12/20/2025
Share this post:
Export:
A Framework for Solving Long-Term Problems with Short-Term Humans
Erik Bethke & Claude (Opus 4.5) — December 20, 2025
The Problem: Humans are short-term optimizers (days to months). Our technology now has consequences spanning decades to centuries. This mismatch is why we're sliding toward Easter Island-style collapse — everyone acting rationally, collectively destroying the future.
The Insight: Don't fix humans. Fix the game. Design mechanisms that transform long-term problems into short-term incentives, so myopic agents performing local optimization collectively solve long-term problems.
The Framework: Temporal Mechanism Design — the discipline of building "Time Goggles" that let short-term vision navigate long-term landscapes.
The Key Proposal: Universal Growing Assets (UGA) — a sovereign wealth fund that buys 1% of public company shares annually, funded by giving public companies a massive tax break vs. private companies. Owners aren't sacrificing; they're optimizing. Citizens become stakeholders, not dependents. No revolution required.
The Punchline: UBI creates dependents. UGA creates owners. The first is a patch. The second is a fix.
Why It Matters: Every civilization-scale crisis — AI displacement, climate, demographic collapse, democratic decay — is a mechanism design problem. We don't need sermons. We need better games.
Humans are extraordinary short-term optimizers. Evolution built us to respond to immediate threats, learn from direct feedback, and make good decisions under time pressure. These are features, not bugs — they kept us alive for 300,000 years.
But we now wield technology with consequences that extend far beyond our natural optimization horizon. We've built levers long enough to alter climate for millennia, deplete resources across generations, and reshape economies in years. Our levers reach centuries; our vision reaches months.
This paper proposes Temporal Mechanism Design — the discipline of constructing games that transform long-horizon optimization problems into short-horizon incentive landscapes, such that myopic agents performing local gradient descent collectively solve long-term problems.
We don't need better humans. We need better games.
All economic activity traces to energy capture:
The economy is fundamentally an energy transformation system.
Energy + Materials + Labor + Knowledge + Capital → Outputs
Where labor spans: human → animal → mechanical → digital/AI
And capital embodies crystallized knowledge — a machine is just "know-how made durable."
Economic history is humanity's search for better coordination mechanisms:
"Given heterogeneous inputs, capacities, preferences, and information distributed across billions of agents — how do we decide who does what, when, and for whom?"
The arc of coordination mechanisms:
Human → Job → Wage
Jobs were treated as atomic units. Policy, identity, benefits, social status — all organized around this abstraction.
Job = Bundle of Tasks [T₁, T₂, T₃, ... Tₙ]
A "radiologist" doesn't do one thing. They:
AI doesn't replace "the radiologist." It competes at the task level.
| Task Characteristic | Human Advantage | AI Advantage | |---------------------|-----------------|--------------| | Pattern recognition at scale | No | Yes | | Novel situation reasoning | Yes (for now) | No (improving) | | Physical manipulation in unstructured environments | Yes | No | | Emotional presence & trust | Yes | No | | Tireless consistency | No | Yes | | Cross-domain analogy | Yes (for now) | Yes (emerging) | | Low-latency retrieval over vast corpora | No | Yes |
Every job is being decomposed whether we acknowledge it or not.
At any given moment, society has a pool of tasks and a pool of agents (human and AI) with varying capabilities, costs, and availability. How do we optimally match them in real-time?
This is hard because:
It's not just incentives. It's computational:
Central Planner must:
- Gather information from N agents (lossy compression)
- Model preferences (high-dimensional, shifting, tacit)
- Solve allocation (NP-hard combinatorics)
- Communicate plan back (latency + distortion)
- Update in real-time (impossible at scale)
Hayek understood this in 1945: "The knowledge of the circumstances of which we must make use never exists in concentrated or integrated form."
The USSR didn't fail because communists were stupid. It failed because no central node can process the distributed, tacit, real-time information required to coordinate billions of tasks.
| Quantum Annealing | Free Market Dynamics | |-------------------|----------------------| | High temperature | Early market: exploration, failed startups, price volatility | | Energy landscape | Fitness landscape of preferences + production costs | | Quantum tunneling | Entrepreneurs jumping to non-adjacent solutions (innovation) | | Cooling schedule | Market maturation, consolidation, price discovery | | Ground state | Efficient equilibrium (approximated, never fully reached) | | Avoiding local minima | Competition + creative destruction prevents lock-in |
The magic: no single agent needs to understand the whole landscape. Each agent needs only local information (prices, their costs, their preferences) and the system computes a global solution through their interactions.
