Google starts pulling the last code that lets old ad blockers run in Chrome

Information Technology 4 min 80 sources

Two Chrome updates this summer finish off Manifest V2 — the extension system that gave ad blockers like uBlock Origin their power. Plus a $22bn Fox–Roku deal, a chip-buyout wave, and automakers scrambling for licenses on cars they've sold for years.

Key takeaways

Google is removing the last code that lets full-strength ad blockers like uBlock Origin run in Chrome — and because most other browsers are built on Google's engine, they inherit the limit too.
It was a heavy dealmaking day: Fox is buying Roku for $22bn, Salesforce is buying AI-support firm Fin for $3.6bn, and Qualcomm is in talks to buy chip designer Tenstorrent for up to $10bn.
Carmakers like Ford are scrambling for U.S. licenses to keep selling models they've sold for years, after a ban on Chinese software in connected cars exposed how deep their supply chains run into China.

The browser quietly closes a door

Google confirmed that Chrome versions 150 and 151, due in late June and July, will strip out the last working code that lets old-style ad blockers run [74]. The casualty is Manifest V2 — the older of two rulebooks that govern what browser extensions are allowed to do. The full-strength version of uBlock Origin, the most-used ad blocker, was built on it.

Google began phasing out Manifest V2 in 2024, pushing extensions onto Manifest V3, which limits how deeply an extension can inspect and block the network requests a page makes [74]. Most users either switched to the cut-down “Lite” version of their blocker or left for another browser. This summer’s updates remove the final references to the old system, so from Chrome 151 onward only V3 extensions work.

A Google developer, Devlin Cronin, said on the public code change that the old system carried security risks and “tech debt” — old code that’s costly to keep alive — and pointed out that other browsers can keep supporting it if they want [74]. That last line matters. Chrome runs on roughly two-thirds of the world’s browsers, and most rivals — Edge, Brave, Opera — are built on Chromium, Google’s open-source engine. When the engine’s owner narrows what’s possible, almost everyone downstream inherits the narrower world.

A buying spree across tech

Monday was unusually busy for dealmaking. Fox agreed to buy Roku for about $22 billion, at $160 a share — one of the largest media deals in years [4][15]. Roku makes the streaming boxes and smart-TV software in roughly 100 million households; Fox wants that direct line into living rooms for its sports and news [4]. It’s a bet that owning the screen beats renting space on someone else’s.

Salesforce agreed to buy Fin for about $3.6 billion [3][76]. Fin — formerly the customer-messaging company Intercom — sells an “AI agent,” software that handles support tickets on its own rather than routing them to a human [3]. Salesforce, which sells the customer-database software most big companies run their sales on, is buying its way into the agent business rather than building it.

And in chips, Qualcomm is in talks to buy Tenstorrent for $8–10 billion [1][10]. Qualcomm makes the processors in most Android phones; Tenstorrent designs AI chips and builds them on RISC-V — an open, license-free instruction set, the basic vocabulary a chip understands, that competes with the proprietary Arm designs Qualcomm currently licenses [1]. Buying Tenstorrent would be a step toward owning that foundation rather than renting it.

The thread across all three: companies paying billions to stop depending on a platform someone else controls.

Cars built years ago, now needing permission to sell

A quieter story shows the same dependency from the other side. Ford and other carmakers are scrambling for U.S. government licenses to keep selling models they’ve shipped for years, after a rule banning most Chinese software in connected cars [40]. Ford’s China-built Lincoln Nautilus needs authorization because, although its software is written in the U.S., it’s loaded onto the car in China [40].

The rule — adopted under Biden in January 2025, kept by Trump — bans Chinese-developed software in cars from model year 2027, with tougher hardware limits arriving in 2030 [40]. Reuters notes the scramble is “exposing the extent to which the U.S. auto industry has intertwined its supply chains with China” [40]. Decisions made over a decade of cheap, deep integration can’t be undone in a season. The 2030 hardware ban is expected to be far harder to comply with than the software one [40].

The CrowdStrike fallout, finally closed

The U.S. Transportation Department closed its investigation into Delta Air Lines’ July 2024 meltdown without any penalty [30]. That meltdown was triggered by the global CrowdStrike outage — a faulty security-software update that crashed millions of Windows machines worldwide and grounded flights for days [30]. Delta’s recovery was far slower than rivals’, disrupting 1.3 million passengers and costing the airline $500 million [30].

The regulator said Delta gave passengers prompt refunds and adequate help, so it dropped the probe — part of a broader Trump-administration rollback of aviation penalties [30]. The episode remains the clearest recent reminder that a single vendor’s bad update, pushed to machines everywhere at once, is now a systemic risk, not a local one.

Consumer hardware: thinner, longer-lived

On the device side, Honor’s Magic V6 foldable claims three firsts — thinnest, biggest battery, best water resistance — though the reviewer’s verdict was that only the bigger, two-day battery actually matters; foldables have matured to where new ones feel only fractionally better [47]. Samsung added a Snapdragon-powered model to its Galaxy Book 6 laptop line at $2,100, swapping Intel’s chip for a Qualcomm one [44].

