TOPIC HUB // 6 ESSAYS
AI Economics
Capability is commoditizing; the cost curve is the frontier. These essays trace the shift from competing on model capability to competing on useful intelligence per joule - and what that does to margins, infrastructure and strategy.
The useful unit is not a token and not a parameter. It is an accepted result: generation, retries, orchestration, verification and human exceptions priced together.
When models converge in capability, the durable advantage moves down the stack to energy, routing, context discipline and the operator who turns those constraints into completed work.
Capability Is Commoditizing. Cost Is the Frontier.
The moment a capability stops being scarce, the market reprices around delivery, not intelligence. Capability is commoditizing; cost is the new frontier.
READ ->Why Humanoid Robots Turn AI Efficiency Into a Battery Problem
A humanoid may carry only 1-3 kWh. Inference, perception and motors all draw from the same battery, so intelligence per joule becomes a design constraint.
READ ->The Social Permission to Burn Tokens
AI won't lose legitimacy for being imperfect. It'll lose it when its energy curve outgrows its impact. The next phase earns trust, not attention.
READ ->OpenAI Is GPU-Constrained, Not Demand-Constrained
OpenAI isn't running out of money. It's burning capital to capture the dominant inference surface before models commoditize. The real constraint is GPUs, power, and grid capacity.
READ ->The Biggest Customer Becomes the Competitor
OpenAI went from design to silicon in nine months and pointed it straight at Nvidia. When the compute bill gets big enough, the biggest customer always becomes the next competitor.
READ ->Models Are Commodities. Clean Data Is Not.
Your agent made headlines. Did it make money? Trillions in market cap sit silent while startups chase likes. Models are commodities; clean data isn't.
READ ->What is AI infrastructure economics?
AI infrastructure economics measures the full cost of producing accepted AI work: inference, context, retries, orchestration, verification, energy and human exceptions.
Why is cost per verified task better than cost per token?
A cheap token can still produce expensive work when failure rates and review costs are high. Cost per verified task prices the result an organization can actually use.
How much does an AI agent cost per completed task?
Divide the fully loaded agent cost - inference, context, tool calls, retries, verification and human exceptions - by the number of outputs that pass acceptance. Token price alone cannot answer the question because failed and unverified attempts still consume money and energy.
How much energy does a humanoid robot use per day?
Daily energy depends on battery size, actuation duty cycle and always-on inference. A 2 kWh battery, 400 W actuation at 60% duty, 150 W inference and 40 W overhead lasts about 4.7 hours; the Humanoid Energy Budget calculator lets operators replace those assumptions.
Read these before the feed catches up.
Every essay here is published first at piszczek.pl. Follow along on LinkedIn or Substack.
ALL ESSAYS SUBSTACK