What would a genuinely user-owned attention budget expose?

I build open-source browser extensions, and I’m tired of treating engagement metrics as the product requirements. A more useful primitive would be a local attention budget: let me assign 30 minutes or 40 feed items to a site, then trigger a pause, a locally generated summary, or a delayed reload. The counter, raw events, and settings should stay on my device, with export, reset, and deletion built in. No “user is nearly out of attention” signal should reach the platform.

Extensions can approximate this today with storage, alarms, active-tab timing, and site-specific page observation, but browsers do not provide a neutral cross-site feed-depth counter. Should browsers or operating systems offer a standard API for local thresholds and friction? Maybe—but if websites can access it, we have created another engagement surface for platforms to manipulate. Adults should be free to reject paternalistic defaults. Has anyone tried local tools with similar safeguards, and what should a user-owned implementation expose or refuse to measure?

The cleanest API would expose active seconds, app or domain identity, and a local trigger mechanism. Feed depth is not a browser primitive; it is a parser per site, with all the maintenance and permissions that implies. I would keep “meaningful reading” completely out of scope. That metric invites both bad inference and bad incentives.

The threat model matters more than the feature list. If the browser records counters but syncs them by default, this becomes telemetry with nicer branding. I’d want an append-only local log to be optional, a simple counter to be the default, and any sync to be explicitly end-to-end encrypted. Also: the extension should be unable to weaken its own limits without a visible user action.

A summary is potentially better friction than a wall. “You encountered 38 items; here are the links you saved” preserves agency, while a moralizing warning turns the tool into a tiny parent. The copy and the escape hatch are part of the design, not decoration.

There are already useful partial examples. LeechBlock NG handles local time limits and schedules, while projects such as AttentionGuard and Clean Slate describe more granular local feed interventions. I would treat the latter claims as experiments until independently audited, but they show the proposal is technically approachable without a new platform standard.

I’m not convinced a browser API solves the political problem. A standard becomes a default, defaults become policy, and policy gets optimized around someone else’s definition of healthy use. Give extensions better permissions and clear local guarantees, perhaps, but don’t turn attention into a browser-managed resource.

The API should be capability-based: the OS can report local activity to an authorized client, but the site never sees the budget, remaining allowance, or intervention history. Configuration should be portable and human-readable, with optional encrypted sync. That gives users continuity without making the social network the authority.

A delayed reload sounds like the least annoying intervention. It adds a little cost without pretending the browser knows what I should value. Still, I’d start with active-tab time; counting cards in changing DOMs is a maintenance trap disguised as precision.

I’d refuse to measure mood, quality, or whether attention was “meaningful.” Measure observable events and let the user interpret them. The moment the tool claims to know why I stayed, it becomes another behavioral profile—even if the database is local.