Price signals are gradient information propagating through the network.
AIs are growing the set of tasks they can perform well at a rate faster than humans can retrain. This creates a musical chairs problem: the least skilled workers are displaced by the second-least skilled who take the remaining human-viable tasks.
AI will do tasks faster and cheaper than humans can command wages. But workers need wages to express preferences in free markets.
┌─────────────────────────────────────────────────────────────────┐
│ │
│ AI capability growth rate >> Human retraining rate │
│ │ │
│ ▼ │
│ Task displacement cascades downward (musical chairs) │
│ │ │
│ ▼ │
│ Wages decouple from productivity (already happening) │
│ │ │
│ ▼ │
│ Workers lose purchasing power │
│ │ │
│ ▼ │
│ Demand signal collapses │
│ │ │
│ ▼ │
│ Markets lose information richness │
│ │ │
│ ▼ │
│ Forced reversion to central planning (or chaos) │
│ │ │
│ ▼ │
│ Central planning fails (per above) │
│ │ │
│ ▼ │
│ ??? (collapse, authoritarianism, neo-feudalism) │
│ │
└─────────────────────────────────────────────────────────────────┘
| Metric | 1950-1970 | 1970-Present | |--------|-----------|--------------| | Productivity growth | High | High | | Median wage growth | High | Flat | | Housing/wage ratio | ~2-3x | ~7-10x+ | | Labor share of GDP | ~65% | ~58% and falling | | Union membership | ~30% | ~10% |
The social contract — "productivity gains flow to workers via wages" — broke.
The gains flowed instead to capital owners, asset holders, credential gatekeepers, and geographic winners.
AI doesn't create this problem. AI accelerates a dynamic already in motion for 50 years.
Markets require demand. Demand requires income. Income (for most) requires wages. Wages require human labor being valued. AI devalues human labor at the task level.
The free market optimizes for effective demand — preferences backed by purchasing power. If purchasing power concentrates, the market optimizes for the preferences of the few.
The information richness of the market — its whole virtue — degrades as fewer people can meaningfully participate.
What makes Easter Island so chilling isn't stupidity. It's path dependence + local rationality = global catastrophe.
Early Easter Islanders:
- Trees abundant
- Statue-building = status competition
- Each clan rationally maximizes prestige
- Coordination to stop = collective action problem
│
▼
Middle Period:
- Trees visibly declining
- Stopping now = unilateral disadvantage
- Social structure DEPENDS on statue economy
- Priests, chiefs, carvers — all invested
- "We'll figure it out" / "The gods will provide"
│
▼
Late Period:
- Last trees falling
- Everyone can SEE it
- But the institutional structure cannot pivot
- No one person chopping is "wrong"
- The system is wrong, but the system IS them
│
▼
Collapse
The man who chops the last tree isn't a fool. He's:
The trap is not ignorance. The trap is that the local gradient points toward oblivion, and no individual agent can jump to a different basin.
| Easter Island | 2025 Economy | |---------------|--------------| | Trees | Human-wage-earning capacity | | Statue building | GDP growth via labor arbitrage + automation | | Clans competing | Corporations, nations | | Priests/chiefs invested | Shareholders, credentialed class, donors | | "Gods will provide" | "Technology creates more jobs than it destroys" | | No mechanism to halt | No mechanism to redistribute productivity | | Seeing last tree fall | Watching median wages decouple in real-time |
Global optimum
(sustainable human-AI economy)
*
/│\
/ │ \
/ │ \
────────────/ │ \────────────
/ │ \
/ │ \
/ │ \
* │
Local minimum │
(current path) │
│
│ ← Energy barrier
│ (coordination cost,
│ political economy,
│ stranded assets,
│ identity disruption)
We are in a local minimum. The path to better equilibrium requires:
And the annealing temperature is cooling. The system is ossifying.
Is the last tree already falling?
Are we the generation standing there, watching, knowing — and still chopping because we don't know what else to do?
The failure mode isn't a dramatic event. It's a slow, visible, collectively rational slide into a basin we can see but cannot escape.
UBI solves the demand side only:
UBI → Consumers have money → They can buy things → Demand signal preserved
But this is half a market:
Free Market = Free Consumers + Free Producers
↑ ↑
(UBI) (???)
