posts. Shorter pieces are also collected on the Notes page.
-
Note This Week in Extensibility: MCP's spec locks down its auth and skills, WebMCP gets conformance tests, WebAssembly components bind to real-world interfaces
-
Note Loop Engineering: always be compounding
Loop engineering for AI agents should raise the floor on each iteration by creating tools and correcting instructions.
-
Note This Week in Extensibility: MCP trims to a stateless core, extensions weigh softer permission prompts, ChatGPT adds plugin governance
Week of June 26 – July 3, 2026: MCP ships beta SDKs for a leaner stateless spec, the WebExtensions group takes up softer permission prompts, and ChatGPT Business adds a plugin control plane for admins.
-
Note This Week in Extensibility: hyperscale sandboxes, extensions get a standards track, and prompt-to-plugin
Week of June 19–26, 2026: AWS makes per-user code isolation a managed product, the WebExtensions API gets a chartered W3C Working Group, and Figma turns plugin authoring into a writing task.
-
Note Beliefs hitchhike like genes
A new explorer that clusters public thinkers by the genome of their beliefs, inspired by genetic hitchhiking.
-
Note agent-browser-shield is live on Product Hunt
agent-browser-shield, a source-available defense layer for AI agents in the browser, is on Product Hunt today.
-
Note agent-browser-shield June 3 update: 14 new rules and the Chrome Web Store listing
14 new rules shipped in agent-browser-shield, now installable from the Chrome Web Store, and using it for daily driving.
-
Note Introducing agent-browser-shield (alpha): keeping AI agents safe in the browser
Announcing the alpha of agent-browser-shield — a source-available defense layer (also available as an OpenClaw skill) that blocks prompt injection and dark patterns before they reach your agent.
-
Note Letting OpenClaw loose on Boston's open data
At the Boston OpenClaw 2026 hackathon, I let an agent autonomously connect to the city's open-data MCP server, devise its own corruption-signal queries on contract data, and package the workflow...
-
Note Did Google sneak a local LLM model into Chrome?
A response to FUD around Chrome's new Local AI models, plus a demo using PixieBrix + Local AI to filter my LinkedIn feed.