The Squeeze: 5 Rules for Surviving the AI Compression of Tech Jobs

78,000 tech workers were laid off last quarter. Half of those cuts were AI-driven. The org chart is compressing: Block cut 40%, Meta restructured into AI-native pods with 1:50 manager-to-builder ratios, Shopify requires managers to prove AI can't do a job before hiring. Here are five rules for surviving what Pete (a NYC software agency owner working with startups and Fortune 500s) calls The Squeeze.

The Squeeze — What's Happening

The org chart is flattening because AI enables fewer people to produce more:

Company	What they did
Block	Cut 40% of workforce, went all-in on AI
Meta	Restructured 1,000 people into AI-native pods — each builder owns features end-to-end, one manager per 50+ builders
Shopify	Managers must prove AI can't do a job before they're allowed to hire for it

The pattern: companies are reducing layers, expanding individual scope, and replacing specialist roles with AI-augmented generalists.

The Five Rules

1. Generalists Win

Corporate America spent a century telling people to specialize. That era is ending. The builders thriving now own the whole process — idea to shipped product:

Talking to users one minute
Prompting Claude Code the next
Automating testing after that

Your agents handle the deep specialist work. Your job is directing them.

2. You're Now a Multitasker

The single-threaded deep-focus workday is over. You're juggling:

Multiple agents running in parallel
Multiple work streams
Multiple roles simultaneously

Jensen Huang expects every NVIDIA engineer to get a token budget on top of their salary so they can amplify themselves 10× with agents. Your value is measured by how well you deploy your agents, not how well you do one task.

3. Surrender the Details

Your job is output — high-velocity shipping. To do that:

Ditch the ego about hand-crafting every line
Treat work as a flow state exercise: directing, reviewing, shipping
Don't compromise quality — achieve higher quality in a fraction of the time
If you nitpick every line while someone else ships 10× faster at the same quality, you lose

4. Develop Taste

More autonomy means more decisions made by individuals, not committees. No more waiting for product specs or final designs — you're making those calls in real time.

Paul Graham's framing: "When anyone can make anything, the differentiator is what you choose to make."

The prediction: CTO gets redefined as Chief Taste Officer — a Rick Rubin-like figure receiving product decisions from machines and deploying taste at scale. Yes/no, yes/no, yes/no.

5. Look Hard at Where You Work

If your company is barely investing in AI, you're a sitting duck. An MIT study found ~95% of AI pilots never make it past proof of concept. If your company can't figure AI out, a competitor will — and the gap compounds exponentially.

People with AI skills are already earning more. Find companies actually embracing this.

Why "Pivot to Something Human" Is a Losing Strategy

The instinct when AI threatens your role: pivot to work that's "more human" — creative, empathetic, strategic. Pete's counter:

"That buys you some time, but it's a losing strategy. AI will inevitably catch up."

The sustainable strategy isn't to run from AI. It's to operate differently — become the person who deploys AI at scale, not the person who does the work AI replaces.

Key Takeaways

78K tech layoffs last quarter, half AI-driven — the org chart is compressing
Generalists win — own the whole process, let agents handle specialization
Multitask with agents — your value = how well you deploy your token budget
Surrender the details — ship fast at high quality, don't nitpick
Develop taste — when anyone can build anything, what you choose to build is the differentiator
Pick your company carefully — 95% of AI pilots fail; be at a company that ships

Resources

The Squeeze (Substack) — Pete's original article and follow-up "Surviving The Squeeze"
Paul Graham on Taste — Source of the "what you choose to make" framing
Shopify AI Hiring Policy — Tobi Lutke's directive on proving AI can't do the job