Everything that follows gets cheaper for the attacker.
Part 1 described what has already happened. This part is the only one that looks forward, and it argues one thing: the defensive assets that matter cannot be bought quickly, which means the decision to start is worth more than the budget behind it.
- The next shift is autonomy, not better malware. Operations are moving from scripted tools run by people to goal-directed agents that plan, adapt and persist without an operator watching.
- Four defensive assets take years and cannot be purchased in a crisis. Legal authority, sensor coverage, a detection corpus, and a trained workforce. Every one of them accrues only by starting.
- Delay is priced in exposure, not in budget. The model in section 3 shows a nation starting three years late carrying roughly twice the accumulated exposure of one that starts now, with the same spend.
What becomes routine next
The mistake in most forecasting here is to imagine more sophisticated malware. The trajectory is not sophistication, it is autonomy and scale. Each step below removes a human from a loop that currently limits how fast and how widely an operation can run.
Automated preparation as standard practice
Continuous reconnaissance, rapid weaponization of published patches, fluent localized deception and per-target malware builds are in routine use by actors well below state level. This is the present, not the forecast.
Goal-directed intrusion agents
Operations given an objective rather than a script: enumerate, decide, pivot, and persist while adapting to what the defender does. The operator supervises several campaigns instead of running one, which multiplies a small team the same way it multiplies a defensive one.
Discovery to impact in a single loop
Finding a defect, building the exploit, delivering it and acting on the objective collapse into one automated chain. Dwell time stops being the window in which a defender investigates, because the operation completes before a human is paged.
The defensive model becomes the target
As defense moves to learned baselines, those baselines become worth attacking: poisoning the training of a detection model, shaping traffic during its learning window, or extracting its thresholds to plan an operation that stays underneath them. Model supply chain becomes a national security dependency in its own right.
Sector-wide simultaneity, and harder attribution
Campaigns run against every operator in a sector at once rather than the largest one, because marginal cost per additional target approaches zero. Tooling becomes generic enough that code similarity carries less attribution weight, which complicates the political response as much as the technical one.
Harvest now, decrypt later
Encrypted government traffic and archives collected today remain valuable to an adversary that expects to decrypt them within the lifetime of the secrets they carry. Cryptographic migration is a multi-year inventory and replacement program that has to begin before it is urgent.
What cannot be bought in a crisis
A government can approve a large budget in a week. None of the four assets below can be delivered in a week, a quarter, or in most cases a year. They accrue only by having started, which is what makes the timing decision more consequential than the funding decision.
| Asset | Why it resists acceleration | Typical lead time |
|---|---|---|
| Legal authority | Mandate, telemetry access, mandatory reporting and carrier cooperation require legislation or regulation, consultation, and often constitutional review. No procurement shortens it | 18 to 36 months |
| Sensor coverage | Physical placement, peering and tap agreements, and access negotiated operator by operator across sectors that do not report to the same ministry | 12 to 30 months |
| Detection corpus | Accumulates only by operating: every incident, every false positive tuned, every technique seen on your own networks. It cannot be bought because it is specific to your estate | Continuous |
| Trained workforce | A nucleus can be recruited. An operations tier has to be grown, and the people who train it must first exist inside the country | 24 to 48 months |
| Interagency trust | Who defers to whom at three in the morning is settled by exercises and by shared experience, never by an organizational chart issued during a crisis | Peacetime only |
Table 3. Lead times are planning estimates drawn from national programs of comparable scope and will vary substantially with legal system and institutional maturity.
The cost of starting later
This is a model, not a forecast, and its assumptions are listed beneath it so they can be argued with. It holds spend constant and varies only the start date, which isolates the one variable a leader actually controls today.
Cost of delay model
Assumptions: adversary effective capability compounds at 35 percent per year, indexed to 1.0 in 2026. A program reaches full capability 36 months after it begins, ramping linearly from today's level. Exposure debt is the accumulated shortfall between threat level and capability across a fixed 2026 to 2032 window, so start dates are compared on the same horizon. Adequacy is capability at or above 60 percent of the prevailing threat level.
The decision in front of a government
Two nations with identical budgets and identical institutions will not end up with identical capability, because one of them started while the legislation was still being drafted and the other waited for the legislation to pass. The first has three years of detection corpus, trained people and placed sensors when the second is signing its first contract.
Nothing in this section argues for spending more. It argues that the same money spent earlier buys a different category of outcome, and that the components that compound are precisely the ones that cannot be accelerated later with a larger cheque.
The practical first step is small and reversible: establish what an adversary can reach today, in evidence rather than assumption, and put the legal authority into drafting in parallel. Everything else can follow the evidence.
The cheapest year to begin was last year. The second cheapest is this one.
A briefing sets out what your country is exposed to now, where the four compounding assets currently stand, and what the first twelve months would produce. It commits you to nothing.
Autogon Inc.. Section 1 is projection built from demonstrated capability and should be treated as planning input rather than intelligence. The model in section 3 runs locally in the browser, transmits nothing, and is a planning aid whose assumptions are stated in full above it.