- Uncovering AI
- Posts
- ⚖️ Meta algorithm targeted employees on leave for layoffs
⚖️ Meta algorithm targeted employees on leave for layoffs
A US state you didn't expect just froze AI infrastructure cold.
In partnership with
My fellow AI explorers
Two stories this week that feel like they're from different planets, except they're on the same planet. One is about an algorithm deciding who keeps their job. The other is about a governor deciding that an algorithm can't have a building.
Both are about the same thing underneath: AI is no longer abstract. It's showing up in your HR file and your backyard.
In today’s edition:
⚖️ Meta employees say an algorithm picked them for layoffs, and picked them badly
🚫 New York just became the first state to freeze hyperscale AI data centers
🎤 How to actually make an AI coding diss track (yes, really)
📰 Suno's scraping habits, Anthropic's state-by-state playbook, and more
You've seen the AI demos. Viktor does it without you watching.
The AI tool you tried last quarter waited for a prompt, hallucinated a number, then asked if you'd like a summary.
Viktor opened a PR at 2am, rebased it against main, ran your test suite, and posted a note in #eng: "Two flaky tests in payments service, both pre-existing. Recommended merging after fixing them." Then drafted the customer reply for the support ticket the bug created.
That's 619K autonomous actions per day across 20,000+ teams. Not chat replies. Real work shipped to GitHub, Stripe, Linear, Notion, and 3,000+ other tools, from inside Slack and Microsoft Teams.
You don't supervise him any more than you supervise a senior engineer.
SOC 2 certified. Your data never trains models.
"It's what you probably originally thought AI was going to be when you first heard of it in sci-fi movies." Tyler, CEO.
AI Firing
Meta's Layoff Algorithm Is Now a Lawsuit
Twenty-six former Meta employees filed a suit this week, and the allegation is blunt: Meta used AI tools and keystroke and activity-monitoring data to pinpoint who to lay off. This isn't a vague "AI was involved somewhere" claim. It's specific.
Here's what the lawsuit alleges:
AI scoring systems ranked employees using productivity metrics and AI token usage, rather than the considered judgment of managers who knew the work.
Protected leave got punished. The scoring inputs by design cannot be accumulated by an employee on protected medical or family leave, or whose output is reduced by a disability.
All 26 plaintiffs took protected leave or requested a disability accommodation before being flagged for the cuts.
The stakes are immediate. Though notified of their layoffs, all 26 remain employed by Meta, with separations set to begin July 22, which is why the suit is asking a judge to freeze the terminations while arbitration plays out.
Meta's response has been flat denial. A spokesperson told the Guardian that workforce management and organizational decisions were and are made by people, not AI. Which is a hard sentence to square with a lawsuit built entirely around AI monitoring tools and token-usage dashboards.
There's also a bigger backdrop worth naming. Meta introduced an AI employee-monitoring program earlier this year, and Mark Zuckerberg reportedly told an internal meeting that the average intelligence of the people at the company is significantly higher than the average set of people you can get to do tasks, framing the monitoring as a way to train Meta's AI systems on how smart employees work. So the same observational infrastructure that's now central to a discrimination suit was, by the CEO's own account, partly built to teach the AI what "good" looks like.
Zoom out and the timing makes it worse for Meta's optics. These layoffs sit inside a much bigger reallocation, with the company earlier saying it planned to lay off 10% of its global workforce, or nearly 8,000 people, beginning in May. The savings are funding an infrastructure arms race that makes this lawsuit look less like a one-off HR mistake and more like a preview of how every AI-heavy company will eventually manage headcount.
🔮 Prediction: This won't be the last "the algorithm did it" lawsuit. It'll be the template. Expect plaintiffs' attorneys everywhere to start subpoenaing performance-scoring systems by default, the same way they now subpoena Slack messages.
Reply and tell me: would you trust an algorithm to review your own performance if a human never touched the final call?
AI Data Center ban
While Meta was fighting a lawsuit about AI deciding who works, New York was busy deciding where AI gets to exist at all. Governor Kathy Hochul signed an executive order this week that makes New York the first state to create a moratorium on new hyperscale data centers, establishing the strongest standards for data center development in the country.
The mechanics:
Scope: The pause hits any new large-scale data center project, temporarily pausing certain state environmental permits for up to a year while New York develops a statewide regulatory framework.
