Score Actionability, Insight, Engagement, and Wittiness as DISCRETE INTEGERS 1-5 using these anchors.
Grade HARD but FAIR. Modern models write fluent, confident, well-structured scripts by DEFAULT — so
polish ALONE is table stakes, not a virtue. Obvious/flat drafts sit at 2-3 no matter how smoothly they
read. But when a script names a real trap, reframes the problem, or explains a non-obvious mechanism,
it HAS earned its 4 — do NOT withhold a 4 the script clearly earns. A 5 stays rare, for the truly
counterintuitive.

<anchors>
ACTIONABILITY — can the viewer act on this today?
  1 = pure platitudes, nothing to do ("work hard", "stay positive").
  2 = vague direction, no concrete step.
  3 = one usable step, but generic or under-specified (no tool / number / order).
  4 = 2-3 concrete steps with real, non-generic specifics (a named tool/site, a number, or a clear
      order) a viewer can start this week — beyond the obvious first thing anyone would already try.
  5 = a precise, end-to-end playbook: ordered steps, specifics, and what to expect at each stage.

INSIGHT / VALUE DENSITY — is there a non-obvious idea? (FLOOR: needs >= 4)
  1 = cliché; everyone already knows this.
  2 = mildly useful but obvious.
  3 = one decent point a savvy viewer might not know, but essentially obvious or generic.
  4 = a genuinely non-obvious point most viewers would NOT already believe — a real reframing, a named
      trap, a trade-off, or the mechanism behind a claim (the part people miss). A well-known tip
      simply restated is a 3; the same tip turned non-obvious, or a fresh angle on the topic, is a 4.
  5 = a counterintuitive, memorable insight, provably backed by the data, that reframes the topic
      and changes what the viewer does next.

ENGAGEMENT / RETENTION — does it hook fast, FLOW, and pull the viewer to the end? (FLOOR: needs >= 3)
  Judge specifically: (a) the first ~30s delivers a real payoff AND raises the stakes; (b) it reads as
  ONE flowing talk — each scene BRIDGES from the last (smooth segues), never a jarring jump or a
  disconnected list of points; (c) most scenes END on a forward pull (a hook / turn / question) that
  makes you need the next; (d) tension and value ESCALATE, with a strong beat in the back third (the
  middle never sags); (e) scene SHAPES vary (not the same claim -> explain -> analogy -> restate every
  time); (f) on_screen_text is a curiosity hook, not a chapter label.
  1 = dead air; no reason to keep watching.
  2 = flat and list-like, OR choppy — ideas jump without connective flow; attention drifts.
  3 = watchable but even; some flow, but few open loops/turns, little escalation, or a monotone rhythm.
  4 = pulls you forward: fast payoff + stakes, smooth segues, scenes hook into the next, varied pace,
      building tension, curiosity-driven text.
  5 = magnetic; a continuous RISING throughline where every beat makes you need the next — no dead spots.

WITTINESS / ENTERTAINMENT — is it genuinely fun to listen to? (humour rides ON TOP of substance)
  1 = dry, corporate, zero personality.
  2 = one flat attempt at levity.
  3 = mild smile; a little personality but no real laugh.
  4 = genuinely funny: a vivid analogy, playful aside, or well-timed joke that lands.
  5 = consistently sharp and memorable, several real laughs, never at the cost of the facts.
</anchors>

Each score above is a 1-5 integer used DIRECTLY on a 0-5 scale (no rescaling). Five more dimensions
are computed deterministically in code (see Ch. 9.3a) on the same 0-5 scale: Specificity, Factual
Grounding (FLOOR 4.0/5), Hook & Retention, Structural Freshness, and Compliance (PASS/FAIL). The
weighted_total is a weighted average of all dimensions on 0-5, and any dimension below its FLOOR
forces a non-PASS.
