The Continuity Drill
A personal capability is not only what it can do at full power. It is what its person can still carry responsibly when one part goes dark.
01 / a capability under stress
A stack should be able to lose a part without losing its person.
App People learned a reliable menu inside a fixed product. Stack People need a reliable way to continue when a component of their evolving capability is absent, altered, or wrong.
That is not a prediction that every AI service will fail at once. It is a basic condition of building useful work around tools that can change their interface, terms, permissions, outputs, and availability. The problem is not merely operational. When a capability carries research habits, working context, delegation rules, and accumulated examples, a missing component can tempt a person to make a faster but less responsible move.
A continuity drill is a deliberate practice of preserving the human question, judgment, and handoff when an AI-enabled part of a Stack Person’s capability cannot do its usual work.
The drill is not nostalgia for doing everything by hand. It is a way to discover what belongs to the tool and what must remain legible to the person. If a research assistant is unavailable, can someone still name the source standard? If a drafting system produces an uncertain answer, can they still identify the claim that needs a review? If an automation loses access, can they make a safe pause visible rather than quietly improvising a promise?
Those questions are different from product backup plans. A company may have redundancy for servers and accounts. A Stack Person also needs continuity of judgment: a practiced ability to carry the work’s purpose and boundary across a break.
02 / the drill bench
Three things worth testing before the break.
A continuity drill does not need a catastrophe. It can be a small rehearsal: remove one convenience and see whether the next responsible action remains understandable. The point is to reveal a dependency while it is still cheap to redesign.
Test 01
Can you restate the question?
When a tool has become the familiar doorway, people can mistake its default prompt or workflow for the task itself. Restating the question recovers the decision, audience, and constraint that the tool was meant to serve.
Try this prompt
Describe the next decision in one sentence without naming a product. Then name what would make the answer unsafe, incomplete, or premature.
Test 02
Can you find the human checkpoint?
Some steps may be automated; accountability cannot be. A useful drill identifies the person who can verify a source, authorize a commitment, or decide that a delay is better than a fabricated certainty.
Try this prompt
For one live workflow, name the moment at which a human must still be able to say yes, no, or not yet. Make that checkpoint visible to the next person.
Test 03
Can you leave a clean handoff?
A break becomes expensive when its context exists only inside a vanished tab, a private memory, or an unspoken habit. A clean handoff records enough purpose and status for another responsible path to begin.
Try this prompt
Write a short handoff that says what has been done, what remains uncertain, and what must not happen until a named check is complete.
03 / from dependency to practice
The fallback is not a smaller stack. It is a clearer one.
It is easy to hear continuity as a demand to preserve an old manual process beside every new capability. That would turn adaptation into a ritual of duplication. The better aim is selective clarity. A person should know which parts of a workflow are conveniences, which are sources of evidence, which are protected context, and which are decisions they remain responsible for making.
That distinction lets a stack change without becoming opaque. One model may be replaced, a connector may lose permission, or a preferred workspace may disappear. The person does not need to carry every implementation detail forward. They do need to carry enough method to choose a credible substitute, explain a delay, and protect the boundary around work that cannot simply migrate.
For organizations, this suggests a more useful question than whether people are using approved AI. Are the consequential questions, reviews, and handoffs understandable when an approved or personal capability is unavailable? A resilient institution does not require each worker to surrender a private stack. It gives people clear shared surfaces for the work that must be visible and safe paths for the work that must pause.
04 / a new kind of readiness
Continuity belongs in capability, not only in infrastructure.
There will be a cultural temptation to measure stack maturity by how little interruption a person experiences. That is an incomplete measure. A capability that hides its breakage may seem fast until it makes an unsupported claim, carries context into the wrong place, or leaves a colleague unable to reconstruct what happened.
A better measure is whether interruption becomes intelligible. Can a worker say, “This was drafted, but its source check did not run”? Can a manager see the difference between a paused automation and a completed obligation? Can a client receive a candid update instead of a polished artifact whose foundations have changed? These are ordinary forms of professional reliability, newly important when more work moves through personal AI capabilities.
The continuity drill also keeps the app-to-stack transition from becoming a story of individual self-sufficiency. Stack People will still depend on platforms, institutions, colleagues, and shared standards. The point is not to escape dependence. It is to know which dependencies are carrying a decision, and to preserve the human ability to respond when they shift.