Milestone 0 · scenario replay · seed 20260807
Something in the belt began converting asteroid material into more of itself. Over eleven days a picket watched it, a directorate authorised a strike on information half an hour stale, and a two-kilometre beam array put 131 megawatts onto its radiator. Nobody involved can tell you whether it worked.
The engagement
The amber marker is where the seed actually is. The cyan marker is where the Heliocentric Commonwealth believes it is, with the disc showing how wrong that belief could be. The Commonwealth never sees the amber marker. Neither does its beam.
Eleven days of watching, six of closing, and one shot.
Finding one
Each observation travels twice: photons from the seed to the picket, then a report from the picket to Earth. Amber is the light in flight; cyan is the report in relay. By the time anything reaches a decision-maker it describes a world roughly half an hour gone.
| look | range AU | SNR | light min | relay min | age on arrival |
|---|---|---|---|---|---|
| 1 | 0.90 | 48 | 7.5 | 20.8 | 28 |
| 2 | 0.90 | 47 | 7.4 | 20.7 | 28 |
| 3 | 0.89 | 46 | 7.4 | 20.6 | 28 |
| 4 | 0.89 | 45 | 7.4 | 20.4 | 28 |
| 5 | 0.88 | 44 | 7.3 | 20.3 | 28 |
| 6 | 0.88 | 43 | 7.3 | 20.2 | 28 |
| 7 | 0.87 | 42 | 7.3 | 20.1 | 27 |
| 8 | 0.87 | 41 | 7.2 | 20.0 | 27 |
Finding one and a half
Authority is not a switch. The directorate reviews its policy every time the picture changes — 8 times in eleven days — and declines 7 of them, because release requires eight independent looks and not one confident classification. A rule that fires on a single good look is exactly what a decoy is for.
| day | review | why |
|---|---|---|
| 0.02 | withheld | class confidence 0.72 below required 0.80 |
| 1.52 | withheld | track rests on 2 observation(s); 8 independent looks required before release |
| 3.02 | withheld | track rests on 3 observation(s); 8 independent looks required before release |
| 4.52 | withheld | track rests on 4 observation(s); 8 independent looks required before release |
| 6.02 | withheld | track rests on 5 observation(s); 8 independent looks required before release |
| 7.52 | withheld | track rests on 6 observation(s); 8 independent looks required before release |
| 9.02 | withheld | track rests on 7 observation(s); 8 independent looks required before release |
| 10.52 | released | policy pol.scd.0001 permits beam_engagement; track trk.belt.0001 meets every confidence predicate |
Finding two
Eight looks pulled the track from 164 km of uncertainty down to 56 km. That is a real improvement and it is three orders of magnitude short of what the shot needs. Uncertainty enters the beam's spot budget directly, so a vague estimate is not merely a riskier shot — it is a physically weaker one, spread across every place the target might have been.
This is why the interceptor exists. Terminal designation is not a refinement of the engagement; it is the precondition for one.
Finding three
The beam does not breach anything. It adds 131 MW to a machine already shedding 125 MW, and asks its radiator to find a new equilibrium. The radiator cannot. It fails, and the seed throttles to whatever the surviving panel can carry.
Finding four · the one that decides whether this is a game
Two days later the picket looks again and measures 18 MW where there had been 125. Something changed. But a half-sized radiator running hot and a full-sized one running cool are the same measurement — flux cannot separate area from temperature. The seed may be crippled. It may simply have noticed the beam and stopped advertising.
Run again with the strike scripted to fail, and the posterior lands at 0.54 — still uncertain, in the other direction. Both sit inside the (0.15, 0.85) band the milestone requires, which was written to be able to fail.
That test is the whole reason to build M0 before anything else. If battle damage assessment had come out legible, the fog would have been decorative and the design premise false. It did not, and the ambiguity is structural rather than tuned: it survives because of what a radiometer physically cannot measure.
What the machine guarantees
Beam emission reappears precisely as absorbed plus missed. Of 6.6 GW leaving the array, 131 MW reached the target and the rest lit empty space. Waste is accounted, not ignored.
Event identity is a hash of causal position rather than a counter, so a counterfactual that changes one branch leaves every unrelated event untouched and a replay diff points at a cause instead of a symptom.
The sky is 8.4 trillion diffraction-limited pixels wide. A processing seed is detectable to tens of AU, so sensitivity was never the constraint. Where you point is.
It received a frozen view holding tracks, arrived messages, and last-reported own-unit status — with no reference that leads back to world state. The firewall is structural; there is nothing to police.