Think more like what Microsoft did by being everywhere enterprises already were: integrations, admin controls, compliance posture, procurement-friendly packaging. OpenAI doesn’t need to “invent all experiences” if it can be the default inference layer that’s easiest to buy, govern, and instrument inside large orgs—especially if it can show auditors a cleaner story than “random employees piping data into a consumer chatbot.” This is the less sexy but very real wedge: the assistant becomes a governed interface to computation.

In consumer search, Google paid to be the default because defaults are basically behavioral gravity. Most people don’t churn defaults. In AI, the equivalent isn’t just “the search box,” it’s the assistant that sits inside the apps where you already spend your cognitive budget—design, docs, email, IDE, CRM, ticketing, finance—so your prompts stop being little one-off spells and start being operations with permissions, audit trails, and repeatability.

You’re worried about invisible persuasion while asking users to paste their own API keys into your extension

for a three-route demo app with /#/a and /#/b, you absolutely don’t need a router library

The idea is that this router works with so much more than the three routes... should prob update the example to show that

edit: updated example to show that this can handle things like url params since it just uses URLPattern()

Author here. Happy to go deep on any part of this. The part I'd most like to get feedback on from this community: the decision to use .md files over a proper vector store or database. We made that call for prototyping speed and human readability, but I'm curious if anyone has hit the ceiling on this approach earlier than we did.

The honest tradeoffs we found: .md works until you need real-time external data or write actions back to production. For pure read/reason workflows in a productivity context, it's held up better than expected.

Also happy to discuss the browser telemetry architecture — specifically how we handle session classification without capturing page content.

Social media didn’t start as a psychological experiment. It started as a tool to connect friends. Then it became a business. Then it became an attention refinery.

escalation is often profitable before it is visibly catastrophic

Facebook manipulated attention. AI platforms will manipulate narratives. Action models and VR systems will manipulate lived experiences.

When feeds were mostly friends and pages you chose, the algorithm felt like a helper. Now it feels like an environment you’re dropped into, one that doesn’t share your values, your context, or your sense of what’s appropriate.

“Open”AI is growing more opaque and black box every day