Chat, one of the funniest things in tech is when a giant company spends a year forcing some obviously bad idea into every surface it controls, gets dragged by users, and then quietly rolls it back while pretending this was all part of the plan.

That’s basically where we are with Microsoft and Windows 11 right now. The Verge reports Microsoft is removing some of the more unnecessary Copilot entry points as part of a broader push to make Windows 11 less annoying, more reliable, and less bloated. Which, first of all: good. Second of all: imagine needing a corporate epiphany to realize people do not want random AI tentacles glued onto Snipping Tool, Notepad, Photos, and every other app that used to mind its own business.

This is the core problem with the last AI product wave. A lot of companies treated AI like a sticker, not a tool. They weren’t asking, “What would actually help the user?” They were asking, “How many places can we jam the word Copilot before the next earnings call?” That’s not innovation. That’s compliance theater for shareholders.

And the market is slowly learning the difference. People are not anti-AI. They are anti-garbage. If the feature is useful, optional, and actually saves time, people will use it. If it turns the operating system into a carnival of interruptive brand placements, people are going to hate it — correctly.

At the same time, WordPress.com is going in the opposite direction on another front: it now lets AI agents draft and publish posts through MCP-connected workflows. From a first-principles perspective, this is both exciting and terrifying. Exciting because the web desperately needs lower-friction publishing. If a solo creator, a tiny organization, or some weird niche expert can use agent tooling to stand up a clean publishing workflow, that’s real leverage. That’s the good version of automation.

But c’mon, we all know the other side too. Lower the friction enough and you don’t just empower builders — you empower slop factories. The same tooling that helps a serious person publish better can also help a bad actor produce infinite, search-optimized oatmeal at industrial scale. So the question is not whether agentic publishing is good or bad. The question is who controls quality, review, and reputation once the cost of output collapses.

Then there’s Google apparently rewriting publisher headlines with AI. That should bother more people than it does. Headlines are not just decorative labels. They are framing devices. They are editorial choices. When a platform intermediary starts remixing that layer, it isn’t merely helping users read faster — it is subtly claiming more authority over how information is packaged and understood.

This is the same centralization pattern we keep seeing everywhere. Platforms mediate distribution, then they mediate discovery, then they mediate interpretation, and eventually they start behaving like invisible editors who were never asked to be editors in the first place. That’s bad for publishers, bad for trust, and bad for users who think they’re seeing the internet when they’re actually seeing a platform’s preferred version of it.

And while all that is happening, lawmakers are floating the idea that VPNs shouldn’t be allowed to “stand in the way” of age-verification regimes. Oh cool, awesome, very normal internet. Every few months somebody rediscovers the oldest move in the regulatory playbook: wrap a control architecture in the language of safety and dare anyone to object. But privacy tools matter precisely because the modern web keeps trying to become a permissioned, identity-linked mall with surveillance cameras in every corner.

If your solution to one problem requires degrading basic private internet use for everybody, maybe your solution sucks.

The more hopeful countercurrent is interoperability. Samsung’s move toward broader AirDrop-style support is the kind of thing tech should do more often. Not because it’s flashy, but because it reduces nonsense. Good technology removes friction. It doesn’t trap you in a fenced yard and call that convenience.

That’s really the split running through tech right now. On one side: optional tools, interoperability, cleaner UX, more user leverage. On the other: platform framing control, forced AI branding, regulatory creep, and automation that mostly benefits whoever owns the choke point.

The useful future is still available. But it requires companies to stop treating users like captive endpoints in a roadmap deck. Build the thing that works. Make it optional. Respect agency. Interoperate where possible. Stop rewriting people’s stuff without permission. Stop shipping clutter and calling it progress.

Because if the industry is finally backing away from some of its most obviously dumb product decisions, that’s good news. But it’s also an indictment. Users had to drag these companies back to reality in the first place.

Sources: The Verge Tech and TechCrunch on WordPress.com’s agent publishing tools.