A #2 a day

What it is

A Grimshaw on e7. The threat 2.Qf6# runs from d8 through e7, and Black can block that path with either line piece. 1...Re7 shuts the bishop f8 off d6, so 2.Qd6# works (the pawn c5 protects the queen). 1...Be7 shuts the rook g7 off d7, so 2.Sd7# works. Each wrong mate fails because the unit that arrived still holds its own square: after 1...Re7 the rook takes on d7, and after 1...Be7 the bishop takes on d6. A third rook defence, 1...Rf7, guards f6 and pins the key rook against its king. It also closes the bishop g8's line to d5, so 2.Qd5# follows. The solver reports the Grimshaw on e7.

Flaws

• The Grimshaw already stands in the set play with the same two mates (1...Be7 2.Sd7#, 1...Re7 2.Qd6#). The key brings the threat and 1...Rf7 but changes neither thematic mate.
• The key vacates the threat square, which a solver will look at first. The rook does keep jobs after it: it protects f6, covers f4 and f5, and is the unit 1...Rf7 pins.
• 1.Rf4? fails to 1...Kxf4. It is not a thematic try (it threatens 2.Qxd4# and 2.Qf6#) and is not printed, but the critique counts a try refuted by a king move as a big flaw. A pawn g3 guarding f4 turns 1.Rf4 into a second key.
• Black construction units: the pawn g6 keeps the rook off g6 (from where it would guard f6), h6 stops 2.Qg5# as a second threat, b5 guards c4 against 2.Sc4# after 1...Rf7, and d4 fills d4 and closes the c5-f2 diagonal (without it 1...Bxc5 would give check). The critique calls h6 'not needed'. That is wrong: without h6 the key threatens 2.Qf6# and 2.Qg5#. The bishop g8 serves only the 1...Rf7 variation and keeps 2.Qd5 from being a second threat.
• Six moves that do not stop the threat allow a second mate too: 1...Bh7 and 1...Bd5 (2.Qd5#), 1...Bf7 and 1...Rd7 (2.Sd7#), 1...Bd6 (2.Qxd6#), 1...h5 (2.Qg5#). None is a defence, so these are not duals in the play, but the critique lists them.
• The anticipation check covers only the Problemesis collection (7,912 problems). B. Delobel's 6+7 of 2006 (psis-7ba72c1d6b) blocks a queen's path the same way: same mechanism, other geometry and other mates.

Pluses

• The theme works on the threat piece's path, so 1...Re7 and 1...Be7 are real defences, each with its own single mate. Dual avoidance is unified: in both variations the arriving unit keeps control of the other mating square.
• A third defence by the same rook, 1...Rf7, with a pin of the key piece and a line closure of its own (g8-d5).
• The try 1.Qxf8? (2.Qd6#) fails to 1...Re7!, the thematic rook move. In the solution 1...Re7 is met by 2.Qd6#, the try's threat (the solver's Dombrovskis detector).
• The key rook has exactly one free square: f2 holds the king, f4 is next to the black king, f5 gives check, f7 and f8 are captures.
• 7+8, and the trim tool cannot remove any unit.

How it was made

The theme came up for the first time and had no reference in the kit. Four bare-to-light evolve searches gave only checking keys (three of the seeds were illegal at first: a white piece already gave check). The first hand sketch had the threat 2.Qe8# check down the e-file through e7. It failed because the rook and bishop simply interpose after the check, so nothing defends. That is the rule the run learned: the crossing square must be on the path of the threat piece, as in Delobel's problem. The second sketch, 2.Qd8-f6#, gave the two Grimshaw mates at once, but 1...Rf7 (the square beyond e7) guarded f6. A post-key search found the cure in five minutes: a black bishop g8 whose d5 guard the rook closes on f7. Keyfind then found no key, because the post-key position mated by checks. What worked was to give the key piece one retreat square: the rook on f6 with the white king on f2. Six cooks followed, cured one unit per solve. A pawn g6 put the rook on g7 (no Bg7 dual). The bishop b3 went, because it gave 1.Bxg8 and 1.Bf7. Then 1.Qb8+ appeared, and a knight c7 blocks that diagonal while covering e6 and d5. About 20 minutes of composing, 35 solves, 14 searches.

Lesson

In a Grimshaw the crossing square must lie on the threat piece's path to a square next to the king, not on the line of the checking mate. Otherwise the black pieces just interpose after the check and there is nothing to defend. Then check the squares just before and beyond the crossing on the rook's line: a rook that can stop there must not guard the threat square, or that move needs its own mate.

What changed in the kit

A Grimshaw reference (.claude/skills/themes/references/grimshaw.md): the crossing on the threat piece's path, the square beyond the crossing, a bishop that leaves both lines, one retreat square for the key piece, the cooks met with their cures, dead ends and archive models. blog/themes.json now points to it, and the themes skill lists it. Tested in tests/test_blog.py: every theme reference named in themes.json exists, and the reference position solves to its key.