AI is becoming more useful at the same time that it is becoming easier to take for granted.
The important question is no longer only what increasingly capable models can do. It is what happens when companies, institutions and eventually parts of society reorganize themselves around the assumption that machine intelligence will remain continuously available.
This series examines that dependency from three different directions. The first is physical: AI ultimately depends on electricity, chips, data centers, networks and other infrastructure that can become constrained even when the software itself is working perfectly. The second is organizational: productivity gains can change staffing, training and workflows until human fallback capacity is no longer what it once was. The third is economic and political: when advanced computing capacity is scarce, access is not distributed automatically or equally but through prices, contracts, infrastructure ownership, public programs and other allocation mechanisms.
None of these essays assumes that AI will suddenly disappear or that society is already incapable of functioning without it. The argument is narrower. Technologies become systemically important not only when they are powerful, but when institutions quietly redesign themselves around their reliability.
01 — Energy & Availability
When Society Depends on AI, Can It Withstand an Energy Crisis?
AI looks like software to the person typing into a browser, but the systems behind it are industrial: chips, cooling, data centers and large amounts of reliable electricity.
This essay asks what happens when AI becomes embedded deeply enough in ordinary work that an energy or computing constraint no longer produces only slower software. If organizations have already redesigned themselves around permanently available machine intelligence, reduced AI capacity can become an operational constraint.
02 — Human Fallback
What Happens When Companies Forget How to Work Without AI?
AI can raise productivity, accelerate learning and distribute expertise. It can also change which skills employees practice, which manual processes companies maintain and where organizational knowledge is stored.
This essay examines the distinction between technical recovery and organizational recovery. Restoring an AI service is one process; rebuilding human skills, procedures and institutional memory that have been allowed to decay is another.
03 — Compute Allocation
When AI Compute Becomes Scarce, Who Gets It First?
High-end AI computing is not distributed through a single global pool. Access already depends on cloud capacity, advance reservations, infrastructure ownership, public programs, geography and the ability to pay.
This essay examines a question that becomes more important as AI moves into essential work: when computing capacity is constrained, who can secure reliable access, and what mechanisms decide where scarce machine capacity goes?
One System, Three Questions
The three essays describe different parts of the same system.
Energy and physical infrastructure determine how much machine capacity can be supplied. Organizations decide how much human fallback they retain as they adopt that capacity. Markets, governments and infrastructure owners determine how access is allocated when supply cannot immediately meet every demand.
The resulting risk is not simply that an AI model might fail. It is that dependency can grow faster than the systems designed to absorb a loss of capacity.
That does not make dependence itself undesirable. Modern economies depend on electricity, telecommunications, cloud computing and many other infrastructures because those systems create enormous value. The relevant question is whether AI-dependent organizations and societies understand which capabilities they are giving up, which dependencies they are concentrating and how they would continue operating when machine capacity becomes less available than expected.
This series follows that question across energy, organizations and compute.