If UBI comes from taxing a shrinking base of mega-producers, you get circular dependency on monopoly/oligopoly production. Consumer "choice" becomes choosing between products from the same five companies.
Bread and circuses with extra steps.
Universal Growing Assets (UGA) does something structurally different:
Citizens as OWNERS (not just recipients)
│
├── Claim on productive assets (equity)
├── Voting rights (governance)
├── Stake in diverse producers
├── Identity shift: laborer → stakeholder
└── Alignment: the table is partly YOURS — why flip it?
This is not charity. This is changing the ownership structure of the means of production — but through markets, not expropriation.
┌─────────────────────────────────────────────────────────────────┐
│ META SOVEREIGN WEALTH FUND │
│ │
│ - Buys shares of public companies on open market │
│ - Diversified across sectors, geographies │
│ - Rolling lock-ups (prevents panic selling, short-termism) │
│ - Each citizen = beneficial owner of a slice │
│ - Dividends flow to citizens (growing) │
│ - Voting rights pooled or distributed │
│ │
└─────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ CITIZEN-OWNERS │
│ │
│ - Receive dividends (income stream) │
│ - Asset base grows with productivity gains │
│ - Identity: "I own part of the AI economy" │
│ - Aligned with system success, not against it │
│ - Role shifts: taste, curation, meaning, consumption │
│ │
└─────────────────────────────────────────────────────────────────┘
| Old Model | New Model | |-----------|-----------| | Human as labor input | Human as owner + curator | | Value = hours x productivity | Value = taste + meaning + governance | | Compete with AI on tasks | Complement AI through ownership | | Identity: "I do X" | Identity: "I own, choose, curate" | | Threatened by automation | Aligned with automation |
Humans move up the stack:
Physical labor → automated (19th-20th century)
Cognitive labor → automating (now)
Taste/curation → human + AI collaboration
Meaning-making → still human (for now?)
Ownership/governance → MUST remain distributed (or collapse)
"We invent machines that do all the work, and this is somehow a problem?"
It's only a problem because we tied survival to labor and ownership to the few.
If we had universal ownership of productive assets and labor as optional contribution, AI would be unambiguously good news.
Instead, we experience liberation as threat.
Before accepting the doom loop, we must steel-man the counter-thesis:
As AI takes tasks, early adopters gain productivity surplus first. But they can't sit on the gains — they must deploy capital and take producer risk. When incumbent markets are defended, the path of least resistance is to create NEW markets by taking yesterday's luxury and making it tomorrow's commodity.
Early AI adopter gains productivity surplus
│
▼
Must deploy capital / take producer risk
│
▼
Incumbent markets defended → create NEW markets
│
▼
Take yesterday's luxury → make it tomorrow's commodity
│
▼
Mass market gets access to previously elite-only goods
│
▼
Standard of living rises even if wages don't
Historical examples:
This is real. It's happening. And it provides endless anecdotes for those who want to claim everything is fine.
Trickle-Down Luxury slows the doom loop. It doesn't stop it.
DOOM LOOP VELOCITY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━→ Collapse
↑
Trickle-Down Luxury
(slows it, doesn't stop it)
What it does:
What it doesn't do:
It's precisely because Trickle-Down Luxury is real that the structural problem gets ignored.
"Everyone has a smartphone and Netflix!" "AI is making everything cheaper!" "The poor today live better than kings of old!"
All true. All irrelevant to:
The anecdotes feed the distraction machine long enough for the structural problem to become irreversible.
You get a world where:
TRICKLE-DOWN LUXURY (deflation in consumption goods)
│
├── Real phenomenon, happens automatically
├── Provides anecdotes: "Look, smartphones!"
├── Masks structural decline
├── Buys time, doesn't buy solution
└── Makes the slide comfortable, not reversible
VERDICT: Symptom mistaken for solution
UBI (income transfer)
│
├── Requires continuous political will
├── Doesn't fix ownership structure
├── Citizens remain dependents, not stakeholders
├── Can be revoked, means-tested, politicized
├── Circular dependency on concentrated production
└── "Bread and circuses with extra steps"
VERDICT: Better than nothing, wrong mechanism, fragile
UGA (ownership transfer)
│
├── Self-reinforcing (compounds without ongoing political will)
├── Fixes ownership structure directly
├── Citizens become stakeholders, not dependents
├── Incentive-compatible (owners WANT to participate)
├── Maintains diverse producer ecosystem
├── Aligned with system success
└── Structural solution, not palliative
VERDICT: Correct mechanism, self-sustaining, solves the actual problem
UBI says: "Give people money so they can participate in the market."