And Apple says iOS 27 will make older iPhones, back to the 2019 iPhone 11, noticeably faster — app launches up to 30% quicker — partly by bringing newer CPU-scheduling tricks (how the phone decides which task to run first) down to old hardware [32]. Keeping aging phones useful is partly goodwill and partly self-interest: a phone that still feels fast is a customer who stays inside Apple’s world.

02 · Lesson · why it matters

The track was laid years ago, and you're still riding it

A choice that made sense when nobody was watching can quietly close every door that comes after it — and by the time you'd want a different door, you no longer get to pick.

A small line of code, a large amount of power

Google is deleting the last bit of code that lets the full version of uBlock Origin — the ad blocker — run in Chrome. From this summer, only the weaker kind works.

On its own, this is a footnote. Ad blockers, browser updates, a system called Manifest V2 most people have never heard of. But look at why Google can do it at all, and you find a pattern that runs under far more than browsers.

Nobody voted for this. No regulator approved it. One company changed one rulebook, and roughly two-thirds of the web’s browsers changed with it. The interesting question isn’t whether the decision is good or bad. It’s: how did one company end up holding that lever?

The decision that mattered was made long ago

The answer isn’t this summer’s update. It’s a chain of choices stretching back fifteen years.

Chrome launched in 2008, was fast and clean, and won. Then Google did something quieter and more important: it open-sourced the engine underneath, called Chromium, and gave it away. Other companies took it. Microsoft rebuilt Edge on it. Brave, Opera, and most newer browsers are Chromium too.

Each of those was a sensible call at the time. Why build a browser engine from scratch — a job that takes hundreds of engineers years — when Google maintains a good one for free? So they didn’t. They built on the track Google had laid.

And that is the whole mechanism. Every reasonable, money-saving, get-to-market-faster decision to use Chromium also handed Google the steering. The companies weren’t tricked. They each made the locally smart choice. The smart choices added up to a world where one firm decides what a browser extension is allowed to do, and almost everyone else inherits the answer.

Why you can’t just walk back

This is what economists call path dependence: where you can go next is shaped by the road you’re already on, not by what would be best if you were starting fresh.

The tell is that going back is far harder than the original choice was. Switching browsers sounds easy — until you count your saved passwords, your extensions, the sites built and tested only against Chrome, the muscle memory. A carmaker can’t un-pick a decade of building its software in China just because a rule changed; Ford is now begging for a license to keep selling a car it has sold for years. The cost of the first step was low. The cost of the step backward is high. That gap is the trap, and it grows with time.

You see the same shape everywhere once you know it. A startup picks a cloud provider in a weekend because it’s quick; three years later, moving off it would mean rewriting half the product. A country builds its industry around one supplier’s parts; the supplier’s price is now its problem. The track laid down when the stakes felt small carries you into a place where the stakes are large and the wheel is in someone else’s hands.

The people on the track who never chose it

Here’s the part that’s easy to miss. The person most affected by Google’s decision didn’t make any of the choices that led to it.

You didn’t pick Chromium to be the world’s engine. You probably didn’t choose Manifest V2 versus V3 — most people have never heard the words. But if you run an ad blocker, the ads you see this summer change because of a rulebook edit you had no say in. The independent developer who spent years building uBlock Origin didn’t choose it either; they built for a platform, and the platform moved. The small browser company running on Chromium to save money now finds the money-saving choice also bought them a master.

This is the quiet thing about path dependence: the track gets laid by people optimizing for their own near-term reasons — cheaper, faster, easier — and then it carries a much wider group of people who were never in the room. The reader scrolling a page, the developer, the rival browser, the carmaker’s customer. None of them set the direction. All of them ride it.

Seeing the whole of it

So the next time a platform changes a rule and the change seems to come from nowhere, the useful instinct isn’t outrage at the rule. It’s curiosity about the track: when was it laid, who laid it, what early convenience let it harden into something no one can easily leave?

Almost every “they just changed it on us” moment is really the late stage of a decision made years earlier, when it looked free. The hard part isn’t spotting the lock once you’re in it — by then it’s obvious. It’s that the choices which lock you in rarely feel like choices at all. They feel like the sensible, cheap, obvious thing to do right now.

We are all somewhere on a track someone laid before us, steering less than we think. Knowing that doesn’t get you off it. But it makes you a little more careful about the small, free-looking choices in front of you today — because those are the only ones you actually get to make.

03 · Lab · your turn

The Track You Lay

Make four cheap, sensible build choices and feel each one quietly raise the cost of ever leaving — rehearsing how lock-in is built from small, free-looking steps.

04 · Hope · carry this

Every track that locks us in was laid by people, which means people can lay a kinder one — and the open standards quietly running the web are proof that some of our most powerful tools are the ones we chose to hand around rather than hold.