From Muscle to Matrix

A developer writes code. Takes 5 hours at $20/hour. Cost: $100.

An LLM writes the same code. One API call. Cost: $0.01.

Same artifact. 10,000x cost difference. This is not a productivity improvement. This is a phase transition.

What Is Money?

Money is stored work you can trade later. But what counts as "work" keeps changing.

1800s: Muscle x Hours

Physical labor over time
Output: Manual work

1950s: Thinking x Hours

Cognitive labor over time
Computers amplify, humans still required

2000s: Expertise x Hours

Knowledge work over time
Time still the bottleneck

2023+: Inference x Electricity

Pure energy conversion
Human time exits the equation

Key insight: For 200 years, time was always in the equation. In 2023, it dropped out.

The Cost Collapse

Human wages (1990 to 2025):

$500 down to $300
Nearly flat decline

AI cost (2020 to 2025):

2020: $100 (human + tools)
2023: $1 (GPT-3)
2024: $0.01 (GPT-4)
2025: $0.001 (optimized)

The gap between these lines is captured surplus. It does not go to workers.

Three disruptions that got us here:

Internet (1999): Copying became free. Creation still needed humans.
Crypto (2017): Energy converted directly to money.
LLMs (2023): Creation no longer needs human time.

AI inference costs halve every 18 months. Human wages do not.

Three Production Systems

Human Era (Pre-1970s)

Input: Calories, oxygen, 8 hrs/day

Process: Serial thinking, 1 thought/sec

Speed: Hours to days

Cost: $100 (5hrs x $20)

Constraint: Human time (24 hrs max)

Value capture: Worker 70%, Company 30%

Computer Era (1970-2023)

Input: Electricity, human time, 8 hrs/day

Process: Thinking + computer assist

Speed: Hours

Cost: $50 (2.5hrs x $20)

Constraint: Human time (24 hrs max)

Value capture: Worker 50%, Company 40%, Tools 10%

AI Era (2023+)

Input: Electricity, GPU silicon, 24/7

Process: Matrix multiply, 1M ops/sec

Speed: Seconds

Cost: $0.01 (1 API call)

Constraint: Energy only (physics)

Value capture: Capital 90%, Worker 5%, Energy 5%

Same output. Different physics. Different economics.

Human time was linear and fixed. AI time is exponential and improving.

Each Era Hits a Different Ceiling

Agricultural (prehistory-1800)

Ceiling: Land
Equation: Money = Work x Acres
Limit: Cannot create more land

Industrial (1800-1970)

Ceiling: Thermodynamics
Equation: Money = Energy x Efficiency
Limit: Entropy, heat loss

Digital (1970-2023)

Ceiling: Human hours
Equation: Money = Cognition x Time
Limit: Cannot add hours to a day

AI (2023+)

Ceiling: Energy
Equation: Money = Electricity x Inference
Limit: Available joules
Note: First non-human constraint

Each era removes the previous constraint. Hits a new one at higher scale.

We have reached the physics layer. No more biological limits.

The Deepest Layer: Negentropy

All economic value reduces to one thing: order extracted from energy.

Plants extract order from sunlight (photosynthesis)
Humans extract order from food (cognition)
AI extracts order from electricity (inference)

Negentropy = Information = Structure = Economic value

Money is a claim ticket on ordered output.

For 200,000 years, humans were the only machine that could extract high-level order from energy. Not anymore.

AI is a more efficient order-extraction engine:

Faster: Seconds vs hours
Cheaper: $0.01 vs $100
Scalable: Copy infinitely vs hire and train

And it improves exponentially every year.

The Sandwich Effect (2020-2026)

This section explains why the transition is happening so fast and so brutally.

Knowledge workers are being squeezed from two directions at once.

Top Pressure: Interest Rates

The timeline:

2020: 0.25% (near zero)
2021: 0.25%
2022: 2.0%
2023: 4.0%
2024-2026: 4.5%

What rising rates mean:

Money is now expensive
Every dollar borrowed costs 4.5% annually
Companies must show profitability, not just growth
Bloated workforces become a liability

Bottom Pressure: AI Capability

The timeline:

2020: GPT-3 (basic)
2021: GitHub Copilot
2022: ChatGPT launches
2023: GPT-4 (agent capable)
2024: Multi-modal + reasoning
2025: Autonomous agents
2026: ???

What falling AI costs mean:

Alternative to knowledge workers exists
Costs 10,000x less
Works 24/7
Improves every 18 months

The Middle: Knowledge Workers

Caught between both forces. Getting compressed.

