Unpredictability

The operator side of a sustained targeting campaign is, structurally, a prediction engine. They observe what you post, where you go, who you talk to, what you read, what time you sleep, which platforms you use, how you write, what you fear, what you want. From that observation they build a model of you — what you'll do next, how you'll respond to a given stimulus, where the pressure points are, which contact-channel reaches you. The model is the operator's working tool. Most of what they spend resources on is keeping the model current.

A model that cannot keep up with the target is a model that produces wrong predictions. Wrong predictions mean wrong tactics, wasted operations, false confidences, and missed opportunities — all costly on the operator side. The technique here is to make the model wrong as often as possible.

The mechanism is not complexity — operators have analysts and increasingly have ML to handle complexity. The mechanism is unpredictability. Specifically, the kind that defies pattern recognition because the underlying behavior is not following a pattern at all.

This site is itself an instance of the technique.

I could have been publishing here, day-by-day, from the beginning of the targeting awareness. A daily journal that accumulated entries at a steady cadence would have been ingestible. The operator side would have settled into a monitoring posture matched to the content's cadence, register, and volume — an analyst could've kept up with my output as a tab in their browser.

I did not do that. I let them settle into watching my older, unpolished, lower-stakes social-media presence and treating that as the signal that defined me. Then this site arrived as a coherent body all at once, with a working tactics vocabulary, a foundational set of techniques, an evidentiary timeline, cross-links to six other reference sites in the same portfolio, and a clear authorial identity in the honto.me cluster. The model the operator side had built — target is the kind of person who posts rambling thoughts on social media — no longer fits the input. They now have to upgrade. New tooling, new analyst skills, new time budget, a different reading level required. Their per-target cost just rose by a step function. We did the upgrade once on our side; they have to do the upgrade for every published artifact on their side, continuously.

That is the technique deployed once, deliberately. The same logic applies in smaller everyday ways.

Zig when they expect zag.

If a model expects you to react to a provocation, don't. If it expects you to ignore one, react. If it expects you to publish on Mondays, publish on a random Thursday. If it expects you to take the highway, take the back roads — but not every time, just sometimes, and not on a schedule. Any consistent contrarian behavior is just a new pattern they can model. Inconsistent contrarian behavior is the thing that breaks them.

Random is not the same as planned variety.

A defender who "varies their routine" by alternating between exactly two options has created a new two-option model. A defender who rotates through a list of five behaviors on a schedule has created a five-option-on-schedule model. Real unpredictability requires actual stochastic behavior — outcomes that depend on a coin flip or a die roll or genuine impulse, not on a pre-decided rotation. The defender who introduces randomness into their own decision-making — I'll go to whichever grocery store I think of first when I get in the car — produces behavior that resists modeling because there is no model to extract.

This is one of the more counterintuitive parts. The defender's instinct under pressure is to plan the variation, because planning feels safer than randomness. But planning produces patterns. The patterns are the input the model wants. Surrendering some of your own decision-making to actual randomness is the move that resists modeling.

Bursts and silences.

A steady stream of output is easier to ingest than a series of bursts separated by silences. The operator side can keep up with a daily blog. They have a much harder time with a publisher who goes silent for three weeks, then publishes nine entries in one day, then goes silent for ten days, then publishes one short piece, then silence for a month. Each burst requires a re-engagement; each silence is information they cannot get; the combination is what defeats steady-state monitoring.

The cost-imposition logic compounds: every silence the operator must wait through is an open question about what's happening (more analyst attention spent); every burst is a sudden volume of work to process (more analyst hours).

Tactic-switching mid-stream.

The same content can land through different channels — long-form prose, short video, art, song, satire, a direct message to a specific person, a sworn affidavit. The operator's tooling is usually bespoke per channel. Moving the same message across channels in unpredictable sequence forces the operator to either build new tooling for each channel or accept incomplete coverage. Both are costly.

Why the operator side hates unpredictability.

Modeling is expensive; re-modeling is more expensive. Each unpredicted action by the target triggers an analyst review: was this random or is this a new pattern starting? The wrong answer in either direction is expensive — assume random when it's a new pattern, you miss the shift; assume new pattern when it's random, you waste effort building a model that doesn't predict anything. Repeated enough times, the cost per unit of intelligence rises until the budget question asks: is this target worth the analyst hours?

This is the same exchange-rate logic underneath cost imposition. Unpredictability is the specific cost-imposition vector that hits the modeling layer. Other vectors hit the recording layer (data poisoning), the social layer (mockery as armor), the legal/enforcement layer (provocation as catalysis). Unpredictability hits the part where they try to anticipate you.

The discipline.

• Don't predict yourself out loud. A defender who tells the world "I always do X" has told the model the same thing. The fewer of your patterns you commit to publicly, the more room you have to break them.
• Occasionally do things that defy your own brand. A jazz fan listens to country for an afternoon. A long-form writer posts a one-line tweet. A measured commentator publishes a wild limerick. Small departures from self-pattern matter even at zero practical cost — they teach the model that your brand is not a reliable predictor.
• Use real randomness for low-stakes decisions. Phone alarm at a random minute. Route to the grocery store picked by coin flip. Which store, which day, which time. The cumulative effect is a behavioral noise floor that's hard to read through.
• Unpredictability is a posture, not a constant active practice. The defender who is straining to be unpredictable in every moment is exhausted and the strain is itself a pattern. The point is to have unpredictability available as an option, deployed when it matters, with enough baseline noise in the routine to make the major moves harder to anticipate.

The connection to cost-imposition.

Cost imposition names the strategic logic: deny the easier path, raise the price of the harder one, until the math stops working on a budget review. Unpredictability is the specific cost-imposition vector that targets the operator's most reusable asset — the model of you. The model is what justifies the budget. Every prediction the model gets wrong is a piece of evidence on a future budget memo against continuing the operation.

A defender who is hard to model is a defender who is hard to justify continuing to spend on. That is the win condition this technique aims at.