I’ll begin this rant with a tiny bit of history: look at the HTTP(S) headers a modern Web browser sends with its requests, and you’ll notice that for historical reasons it refers to itself by an odd name: not a browser, not a program, not an App, but a user-agent1.
While the term may have sounded a little obtuse 30-odd years ago, here in 2025 it sounds downright visionary. The World Wide Web Consortium elaborates on this terminology in its Privacy Principles:
A user agent acts as an intermediary between a person (its user) and the web. User agents implement, to the extent possible, the principles that collective governance establishes in favour of individuals. They seek to prevent the creation of asymmetries of information, and serve their user by providing them with automation to rectify automation asymmetries. Where possible, they protect their user from receiving intrusive messages.
The user agent is expected to align fully with the person using it and to operate exclusively in that person‘s interest. It is not the first party. The user agent serves the person as a trustworthy agent: it always puts that person‘s interest first.
In an era of neverending monetization and enshittification, where the web browser with the lion’s share of marketshare is owned by an advertising megaplatform and bespoke content-viewing mobile apps exclusively serve the interests of their platform owners, this bears repeating. In the original vision of the Web, something was expected to actively advocate for the user in competition with the wild unknown, and the user’s choice of access software was intended to fulfill that role.
How does the coming AIPocalypse fit into this?
Cory Doctorow has an interesting and probably-not-even-wrong take on the current AI frenzy being a money-burning bubble the likes of which we have never seen, which will inevitably implode as the Infinite Improfitability Drive propelling it ever forward summons a space whale over Mars and crashes the Hubble constant our 401Ks.
These companies are not profitable. They can’t be profitable. They keep the lights on by soaking up hundreds of billions of dollars in other people’s money and then lighting it on fire.
But I disagree: where there is a will (and a paying customer), there is a way. Here is where the money trail leads, and mark my words.
First, the days of free-as-in-beer AI are numbered. The cute conversational summary that appears at the top of every Web search (without asking), in your word processor (without asking), in your photo editor, email client, operating system, corporate Wiki/Teams/Confluence/Git-Host (without asking), is just the loss-leader honeymoon period talking. Part the usual get-big-fast play of outsprawling competitors for market share, part “first taste’s free”. You, or more likely your boss, will pay handsomely for search that actually works if the organic results degrade into slop badly enough, let alone for vibecoding tools that can justify sacking that smarmy Senior Architect along with an entire expensive QA department.
Second, personal AI-powered user-agents will evolve as a necessity that AI bros are all to happy to put in your corner of the ring – for a monthly fee, of course. Remember that bit about automation asymmetries? Just as Douglas Adams’ electric monk famously offloaded the onerous task of believing all the things the world expected you to believe, your user-agent will outsource jumping through all the technical and emotional labor hoops late-stage capitalism expects you to profitably jump through.
Imagine it, a personal assistant to automatically click the first 30 Google search results, strip all the low-effort AI slop and redundant singsong SEO-speak, deduplicate and summarize the rest. Filter away shopping results when you’re not shopping. Watch that YouTube video with 8 minutes of content padded to 30 minutes to please a ranking algorithm and summarize it back down to 8 minutes. Sit on hold and fight its way through AI-powered phone trees for you until it reaches the “talk to a human” finish line.
As the adblocking wars continue to escalate, your AI user-agent will watch the ads and take quizzes on them for you, extracting the actual content for perusal at your leisure2. Read EULAs, opt out of things, sidestep dark patterns, solve annoying CAPTCHAS with an ironic sneer3, vet callers from unknown numbers, cancel that useless subscription that can only be done via 1-800# during business hours and you can never find the time (as designed). Create an entire parallel digital fingerprint that isn’t your own. Be the product for you. Wade through an eternity of shit, so you don’t have to.
How much would you pay for a de-enshittified experience across the entire technological attack surface of your life?
Hmm. How much would the players who make their billions on said enshittification pay to not let this happen?
Ding ding ding, let the next zero-sum arms race begin. Trillions of your favorite monetary unit and untold tons of dead dinosaurs blowing up in smoke just for the world to run in place.
Welcome to the future.
- This has been dutifully carried forward from the earliest drafts of HTTP. While the specific origins of the term are not elaborated on in documents of the era, it dates back to at least 1992. ↩︎
- What, it has to physically play the content through the DRM pipeline and transcode it back via screen-scraping at gawdawful quality? Built-in AI upscaling to the rescue! ↩︎
- CAPTCHAs will inevitably devolve into reaction time tests, if they don’t straight bust out a mandatory retinal scanner. ↩︎
Leave a Reply