2021: "We cannot hire fast enough"

2023: "We are automating your job"

That is whiplash at civilizational scale.

Phase 1: The Setup (2020)

March 2020: The world shuts down.

The Fed's response:

Print $3.3 trillion
Drop interest rates to 0.25%

Result: Money becomes essentially free. Companies can borrow unlimited capital at no cost.

Phase 2: The Boom (2021-2022)

ZIRP = Zero Interest Rate Policy

When money is free, companies optimize for growth, not efficiency.

What happened in tech:

Hiring exploded (+45% job openings)
Salaries skyrocketed (+30% YoY)
Six-figure signing bonuses became normal
"Growth at all costs" mentality
Every company hired a bloated workforce

The logic: "Money is free, revenue is up, hire everyone before competitors do."

Tech employment reached all-time highs in 2022.

No one was thinking about cost efficiency. Why would they? Capital was free.

Phase 3: The Squeeze (2023 onward)

2022-2023: Inflation hits. Fed panics.

The pivot: "Cut costs immediately."

Tech layoffs:

2023: 260K+ workers laid off
2024: 150K+ more
2025-2026: Ongoing "right-sizing"

November 2022: ChatGPT launches.

March 2023: GPT-4 arrives.

Suddenly, there is an alternative to knowledge workers.

The Calculation Changes

Before (2021):

Capital is free: Hire everyone
No AI alternative: Must use humans

After (2023):

Capital is expensive: Cut costs aggressively
AI can do knowledge work: Replace humans

Why This Is Uniquely Bad

Previous tech disruptions happened during growth periods:

1990s: Internet boom created lots of new jobs alongside disruption
2000s: Mobile boom created new platforms, new roles

This time:

Happening during contraction (expensive capital)
No new platform creating jobs
AI destroys roles faster than it creates them
Companies have financial pressure to cut AND technological ability to replace

It is a perfect storm.

The Numbers

2021:

Tech adds 500K jobs
AI capability: Low
Interest rates: 0.25%
Result: Hiring boom

2023:

Tech cuts 260K jobs
AI capability: Medium-High
Interest rates: 4.5%
Result: Mass layoffs

2026:

Tech stabilizes (but lower baseline)
AI capability: Very High
Interest rates: ~4.5%
Result: Permanent compression

What Makes It A Sandwich

It is not just one force. It is two simultaneous pressures.

If only rates had risen (no AI):

Companies would cut costs, but still need humans
Wages would drop, but jobs remain

If only AI had arrived (no rate rises):

Companies would adopt AI gradually
Cheap capital means no urgency to cut humans

But we got BOTH:

Urgent pressure to cut costs (rates)
Perfect tool to replace workers (AI)

The timing created a vice grip that neither force alone would have produced.

The Irreversibility

Even if rates drop again, the AI genie is out of the bottle.

Companies learned:

"We can operate with 30% fewer people"
"AI can handle tier-1 tasks"
"Why hire when we can automate?"

The sandwich compressed the knowledge work layer.

It will not uncompress.

Even in the next boom, companies will hire AI-first, humans-second.

The 2021-2022 hiring frenzy was the peak of human knowledge work employment.

We will not see those numbers again.

The New Equations

Old: Money = Human Time x Skill

New: Money = Energy x Inference Efficiency

Time is no longer in the formula.

What This Means

For workers:

Knowledge work wages face permanent deflationary pressure
Your time becomes less valuable every year

For companies:

Profit margins explode
As artifact costs approach zero, revenue stays fixed
Surplus approaches 100%

For society:

Wealth concentrates at the energy layer
Whoever controls compute and electricity controls value creation

The Question

We are not debating whether this happens. The cost divergence is already underway.

The question is:

In a world where money equals energy times intelligence, and humans are neither the cheapest energy source nor the fastest processor, what is our economic role?

For 10,000 years, humans were the answer to "how do you create valuable order?"

For the first time, we are not.

Final Note

The transition from "Money = Human Time" to "Money = Energy x Intelligence" was inevitable.

But the 2020-2023 sequence created a uniquely brutal compression event:

COVID hits, money printing, ZIRP begins
ZIRP causes hiring explosion, peak knowledge work employment
Inflation hits, rate hikes, urgent cost pressure
AI breakthrough, perfect replacement tool arrives
Sandwich activates: Cut costs (macro) + Automate (tech)

The sandwich effect explains the speed of the transition.

Without the sandwich, this might have taken 20 years.

With the sandwich, it is taking 3-5 years.

And we are only halfway through the compression.

P.S World is far more complex then I could imagine. Do not get biased by it.