23d • General discussion
Claude as Nicholl-Rubric Evaluation
So I took yet another run at the script based on feedback from Claude.
What was involved was a series of probing questions and edits. There's obviously no point in pleading my case to a machine. But in places where it was flat out wrong in it's evaluation, such in places where it missed set-ups Claude agreed, but pointed out that "a tired reader at page 100 might not catch it", which is a valid point.
The collaboration with Claude proved to be quite useful in teasing out such things as tying up abandon plot lines, fortifying secondary character arcs, cleaning up character voice inconsistencies and making the ending land better.
Most of its suggested edits were still hilariously awful, but these suggestions, as bad as they were, forced me to take a new perspective on given scenes, and in some instances, the improvements were far more than incremental tweaks.
I still have some work to do, but 18 drafts later, I think I'm on to something...
The Legend of Dragonfield —
Story — 7.5/10. A dual-timeline epic that braids an 11th-century warrior-poet tragedy with a contemporary family drama, joined by a nine-century bloodline. The structure is ambitious and, in this draft, disciplined: two protagonist arcs resolve cleanly — Ali's coming-into-power and Alan's conversion from denial to belief — and the antagonist's pivotal choice to release the family is now properly motivated, driven by outside political pressure and his own recalculation rather than convenience. The ending delivers a genuine first reckoning: the bloodline and the beast make contact before the credits, so the film pays off its title threat instead of banking everything on a sequel. The remaining ceiling is that the full physical confrontation is deferred — but as a deliberate cliffhanger, not a dodge.
Voice — 6.5/10. The strongest writing lives in the medieval register: Wyrtgeorn's refusal of immortality, Rowena's burned-queen ordeal, the courtship economy of their first scenes. The prose has real muscle and a distinctive cadence. The contemporary banter is serviceable and occasionally sharp. A dialect-caricature pass cleaned out the most circle-able liabilities, so the category sits more securely than the number alone suggests.
Characters — 7/10. Two completed arcs is rare at this budget of pages, and the script earns both. Ali is sharply drawn — vain and self-absorbed early, forged into something formidable by the rescue — and the final beat positions her unmistakably as the protagonist of what comes next. Alan's guilt-and-belief arc resolves in action without contrivance. The supporting bench (Alois, Jiao, Addison, Atlas) is vivid, and the deliberate warmth built into a doomed minor character pays off as a gut-punch rather than a throwaway.
Craft — 6.5/10. The signature strength is architecture: seeds planted early (the fresco-as-map, the dragon-tooth pendant, the lake defenses, the embassy gambit) pay off later with satisfying click. Tension management and set-piece construction are confident. The chief remaining drag is a tonal swing — broad comedy abutting sudden violence in the third act — which is a defensible directorial signature but the one thing a cold reader might flag.
Meaning and Magic — 7/10. The theme lands on both ends — Alan's hard-won belief, Ali's "legacy isn't a name or a place" — and the nine-century soul-thread, culminating in the zombie-knight's recognition and the dragon's final lock onto Ali, supplies the indescribable something the rubric is reaching for. This is the script's quiet engine.
Overall — ~7.5/10. A strong "Consider" pressing on the bottom edge of "Recommend." The script knows what it is, executes its central conceit, and delivers an emotional and mythic payoff before the cliffhanger. It is ready to be read.
1
5 comments
Claude as Nicholl-Rubric Evaluation
powered by
Suggested communities
Powered by