▶ YouTube Summaries

• https://www.youtube.com/watch?v=7QKcphdM5Cw
• Original title: AI Isn't Replacing Developers

Forrest dissects the 2026 wave of "AI layoffs" (Block, Snap, Cloudflare, Oracle, Atlassian, Coinbase, Wix and others) and argues that in most cases AI is a convenient, unfalsifiable scapegoat rather than the real cause. Reading filings, shareholder letters and activist-investor pressure campaigns, he traces the cuts back to pandemic over-hiring, margin defence, and budget reallocation into AI infrastructure — while noting that AI-driven efficiency gains are real and do reduce hiring at the margins.

The pattern

Companies are attributing mass layoffs to AI: Block cut 4,000 (~40% of its workforce, the largest percentage layoff in S&P 500 history), Snap 1,000, Cloudflare 1,100, Atlassian 1,600, Oracle 21,000 over 12 months. All frame the cuts not as weakness but as AI-driven efficiency. Forrest's core observation: this claim is unfalsifiable — nobody audits how AI supposedly took over the work, and some companies even claim AI will do the work in the future, which requires proving nothing at all.

Case study: Block

Jack Dorsey's letter says the business is strong (gross profit up 15% to $10B+, raised guidance, $5B buyback) and cuts are due to AI-enabled "smaller, flatter teams". Omitted context: Block nearly tripled headcount during the pandemic (4,000 to 12,000) on near-zero interest rates, then the stock fell ~70% and rates rose. It avoided the big 2023 correction most peers did, then finally cut 4,000 in early 2026 — with AI as the stated reason. Analysts (WSJ, FT Partners via Bloomberg) didn't buy it either. The stock jumped 25% in an hour on the announcement, adding ~$6B of market cap.

Case study: Snap

Two weeks before Snap's 1,000-person "AI efficiency" layoff, activist investor Irenic Capital took a 2.5% stake and published a letter (savesnapnow.com) explicitly saying Snap over-hired (3,000 pre-COVID to 5,200), hadn't course-corrected like Meta, and pointing to Block's "AI-driven" 40% cut as the template. Snap complied two weeks later and labelled it AI. Either Snap replaced 16% of its workforce with AI in a fortnight, or it did investor-demanded cuts under the best-sounding banner available.

Case study: Cloudflare — the strongest counter-argument, still shaky

CEO Matthew Prince wrote a WSJ op-ed, "How I Choose Which Cloudflare Employees to Replace With AI", splitting staff into builders / sellers / measurers (Drucker, 1954) and laying off mostly "measurers" — finance, legal, audit, compliance, middle management, i.e. the historically most common layoff target anyway. The announcement landed on earnings day: record revenue ($640M, +34%) but a GAAP operating loss of $62M and shrinking margins (~130bps YoY). Stock dropped 20%. Forrest reads it as classic margin defence and bloat-cutting (plus hiring more engineers), with AI as wrapper and earnings-day timing as a hedge.

The real reasons

1. Over-hiring during the pandemic — the honest reason given in 2023 (e.g. Zuckerberg), which mysteriously vanished from 2026 announcements once AI became a stock-positive excuse. In January only 7% of cuts cited AI; by May, 40%.
2. Budget reallocation to AI infrastructure — the verifiable "AI layoffs" are cuts that fund AI capex: Meta cut 8,000 while moving 7,000 into AI roles and raising capex to $125–145B; Oracle needed $8–10B in cash for a $50B data-center build-out. AI claims the budget, not the jobs.
3. Investor optics — a December 2025 survey of 1,000 US hiring managers: 59% admitted companies lean on AI as the layoff reason because it "plays better" with investors and press; only 9% said AI actually fully replaced the roles.

Even AI-industry CEOs call this out: Jensen Huang calls the AI-layoff narrative "lazy" and irresponsible; Demis Hassabis says AI is an excuse for layoffs companies wanted anyway; Sam Altman calls it "AI washing". Though Forrest notes they're doing damage control — public fear of AI job loss invites protests and regulation (as seen with Anthropic and Mythos).

But AI is real

Forrest doesn't claim AI does nothing. Since Opus 4.5 and Codex 5.3 (six-seven months prior), AI genuinely makes developers more efficient when used properly — agents completing tasks per your spec and standards, with human review. His friend's dev shop of 4–5 developers now does work that would have needed 15–20 a few years ago: no layoffs there, but fewer jobs created. The logical gap he keeps hammering: if AI really multiplies output and your numbers are great, why shrink instead of growing output with the same headcount?

Takeaways

• Three buckets: outright scapegoating (layoffs that would have happened anyway), overstating (AI a secondary factor sold as the whole reason), and misleading headlines (cuts funding AI infrastructure, not AI doing jobs).
• Nobody is demonstrably having AI "come in and do the jobs" as the announcements imply.
• Warning to CEOs: blaming AI for layoffs makes you a case study — every subsequent outage and bug gets amplified as evidence against you.
• Watch for the rehiring pattern seen in 2024–25 (Klarna; Atlassian laid off 1,600 "AI-first" right after hiring 2,000 and kept hiring).