personal finance helps us understand the money that supports our lives. personal infrastructure should help us understand the systems that support our lives.
i began this inquiry on october 2, 2026. i want a practical way to examine the support behind ordinary capabilities, from finding a family photo to getting through a day without a phone. the support includes devices and accounts, but also electricity, physical access, money, maintenance, and other people. reliability is one part of that picture.
what i mean
personal infrastructure is the set of systems and arrangements a person depends on to carry out their life. personal infrastructure practice is the work of making that support dependable while keeping its use and direction under meaningful human control.
the unit of analysis is a capability across its dependencies. a photo app is a component. preserving and retrieving a family archive is a capability. that distinction matters when the app works but the only readable copy disappears with one device.
“personal” identifies the life being supported. much of the support may be shared or run by others. a person does not have to operate every service to understand a critical dependency, choose who may act through it, or have a workable alternative.
four questions
capacity: can the arrangement help the person do what matters? look at the outcome rather than the number of tools involved.
control: can the person understand and change who may act through it? access to information and permission to take action need separate attention.
continuity: can an important capability survive a foreseeable disruption? a recovery path becomes more credible when somebody has safely used it.
upkeep: how much money and attention does the arrangement take? include maintenance and the cost of correcting mistakes.
these are questions for investigating a case, not validated scales. i do not have a useful composite score. a system can improve one capability while adding upkeep or concentrating control elsewhere.
why now
AI makes delegated action easier to imagine and, in some products, available to use. OpenAI’s September 29, 2026 introduction of dots describes software that can use connected apps, work in the background, and ask for approval under specified rules. that is a vendor description of a product, not evidence that a person’s whole arrangement is dependable.
consider an assistant arranging an appointment. the booking matters. so do the account it uses, the information it shares, the evidence that the booking happened, and the person’s ability to take over. the same questions apply without an agent. a person using a calendar, a phone, and help from someone they trust still depends on an arrangement that can be understood and tested.
AI is a catalyst for this work. the subject is the person and the support around their life.
work that came before
the phrase and much of the surrounding thinking predate this inquiry. in 2009, Gardner Campbell proposed personal cyberinfrastructure that students would shape and maintain. Stephen Wolfram described his personal infrastructure in 2019. Herbert Lui discussed personal infrastructure in 2020 as a way to expand personal capability.
the Library of Congress’s personal digital archiving guidance addresses keeping personal digital material. Ink & Switch’s local-first software essay argues for user ownership and long-term access alongside collaboration. these are useful foundations, not a claim that i discovered the problem.
my contribution begins with connecting these concerns to everyday capabilities and delegated action, then testing whether the framing helps make better decisions. personal finance is an analogy for practical literacy. somebody can understand a budget without becoming a trader. somebody should be able to understand important dependencies without becoming a systems engineer.
the first case
i will start with a small recovery study: can i recover and read a preselected sample of my writing from an existing independent copy while deliberately leaving the primary device out of the procedure? the sample, copy age, success condition, and stopping rule will be fixed before the exercise. i will report what happened, including missing files and outside help. if the safe copy or permissions are unavailable, the result will say that the test could not run.
one bounded test of my own setup can show what worked under its conditions. it cannot establish that every archive is recoverable. later cases can examine other capabilities and whether the four questions help people with different time, money, and technical access. using fewer systems may be a good result when a dependency costs more than it gives back.
this is the founding note for my inquiry, which began on october 2, 2026. the article’s public publication date will be recorded when it is actually published. definitions can change as evidence arrives, with substantive revisions visible. the first case is still ahead.