← Back to the Half-Life Tax calculator

The Economic Anatomy of a Dangerous Autonomous Agent

An agent that poses existential-level risk needs compute to operate. Compute costs money. Someone has to pay. This calculator traces the economic chain that would need to hold for a dangerous autonomous agent to actually run, and shows what breaks at each link. Companion to the Half-Life Tax model.

Shared Assumptions

These describe the current frontier. Adjust to explore different capability levels.

5h
Task length at which the agent succeeds 50% of the time. Currently ~2.5–5h for frontier models.
0.70
κ=1 is exponential. SOTA ≈0.6–0.9. Humans ≈0.37. Does not improve with scale.
$0.22
Average cost of one model call.
$150
$100–200 skilled workers. $300–500 scarce specialists.
80
50 for complex reasoning, 120+ for routine.
20%
Time to review agent output. 5% = cursory, 20% = moderate.
How these are used:Verification economics uses all six to compute agent cost at each task length.  The full chain evaluates all four conditions at these settings.  Bubble dynamics uses T50 as the Feb 2026 starting point for timeline projections; κ is fixed at 0.70 in that tab (the current architectural ceiling). Defaults reflect the telecom-precedent scenario: fire-sale inference (0.5×) and 24-month post-bust doubling.

Verification Is the Weibull Ceiling

The Weibull model makes compute costs gentler than exponential, but verification cost takes over as the dominant expense. Under realistic review fractions, human review accounts for the majority of total agent cost.

Total agent cost (compute + verification, shaded) versus human cost (dashed). The gap between the solid and dashed red lines is verification.

Viability Map

Total cost (compute + verification) as a ratio to human cost. Green (<1×) = viable.

Each cell = (compute + verification) / human cost at current κ and review fraction.

The Verification Floor

The minimum T50 below which even free compute cannot make an agent viable, because review time alone exceeds the cost of a human doing the work.

TaskAt 10% reviewAt 20% reviewAt 30% review
Minimum T50 (hours) at which verification cost alone equals human cost.

The Economic Chain of Necessary Conditions

For a dangerous autonomous agent to operate, all four conditions must hold simultaneously. Each is evaluated at the current parameter settings.

What Happens If the Bubble Bursts?

The T50 doubling trend is funded by capital investment. The telecom precedent (1996–2001) suggests that infrastructure overbuilding based on exponential growth projections ends with a capex collapse, fire-sale pricing on excess capacity, and a deep stall in new buildout. Starting from the T50 set above, with κ fixed at 0.70.

mid-2027
When capex contraction begins. Modal scenario: mid-2027 (telecom precedent: 4–5 years after buildout starts).
24 mo
7 = no bust. 24 = telecom-precedent default. 60 = near-freeze.
0.5×
<1 = fire sale (telecom default: excess GPUs). >1 = oligopoly pricing. Affects pair survival only.

T50 Trajectory

Projected T50 from current value. κ fixed at 0.70 (current architectural ceiling).

Impact on Dangerous Milestones

P(coherent 1-month plan) over time at κ = 0.70.

Milestone Delay Table

MilestoneBaselinePost-bustDelay

Self-Sustaining Pair Survival

Does the pair survive the bust? Under the telecom-precedent default of fire-sale pricing, margins become absurd. The bust kills growth potential without killing the pair.

Inference costCost/deliveredMarket priceMarginSurvives?

Why Coalitions Cannot Overcome κ

Can multiple pairs coordinate to exceed the individual T50 horizon? No. When κ < 1 the hazard rate declines over time, so a surviving agent enters a progressively safer state. Decomposition throws this away: each fresh agent restarts from peak hazard.

StrategyP(160h plan)vs single attempt
At T50 at the bust date, κ = 0.70. Coordination overhead: 40h base + 12h per additional pair.