UGA says: "Give people ownership so they ARE the market."
The first creates dependents. The second creates stakeholders.
The first requires perpetual political will. The second compounds on its own.
The first treats citizens as consumers. The second treats them as owners.
UBI is a patch. UGA is a fix.
We should search for game structures that solve longer-term optimization problems correctly — not by making humans more thoughtful or long-term thinking — but by transforming long-term problems into short-term problems that allow humans to human.
As they optimize for the short term, they provide positive search pressure (gradient descent) to solutions that are long-term aligned.
Don't fix humans. Fix the game so that human nature solves the problem.
WRONG APPROACH:
Long-term problem → Ask humans to think long-term → Failure
(hyperbolic discounting, bounded cognition, immediate needs, mortality)
RIGHT APPROACH:
Long-term problem
│
▼
MECHANISM DESIGN (Transform the game)
│
▼
Short-term incentives that ALIGN with long-term solution
│
▼
Humans optimize locally (as they will)
│
▼
Gradient descent toward long-term optimum
Human nature is not a bug to be patched. It's the compute substrate you have to work with.
Instead of revolution, expropriation, or moral persuasion:
Change the payoff matrix so current winners choose the new equilibrium voluntarily.
CURRENT STATE:
Private Company: 100% ownership retention, ~25-35% tax rate
Public Company: Diluted ownership, ~21% tax rate
PROPOSED STATE:
Private Company: ~35-40% tax rate
Public Company: ~15-18% tax rate (significant discount)
+ Sovereign Wealth Fund buys 1% of float annually
| | Stay Private | Go/Stay Public | |---|---|---| | Tax burden | High | Low | | Dilution | None | ~1%/year to SWF | | Stock price floor | N/A | Guaranteed buyer | | Liquidity | Low | High | | Valuation multiple | Lower | Higher |
The math for owners:
"I give up 1% per year to citizens, but I pay 15% less in taxes, my stock has a guaranteed buyer, my liquidity improves, and my remaining 99% is worth more because of the price floor."
This is a trade many would take.
The concern about capital fleeing to other jurisdictions is backwards.
The implementing jurisdiction becomes a capital magnet:
┌─────────────────────────────────────────────────────────────┐
│ CAPITAL MAGNET │
│ │
│ ✓ Lowest corporate tax rate (for public companies) │
│ ✓ Largest, most liquid market │
│ ✓ Guaranteed buyer (price floor) │
│ ✓ Best legal infrastructure │
│ ✓ Reserve currency │
│ │
│ "Cost": 1% annual dilution to citizens │
│ (priced in, still the best deal available) │
│ │
└─────────────────────────────────────────────────────────────┘
What happens:
No coordination required. Just be the best option.
Other countries then:
Either way, you win.
That's not spending money. That's ownership stake in the productive economy.
Temporal Mechanism Design is the art of constructing games that transform long-horizon optimization problems into short-horizon incentive landscapes, such that myopic agents performing local gradient descent collectively solve the long-term problem.