Threshold: It covers facilities using 50 megawatts or more of power, for up to one year.
The why: Hochul framed it as a ratepayer issue as much as an environmental one. "As data center development threatens to hike up utility bills, deplete our natural resources, and create uncertainty for New Yorkers, it's my responsibility to take action and lead," she said.
The receipts: Residential electricity rates in New York have jumped close to 68% over the past six years, and the U.S. Department of Energy ranks New York as the 4th most expensive state for residential power.
The public backs it. A recent poll from Siena Research showed support for a one-year moratorium at 46% of New Yorkers, with just 21% opposed.
The fine print matters here too. The moratorium stays in place while the Department of Public Service develops a Generic Environmental Impact Statement that’s expected to take up to a year. During that time, the Department of Environmental Conservation will not issue discretionary permits for new data center projects unless applications have already been deemed complete. Translation: this is a pause with an exit ramp, not a permanent ban, but it's the first crack in the assumption that states will just keep waving hyperscale projects through.
And this isn't happening in a vacuum. 14 other states are reportedly considering similar moves. If New York's pause holds and doesn't get challenged into oblivion, expect a copy-paste wave.
🔮 Prediction: The "AI infrastructure is unstoppable" narrative just met its first real regulatory speed bump. Watch Texas and Virginia, the states with the heaviest data center buildout, for whether they follow New York's playbook or double down as the anti-moratorium states.
Reply and tell me: should states be able to pause AI infrastructure over electricity costs, or does that just push the buildout somewhere with fewer protections?
30-Second AI Play
How to Make an AI Coding Culture Song (Without Sounding Like a Bot Wrote It)
You've probably seen the wave of AI-generated tracks poking fun at developer life, the ones full of inside jokes about context windows, terminal tabs, and shipping questionable pull requests. Here's how to make your own in under 30 minutes using AI music tools like Suno or Udio.
Pick your inside joke first, melody second. The best AI-culture songs aren't generic "AI is cool" anthems. They're built on specific developer pain: context limits, flaky merges, prompt engineering rituals. Write down 5-6 phrases your team actually says out loud.
Draft a verse-chorus skeleton in plain text before touching the AI tool. Two verses, one hook repeated 3-4 times. Keep lines short and rhythmic, aim for 6-9 syllables per line so the AI vocal model doesn't cram or stretch words awkwardly.
Feed the tool a genre plus a mood, not just a genre. "Boom-bap hip hop, confident but self-deprecating, fast flow" gets you a far better result than just typing "rap song."
Generate 3-4 versions and Frankenstein them. Take the best hook from one generation, the best verse flow from another. Most people stop at generation one, which is the biggest reason AI music sounds generic.
Master with a light touch. Run the output through a free tool like Auphonic or LANDR for loudness leveling, then drop it straight into a lyric video template.
💡 Pro tip: If your team culture leans niche (your specific tools, your specific bugs), that specificity is your unfair advantage. Generic AI hype songs are everywhere. A song only your dev team fully gets is the one people actually forward.
Advertise to 180k engineers and CTOs choosing what tools their companies build with.
Other Relevant AI News!
🎵 A hacker's breach of Suno's source code appears to confirm what labels have alleged for years: that the AI music generator scraped YouTube, Deezer, and Genius to build its training library.
🏛 Rep. Ted Lieu says he's still unnerved by how fast AI is moving and put House Republicans on blast, arguing he can't name a single AI guardrail law passed by GOP leadership this term.
🗺 Anthropic is quietly building out a state-by-state regulatory playbook, backing bills like California's SB 53 and New York's RAISE Act rather than waiting on Washington.
🏠 A Georgia family says they were pushed into selling their childhood home to make way for a transmission line that will overwhelmingly serve AI data centers, not local homes.
💼 Anthropic and Blackstone are betting the real trillion-dollar opportunity in AI isn't the models. It's getting companies to actually use them, with their joint venture Ode now staffing "forward-deployed" engineers inside client offices.
Golden Nuggets
⚖️ Meta's layoff algorithm just became a discrimination lawsuit, and the "people, not AI" defense is getting harder to sell.
🚫 New York froze new hyperscale AI data centers for a year, and 14 other states are watching closely.
💰 The next AI gold rush isn't better models: it's getting real companies to actually use the ones that already exist.
Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)
Until our next AI rendezvous,
Anthony | Founder of Uncover AI