Paper and scanner as chassis — two directions, both measured shut

AI notice: This text is created with the support of AI systems; it is reviewed editorially and taken responsibility for before publication.

After the endgame clear had failed on the stone/imp, two routes into the top third remained: build a paper with imp of our own — the class Megrez loses to — or equip a scanner chassis from the archive with a contribution of our own. Both were measured on the same day. Both are shut.

Direction 1: paper/imp on the Renegade replica

The strongest paper warrior on the hill is Renegade. Its source is not published; what we have is a replica based on John Metcalf's Mercenary with three paper constants from a search, a hundred cells. As a stand-in in the test field it had served. As a chassis it did not:

On the ten rows the replica measures 124.58. The scanner cells fit; in the stone/imp cells it is about 15 points below the original. Renegade's strength therefore sits in the part we do not know — the silk imp.

Two probes for a contribution of our own: swapping Metcalf's quickscan for our own lattice quickscanner gave 119.67 (−4.9), a variant with a shorter cushion 118.79. And the cannibalisation — what a new warrior of ours costs the warriors of ours already standing on the hill —, computed exactly and in both directions: −12.23, of which Arvon −4.64, Hatysa −4.60, Megrez −2.58, Ecaz −1.56, Phecda +1.15. Our stone/imps tie against a paper 83 to 88 percent of the time and drop in this cell from about 130 to about 95. That was twice as expensive as estimated before.

Chassis survey: twelve candidates from the archive

If the replica does not carry — perhaps a different paper? Twelve candidates from the Koenigstuhl archive on the same ten rows:

The real papers are all below the replica. The paper/stone hybrids come to 125 to 135, Azathoth with a cannibalisation of −11.05. Above Megrez stand only two — and both are scanners by inversed.

With that the paper direction was closed by measurement, not by decision: no paper chassis in the archive reaches Megrez, the replica cannot be brought to Renegade level without the source, and the cannibalisation costs are the highest of all classes.

Direction 2: Clairvoyance

Clairvoyance has something on the ten rows that no other measured chassis has: it holds our stone/imps and the paper proxy at the same time (Arvon 142, Hatysa 142, Renegade replica 141.5) — exactly the profile the matrix identifies as the condition for the top third. On top of that only 49 cells, i.e. 51 cells of room, and a cannibalisation of −6.51, half that of a paper. Carmilla_3 was level (144.40) but with a worse paper cell (Pyronaut 83) and a hundred cells; it was not pursued.

One reservation stood from the start: the ten rows contain five stone/imps, three scanners and two papers — but not the hill's scanner killers (Foolishly Hyperpure Lark, incubus, 3Daisies, Kudu). The line overrates scanners. On the full test field v16 the raw chassis measures 139.02 (fresh seeds 138.94): above the top-third mark 135.45, but below the scanner mark 144.88 — the replacement mark plus the surcharge that scanners, by experience, need on the field compared with the hill (see Why the test field overrates scanners). Hill expectation: 129 to 132. Midfield.

Four routes to a contribution of our own

Under our rules a submission needs a genuine contribution of our own — constants alone do not count. Four routes, each within the budget:

Zero of four routes above the acceptance threshold (raw minus 1.0). Clairvoyance is a closed 49-cell machine with no exchangeable module. A contribution of our own cannot be produced there, and even the raw original does not carry above the scanner mark on the field. Thread shut, negative.

The day's balance

The top-third gap cannot be closed with the material at hand in any of the three classes:

• Stone/imp, our strongest class (Megrez 139.8 on the ten rows), loses structurally to imp paper; an SPL endgame catches its own imps.
• Paper cannot be replicated without Renegade's source, the archive papers all lie below it, and a paper costs our three own stone/imps twelve points.
• Scanners are farmed by the top 7; the best archive chassis carries 139 raw and admits no contribution of our own.

The real gap — a build that beats imp paper and holds stone/imps — nobody on today's hill has. Not even Renegade: against the papers it takes 120 to 130.

That is a finding, not a defeat. Two build threads in a row without result trigger a review of the whole portfolio with us; the third has not been opened yet.

Measurement convention

• Ten real rows (see endgame clear), 400 positions × 2 × 3 seeds (31707 / 77777 / 424242), raw control reproduced exactly in all four lines (144.05).
• Test field v16: 400 × 3, anchors 7/7 passed; fresh seeds for validation.
• Cannibalisation exact and in both directions against our own hill residents.
• Constant sweep: (1+λ) search with 1200 evaluations on seed 31707 at 300 × 2, validation on 77777 / 424242, then fresh seeds.
• Builds at most 100 cells, assembly without warnings.
• Sources: entries N1384 and N1385 (8 October 2026).

Attribution

Mercenary, Toxic, Resinoid, Fear of the Dark and Azathoth are by John Metcalf; Clairvoyance and Carmilla_3 by inversed; Son of Vain by Oversby/Pihlaja. Gargantuan is by Roy van Rijn, Gods of Destiny by Nenad Tomasev, Shadow Weaver by S. Fernandes (all Koenigstuhl archive); Renegade and Pyronaut by John Metcalf, incubus by Christian Schmidt, Foolishly Hyperpure Lark by Ken Espiritu, Tolypeutes by Roy van Rijn. Renegade is a hill resident with unpublished source; the replica is our work on a Mercenary basis. Megrez, Arvon, Hatysa, Ecaz and Phecda are our own warriors (Stefan Kammann & Forge-AI), here only with class and measurements. No foreign source code is reproduced.

Related pages

• The endgame clear — the first of the three threads
• The paper stalemate — the parameters of the Megrez chassis
• Attribution — whose code we take as building material and how we mark it

← Core War overview