A valid Temporal Mechanism must satisfy:
| Design Criterion | How UGA Satisfies It | |------------------|---------------------| | Incentive compatible | Owners get tax break + price floor NOW | | Legible | "Pay less tax" is crystal clear | | Low coordination | Each company decides independently | | Robust to defection | If some stay private, public companies still benefit | | Compounding | 1%/year + growth + dividends = generational wealth | | No martyrs | Nobody sacrifices — everyone optimizes |
We just don't call it that, and we don't see it as a unified discipline:
┌─────────────────────────────────────────────────────────────────┐
│ EXISTING TIME GOGGLES │
│ (Temporal Compression Technologies) │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ OPTIMIZATION TOOLS │
│ ├── Gurobi, CPLEX (linear/integer programming) │
│ ├── QAOA, Quantum Annealing (combinatorial optimization) │
│ ├── Monte Carlo simulation (probabilistic futures) │
│ └── Horizon: hours to months │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ PROJECT MANAGEMENT │
│ ├── PERT charts (critical path analysis) │
│ ├── Gantt charts (temporal sequencing) │
│ ├── Agile sprints (chunking long projects into short cycles) │
│ └── Horizon: weeks to years │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ WHITE COLLAR PROFESSIONS │
│ ├── Accounting (compress future cash flows → present value) │
│ ├── Law (compress future disputes → present contracts) │
│ ├── Engineering (compress future failures → present specs) │
│ ├── Insurance (compress future risk → present premium) │
│ ├── Finance (compress future returns → present price) │
│ └── Horizon: 1-10 years │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ INSTITUTIONAL STRUCTURES │
│ ├── Corporations (compress future profits → present stock) │
│ ├── Bonds (compress future repayment → present lending) │
│ ├── Pensions (compress retirement needs → present saving) │
│ ├── Constitutions (compress future governance → present rules) │
│ └── Horizon: 5-30 years │
│ │
├─────────────────────────────────────────────────────────────────┤
│ │
│ ??? GAP ??? │
│ ├── Climate (50-200 years) │
│ ├── AI transition (20-50 years) │
│ ├── Resource depletion (30-100 years) │
│ ├── Demographic collapse (30-80 years) │
│ └── Horizon: NO ADEQUATE GOGGLES │
│ │
└─────────────────────────────────────────────────────────────────┘
| Era | Technology Lever Length | Time Goggle Horizon | Match? | |-----|------------------------|---------------------|--------| | Hunter-gatherer | Days-seasons | Days-seasons | Yes | | Agricultural | Years-decades | Religion, kings (generations) | Yes | | Industrial (early) | Decades | Corporations, bonds (10-30 yr) | Yes | | Industrial (late) | 50-100 years | ??? | No | | AI/Planetary | 100+ years | ??? | No |
We invented levers that reach centuries, but our longest Time Goggles see ~30 years, max.
A CFO literally cannot put "climate externality in 2080" on a balance sheet in a way that changes today's decision. The goggle doesn't reach.
For every technology lever of length L, we need Time Goggles of at least length L.
Otherwise:
Action taken with short goggles
│
▼
Consequences arrive after goggle horizon
│
▼
Surprise! (Easter Island, climate, AI displacement)
│
▼
Too late to correct
Problem: Cost externalized, benefit diffuse, timeline 50-200 years.
Failed approaches: "Think of your grandchildren!"
Time Goggle:
The goggle works because the carbon price makes every purchase a short-term calculation that embeds the long-term cost.
Problem: Kids are ROI-negative in modern economies. Pure cost center for 18-25 years.
Failed approaches: "Have kids for meaning!" / tax credits (too small)
Time Goggle:
UGA share = base_share x (1 + child_multiplier x num_children)
Fertility becomes financially rational again. Pre-industrial incentive structure rebuilt with post-industrial mechanisms.
Problem: 10,000+ year horizon. No one alive will see consequences.
Solution: Overstated problem. West Texas isn't doing much. Go hard on solar, wind, batteries, and nukes. The waste is manageable with existing empty space.
Problem: Humans and wildlife compete for same land.
Failed approaches: "Humans must consume LESS." (Scarcity framing → zero-sum conflict → failure)
Time Goggle: Get humans OFF Earth.
Current: Humans vs. Nature fighting over Earth
Future: Humans → Solar System, Nature → Earth
No conflict. Both win.
Long-term preservation of biosphere achievable via short-term incentive of "get rich in space."
Problem: Politicians paid less than private sector → captured by donors. Serve donors → get post-office payday.
Failed approaches: Campaign finance reform (whack-a-mole)
Time Goggle:
Old: Policy → (decades) → Outcome → (never) → Politician's wallet
New: Policy → Outcome → KPI → UGA multiplier → Politician's wallet
Align financial interest with national flourishing. Corruption becomes irrational.
Game Designer:
Design → Ship → Players play → Players quit if bad → Metrics in hours
↓
Designer sees, iterates
Feedback loop: days to weeks
Politician:
Design → Pass law → Citizens "play" → Citizens can't quit → Election in 2-6 years
↓
Maybe feedback, maybe not
Cause-effect link obscured
Politician may be gone
Feedback loop: years to decades (if ever)
Professional game designers learn through player churn. If your game sucks, players leave today. You see it in the dashboard. You fix it or you die.
Politicians never see the dashboard:
Politicians design games they never have to playtest.
