No public post URL exists for this piece — no "view online" permalink appears in the email body, only click-tracked workshop-signup links. This is the fifth and final email in the active Commoncog x Shadowbox sequence, following [[2026-08-25-commoncog-why-tactical-decision-games-work]] (the original, unsponsored essay), [[2026-11-01-commoncog-making-of-common-sense-tdg]] (first Shadowbox-sponsored installment), [[2026-11-02-commoncog-tdgs-accelerated-osmosis-expertise]] (second), and [[2026-11-03-commoncog-making-of-hindsight]] (third) and [[2026-11-07-commoncog-tdgs-in-the-age-of-ai]] (fourth, Cedric solo). This is the "surprise guest" finale that the 11-03 and 11-07 notes both flagged as coming: John Schmitt himself — Shadowbox founder, inventor of the Tactical Decision Game, former Marine Corps infantry officer — writing the origin story in his own voice, not narrated by Cedric or Jared Peterson. Confirmed NOT a duplicate of the 11-07 note: 11-07 ("TDGs in the Age of AI") is Cedric's own essay on borrowing expert judgment via exposure; this one is Schmitt's first-person memoir of inventing the TDG format itself. Different author, different argument, different content — same sponsor sequence.
Why this is in the vault
A concrete, reusable design recipe for constructing a decision-forcing training scenario from a blank page — not the abstract theory of why TDGs work (covered in the 08-25 note) but the literal mechanical steps Schmitt used to build the first one, which is directly transferable to agent-evaluation scenario authoring.
The core argument
Schmitt tells the origin story of the Tactical Decision Game, set in 1988-1990 ("The Swirl"). Three threads converged: (1) frustration with the Marine Corps' Military Decision-Making Process — an "extreme instantiation of rational choice theory" requiring a weighted-criteria matrix mid-firefight, which he knew was "obvious nonsense" because no Great Captain of history ever built a decision matrix in combat; (2) discovering Gary Klein's Recognition-Primed Decision model, which argued experts don't compare courses of action analytically, they pattern-match from experience — raising the question of how you train an intuitive process; (3) having just written FMFM 1, Warfighting, the Marine Corps' new maneuver-warfare doctrine — short (77 pages), abstract, full of concepts like "commander's intent" and "surfaces and gaps," and a document Marines struggled to translate into practice, which Schmitt took personally since making it accessible had been his job.
The breakthrough came watching a German military attaché, Brig. Gen. Hasso von Uslar, pull a topographical map out of his briefcase mid-conversation and say "Okay, we're here and the enemy's there. Then this happens. What would we do?" Schmitt drove home and built the first TDG that night using pure bricolage: faced with a blank sheet of paper, he drew a line down the middle ("that's a river") to make the blank page less overwhelming, then added one constrained element at a time — a bridge because rivers have bridges, a road because bridges lead to roads, a town because roads lead somewhere, hills to explain why the roads curved, a wooded triangle that looked like a good assembly area. The dilemma emerged from the terrain itself: what if the assembly area already has an enemy unit in it? He wrote a backstory explaining how the situation arose, added a time limit for stress, and — critically — stopped the narration right before the climax, forcing the reader to supply the ending. He named it "Enemy Over the Bridge," submitted it on a whim to the Marine Corps Gazette, and editor Col. John Greenwood fast-tracked it into the April 1990 issue. Reader-submitted solutions flooded in; Greenwood made it a monthly feature; the TDG movement in the Marine Corps was born from that single submission.
Mapping against Ray Data Co
The concrete, specific connection: Schmitt's bricolage method — pick one arbitrary constraint to escape blank-page paralysis (a line becomes a river), then add just enough concrete, causally-linked detail (river → bridge → road → town → terrain) until a forcing dilemma emerges on its own, then stop the narration right before the climax so the subject has to decide — is a transferable recipe for authoring agent-evaluation scenarios, which RDCO doesn't currently have a method for. RDCO's existing critic stable (verify-vault-write, verify-strategic-output, verify-dispatch, behavior-critic, etc., per the skills index) all score an already-produced artifact against a pre-written rubric — they are judges, not scenario generators. None of them construct the test case itself; they're handed one. Schmitt's method is specifically about constructing the dilemma from scratch in a way that's concrete enough to force a real decision rather than invite abstract theorizing — the same gap his FMFM 1 doctrine had (correct, abstract, and un-actionable until a TDG made it concrete). That's the sharper, non-redundant point versus the 11-07 note's "ask what a novice gets wrong" framing: this piece isn't about eliciting an expert's judgment on an existing artifact, it's about building the artifact that forces judgment in the first place — the step upstream of every critic RDCO runs. If RDCO or phData ever builds TDG-style scenario evals for Claude agents (the "agent-evaluation" tag already flagged on the 11-01 installment), this is the literal construction method: start from one constraint, layer concrete causal detail, cut before the resolution.
⚠️ Sponsorship
sponsored: true / sponsor_entity: Shadowbox. This is explicitly billed in the email as "the final email in a series of five... written specifically for this workshop," repeats two "Purchase a seat" links, and restates the Nov 10 sales deadline and 20-seat cap. Unlike the prior four installments, the guest author here is not Jared Peterson but John Schmitt himself — the actual Shadowbox founder and paid workshop instructor — making the commercial relationship as direct as the sequence gets: the piece is both content marketing and literally written by the person selling the seats. Cedric again discloses he is a paying participant and Jared will assist with teaching. The origin story itself reads as independently credible (specific named sources — Klein, Card, von Uslar, Greenwood — and self-deprecating detail like admitting the "Enemy Over the Bridge" title was a design mistake), but its placement as the sales-sequence finale, timed two days before the deadline, means the same "selection and timing serve the sale" caveat applies.
Tracked-entity note: John Schmitt (Shadowbox founder/CEO, TDG inventor) already flagged as a CRM candidate in the 11-07 note; this installment is the first where he is the sole bylined voice rather than a referenced figure.
Related
- [[2026-11-07-commoncog-tdgs-in-the-age-of-ai]]
- [[2026-11-03-commoncog-making-of-hindsight]]
- [[2026-11-01-commoncog-making-of-common-sense-tdg]]
- [[2026-08-25-commoncog-why-tactical-decision-games-work]]
- [[feedback_workflow_agent_output_integrity]]
- [[project_rdco_dormant_phdata_main]]