Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
The AI layoffs end in 12 months and I know why
TL;DR
Cloudflare became the poster child for AI-era layoffs — On an earnings call announcing a record $639 million quarter, CEO Matthew Prince also said Cloudflare would cut 1,100 people and framed it as getting “fitter,” while claiming employees had become “100 times more productive.”
The real target isn’t just customer support — it’s organizational redundancy — Mo’s key read on Prince’s “support roles” comment is that companies are cutting the bench: backup people who mainly existed to preserve institutional knowledge, because tools like Claude now serve as the fallback.
His core thesis is that AI layoffs are temporary because hidden software demand is enormous — He argues Jevons paradox won’t mean consumers want 50x more random apps, but that companies have a massive backlog of internal dashboards, automations, Slackbots, and ad hoc tools that were never worth staffing before.
AI is especially well-suited to the ugly, neglected software no one wanted to build — Internal tools have historically been low-priority “horror shows,” and he thinks AI will unlock that long tail because quality standards are lower and the economic threshold to build them just dropped.
In the near term, hiring rewards positioning more than craft — Mo says the market now values candidates who present themselves as AI-multiplied operators—think ‘10 agents’ and ‘300 pull requests’ energy—rather than people merely saying they can code both with and without AI.
His practical advice is blunt: stand out now, or wait for demand to rebound — He predicts companies are temporarily saying ‘let’s try less people,’ but once they exhaust what a smaller AI-assisted team can do, they’ll swing back to ‘we need more people,’ likely within roughly 12 months.
The Breakdown
Record revenue, mass layoffs, and the world’s weirdest gym metaphor
Mo opens on Cloudflare’s earnings call, where Matthew Prince announced a record $639 million quarter and then, almost in the same breath, 1,100 layoffs. He hammers the absurdity of celebrating the company’s best quarter ever while cutting one in five people, then mocks Prince’s line that “just because you’re fit, it doesn’t mean you can’t get fitter” as if layoffs were some kind of David Goggins-inspired workout plan.
“100x productivity” and the screwdriver insult
The tone gets sharper when Prince says Cloudflare employees are now “100 times more productive” and compares the shift to going from a manual screwdriver to an electric one. Mo lingers on how dehumanizing that sounds, joking about laid-off engineers hearing their work reduced to Black & Decker metaphors while their families listen in.
What “support roles” really means
Then he lands on the line he thinks actually matters: Prince says the people directly creating code are seeing incredible gains, while “support roles behind them” won’t drive companies going forward. Mo argues this doesn’t just mean customer service — it means the backup layer companies built over decades: extra staff holding institutional knowledge in case a key person quits, goes on vacation, or becomes a bottleneck.
Claude as the new bench player
His analogy is sports: companies used to need starters plus a bench, but now “Claude is the backup.” If a star engineer leaves, you’re not totally stranded anymore; you can spin up Claude, ask questions, and get by well enough that the insurance role starts looking expendable.
Jevons paradox, but not the dumb version
Mo brings in William Stanley Jevons, who noticed in 1865 that more efficient steam engines increased coal demand rather than reducing it. But he rejects the lazy AI version of that argument — the idea that the world simply wants “50x more software” in the form of more apps, more notifications, and more bugs — and says the real missing piece is the long tail of internal software demand.
The software companies always wanted but never funded
This is his optimistic turn: every company has an endless backlog of dashboards, automations, Slackbots, and weird internal tools that never made it into OKRs or budgets because the core product always came first. AI, he says, is perfect for this class of software: low-stakes, ugly, ad hoc, historically neglected systems that no one wanted to build by hand.
Why layoffs happen before rehiring
Mo thinks most companies still haven’t figured out how to actually use these efficiency gains, so for now they’re doing the simpler thing: shrinking the ship to make it easier to steer. That’s why he thinks the layoffs come first, but not forever — once firms max out what smaller AI-assisted teams can do, capitalism pushes them back into “we need more people” mode.
In the meantime, market yourself like an AI monster
His advice for job seekers is less about craft than perception: don’t just say you’re willing to use AI and can also code by hand. He says candidates should signal absurd-looking leverage — “10 agents,” “300 pull requests,” “1000x throughput” — because CEOs want to believe they’re hiring Boris-level AI wizards, and personal websites matter because they make you memorable in a pile where attention is now 70% of the game and skill is 30%.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
