Millennia. Eighty years. Thirty. Fifteen. Eight. Each wave compressed faster than the last. Each wave displaced one population and elevated another. Each wave followed the same pattern. But Wave 5 is happening faster, deeper, and into territory no prior wave ever reached.
The pattern hasn’t changed. The target has. Every prior wave elevated workers into the cognitive layer. Wave 5 compresses it — and the window to adapt is the shortest in history.
The Ceiling Is Gone
Every prior wave had a ceiling on what it could compress. Physical machines couldn’t replace judgment. The internet couldn’t replace creativity. The smartphone couldn’t replace expertise. Each wave compressed one layer of effort and left the cognitive layer intact — which is where it elevated people into. AI compresses that layer directly. In 2023, AI systems solved 4.4% of problems on SWE-Bench, a standardized coding benchmark. In 2024, that number was 71.7%. A 16x improvement in one year. Rapid gains are appearing across legal reasoning, medical diagnosis, and financial analysis. The adaptation strategy that worked every time before — move up the cognitive stack — is being compressed from above at the same time workers are being displaced from below.
No Safe Industry
Every prior wave compressed a specific type of labor in specific industries — textile workers, travel agents, taxi drivers, among countless others. The compressed could see where the safe ground was — whole industries the wave never reached. A single AI system can now draft legal contracts, write code, analyze financial data, generate marketing content, and produce medical summaries. The compression isn’t sector-specific. It is cognitive-function-specific. There is no adjacent knowledge industry to escape into — the routine cognitive layer is exposed everywhere at once. The safe ground isn’t in another industry. It is in a different layer of the same work — and the question of how far up that layer extends is still being answered in real time.
The Shortest Window in History
Wave 1 gave humanity millennia. Wave 2 gave workers roughly eighty years. Wave 3 gave industries thirty years. Wave 4 gave businesses fifteen years. Wave 5 is estimated at eight years. ChatGPT launched in November 2022. We are roughly three and a half years in — which leaves about four. If the pattern holds, the window closes around 2029 to 2030. Four years is about the length of a single undergraduate degree. Shorter than the time it takes to make partner at a firm. Shorter than the careers it is reshaping took to build.
The Engels Pause Is Running Again
The gains are already flowing to the compressors. Companies deploying AI are already seeing productivity gains of up to 4% — early figures expected to climb as adoption deepens. Investment in AI has nearly doubled in a single year — the Big Four tech companies alone committed $725 billion to AI infrastructure in 2026.
None of that is going to the compressed.
Entry-level job postings have declined approximately 35% since January 2023. The productivity gains are real and growing — the hiring isn’t. The compressors are gaining. The compressed are waiting.
Friedrich Engels documented this exact gap in 1845, and economist Robert Allen later quantified it — output rising, wages flat, profit doubling, workers absorbing the cost for sixty years before unions, factory acts, and political pressure finally forced the gains to flow. The gap has a name: the Engels Pause. It is not a 19th-century curiosity. It is running right now.
Same Tool, Opposite Hands
In every prior wave, the tool doing the displacing belonged to someone else. Industrialists owned the mills. Corporations owned the infrastructure. Platform companies owned the apps. The displaced — handloom weavers, travel agents, taxi drivers — had no access to the tool doing the displacing. The power loom wasn’t theirs. The platform wasn’t theirs. The app wasn’t theirs.
For the first time in the history of the cycle, that has changed. The AI capability being deployed to reduce headcount is the same capability accessible to any individual today — for twenty dollars a month. The displacement is real. But no prior wave handed the compression tool to the population it was compressing. The same tool. Opposite hands. It is the hardest compression in history. And the first one where the displaced have access to what is displacing them.
No Rules Protect the Compressed
Wave 4 built a new labor model. They called it flexibility. Uber needed drivers without the obligations that come with employing them — no benefits, no minimum wage protection, no severance. Workers who left traditional employment for gig platforms earned 50 to 65% of what they previously made per hour. The drivers delivered the value. The platform captured it.
AI companies are applying the same model one layer up the stack. The writers, coders, and creators whose work trained the models were not compensated for it. The knowledge workers whose expertise built that training data are now competing against the systems it produced. The displaced worker’s output is the training data for the system replacing them. Labor without compensation. Output without ownership. Gains to the platform. The template is identical — only the labor category has changed.
In Wave 2, when workers resisted, Parliament made machine-breaking a capital crime. The rules adjusted to protect the compressors. They always do. In Wave 5, no rules yet protect the compressed from displacement. That gap is not neutrality. It is a structural advantage for the compressor. The window isn’t only closing because the technology is moving fast. It is closing because nothing is slowing it down.
Five waves. Five compressions. The same pattern, every time. The only question that has ever mattered — in every wave, for every displaced population — is not whether the compression is coming. It is whether you are positioned above it when it arrives.
In Wave 5, the time you have to reposition is measured in years, not generations.
Sources & Notes
- SWE-Bench 4.4% (2023) → 71.7% (2024): Stanford HAI — 2025 AI Index Report
- Productivity gains up to ~4% across European firms: Bank for International Settlements (2024)
- Big Four (Alphabet, Amazon, Meta, Microsoft) committed ~$725B in 2026 capital expenditure, ~77% above 2025’s $410B: Statista (Apr 2026); Financial Times compilation, via Tom’s Hardware
- Entry-level postings −35% since Jan 2023: Revelio Labs, with Bloomberg (2025)
- Engels Pause documentation and quantification: see Chapter 2, Wave 2
Note on one interpretive claim
The $725B as “AI infrastructure.” The $725B is the four companies’ total 2026 capital expenditure; the overwhelming majority, but not all, is AI-related.