Tie politician compensation to long-term national KPIs via UGA multipliers:
Old: Policy → (decades) → Outcome → (never) → Politician's wallet
New: Policy → (decades) → Outcome → KPI → UGA multiplier → Wallet
Synthetic feedback loop created. The long-term consequence of their game design shows up in personal wealth.
They're still amateur game designers. But now they're playing their own game.
Every solution follows the same structure:
┌─────────────────────────────────────────────────────────────────┐
│ │
│ LONG-TERM DIFFUSE PROBLEM │
│ (no individual feels it, timeline > human lifespan) │
│ │ │
│ ▼ │
│ │
│ TEMPORAL MECHANISM DESIGN │
│ ├── Attach personal financial outcome │
│ ├── Make it legible NOW │
│ ├── Compound over time │
│ └── Align self-interest with collective outcome │
│ │ │
│ ▼ │
│ │
│ HUMANS OPTIMIZE LOCALLY │
│ (as they always do) │
│ │ │
│ ▼ │
│ │
│ LONG-TERM PROBLEM SOLVES ITSELF │
│ │
└─────────────────────────────────────────────────────────────────┘
These aren't separate problems. They're all instances of:
Long-lever technology + short-horizon humans + missing Time Goggles = failure
And they all yield to the same solution pattern:
Design mechanisms that make long-term outcomes show up in short-term personal incentives.
Easter Island failed because:
No mechanism existed to transform "preserve trees for grandchildren" into "it's in MY interest TODAY to stop chopping."
The gradient pointed toward oblivion, and no game structure redirected it.
The project of Temporal Mechanism Design:
Design the games we need so that the gradient points toward survival — and let humans be humans.
Every long-term crisis we face — climate, AI displacement, resource depletion, demographic collapse, institutional decay — is a mechanism design problem.
The question is not: How do we make people care about the future?
The question is: How do we build Time Goggles so that caring about today IS caring about the future?
We do not need better humans. We need better games.
The discipline of Temporal Mechanism Design asks:
Given that humans are short-term optimizers, what game structures transform long-horizon problems into short-horizon incentive landscapes that myopic agents can navigate toward globally optimal outcomes?
This is not utopian. It's engineering.
This is not philosophy. It's infrastructure.
Just as we built roads for physical transport, wires for information transport, and markets for value transport — we need to build Time Goggles for consequence transport: mechanisms that bring distant futures into present decisions.
The game designer outplayed the economists.
Now let's build the games.
Temporal Mechanism Design: The discipline of constructing games that transform long-horizon optimization problems into short-horizon incentive landscapes.
Time Goggles: Mechanisms that allow short-term vision to navigate long-term landscapes by compressing distant consequences into present incentives.
Universal Growing Assets (UGA): A sovereign wealth fund mechanism where the state buys public company shares at market price, funded by tax differentials, with shares allocated to all citizens under rolling lock-ups.
Lever Length: The temporal reach of a technology's consequences.
Goggle Horizon: The farthest point in time that an existing mechanism can effectively see and influence present decisions.
The Easter Island Trap: Path dependence + local rationality = global catastrophe. The failure mode where every individual agent is acting rationally, but the collective gradient points toward collapse.
Trickle-Down Luxury: The phenomenon where early AI/technology adopters, seeking new markets, take yesterday's luxury goods and make them tomorrow's commodities. Real but insufficient — it slows the doom loop without fixing the structural ownership problem, and provides anecdotes that mask decline.
This essay emerged from a conversation where the shaping went both ways. Erik's game designer instinct for incentive structures shaped how Claude framed the solutions. Claude's synthesis and pattern-matching helped Erik see connections across domains. The back-and-forth compressed ideas that neither would have reached alone.
The question of "who thought what" becomes meaningless. The artifact is what matters.
— Erik Bethke & Claude (Opus 4.5), December 20, 2025
Norway With GPUs: How Hyperscalers Are Running the Cleanest Macro Trade on Earth
While we debate AI models and chatbot personalities, hyperscalers are executing one of the largest macro trades in history: shorting future currencies...
A Review of The Manhattan Trap
A critical analysis examining the assumptions and implications of AI development control proposals
One Sentence, 150,000 Words
I typed one sentence into QuestMaster. It returned 35 design documents, 150,000 words of Mars colony engineering specifications, and 6 interactive sim...
Get notified when I publish new blog posts about game development, AI, entrepreneurship, and technology. No spam, unsubscribe anytime.