AIpocalypse.Now
Today'sdoom3.9
Doom 6/10 · 8 stories

IBM Cracks, Meta Sued, Secure Boot Rots for a Decade

Enterprise budgets flee to AI infrastructure, a disability-targeted layoff suit lands at Meta, and a ten-year Secure Boot hole was hiding in plain sight.

Published · By · Story-level doom average 4.8/10

The Reallocation Is Not Gentle

IBM lost roughly a quarter of its value in a single move, and the reason is unglamorous. Enterprise buyers are not cutting IT budgets, they are redirecting them, and the destination is AI infrastructure that IBM does not credibly sell. The Financial Times framed it twice in one day because the story deserves the repetition. Markets have decided that consulting revenue and mainframe maintenance are no longer where the growth lives, and they punished a company that spent a decade telling everyone it was an AI leader while quietly being something else.

The lesson generalizes. Every incumbent currently pitching a hybrid story is being repriced against pure-play infrastructure and model providers. There is no polite way to reallocate a hundred billion dollars of enterprise spend, and IBM is what the impolite version looks like on a stock chart. Expect more of these single-day repricings as CFOs finish their 2027 planning cycles and discover that the AI line item has quietly eaten the rest of the budget.

Automated Firing, Manual Denials

A new lawsuit alleges Meta used automated systems to help decide which workers to cut, and that those systems disproportionately selected employees with disabilities. Meta denies it. The denial is doing a lot of work, because if any part of the claim survives discovery, it becomes the template case for every algorithmic layoff to come.

The uncomfortable part is that nobody at any large company can currently produce a clean audit trail for how ranking, stack-ranking, and performance-model outputs feed into termination decisions. The tools exist, they are used, and their outputs are laundered through a human manager who signs the form. The legal question is whether that laundering counts. If it does not, a lot of 2024 and 2025 layoffs are going to look retroactively actionable.

Meanwhile the UK is proposing that teenagers voluntarily stop using social media after midnight, which is the policy equivalent of asking a river to consider not flowing downhill. Critics called it insufficient. Critics are correct.

Ten Years of Unlocked Doors

The security story of the slot is Microsoft's Secure Boot shim problem, which Ars Technica walked through in detail. Unrevoked bootloader shims allowed Secure Boot bypasses for roughly ten years, and the revocation only happened after external pressure made it embarrassing not to. Secure Boot is the feature that is supposed to guarantee your machine is running the operating system you think it is running. For a decade, on a large installed base, that guarantee was a suggestion.

There is no clean way to spin this. The mitigation exists, the revocations are rolling out, and any competent attacker with persistence goals already knew. The interesting question is how many other trust-root primitives have similar decade-long tails of neglected revocation, and the honest answer is that nobody has looked hard enough to know.

The Bank of England's Andrew Bailey used the week to point out that the US cannot manage AI risk alone, which is diplomatically phrased and substantively correct. Governance of frontier systems will either be coordinated or it will be nominal, and current signals point at nominal. A voluntary teen curfew and a Governor's speech are not a regime.

Ambient Chat, Ambient Anxiety

On the product side, OpenAI is reportedly building a screenless smart speaker with a camera, aimed at conversational rather than visual interaction. This is Alexa reimagined by a company that actually has something to say back. Whether households want a camera-equipped conversationalist on the counter is a question the market will answer, probably in the affirmative, probably against their own long-term interest.

NPR spent time with seven teenagers on what it feels like to grow up inside this. The answers were less dramatic than either the boosters or the doomers would want. The kids treat AI as infrastructure, occasionally useful, occasionally annoying, rarely miraculous. That is probably the most honest read anyone has offered this week. The adults are the ones losing composure. The teenagers are just doing their homework with a slightly weirder pencil, and waiting to see which of us breaks first.

Sources cited in this digest
  1. Microsoft's Decade-Old Secure Boot Flaw Finally Noticed · Ars Technica · 7/10
  2. IBM Learns Markets Punish Failure With Ruthless Efficiency · Financial Times · 5/10
  3. UK Proposes Midnight Social Media Curfew for Older Teens · BBC News · 2/10
  4. OpenAI Reportedly Planning Screenless ChatGPT Smart Speaker · The Verge · 4/10
  5. IBM Stock Collapses as Enterprises Stampede to AI Infrastructure · Financial Times · 6/10
  6. Bank of England Warns US Cannot Solve AI Alone · The Guardian · 5/10
  7. Lawsuit Alleges Meta Used AI for Disability-Targeted Layoffs · Jon Brodkin · 7/10
  8. Teenagers Assess Growing Up in the Age of AI · NPR · 2/10
Today's doomWeekly column