Lead-to-Mock — 4-way comparison

Same prospect (petri-pools), four chain runs varying model + prompt-tuning style. Tap a screenshot to open it full-size; "Open live mock" loads the actual rendered HTML.

generic = CHAIN_PROMPT.md (universal overrides + step spec) tuned = model-specific variant (CHAIN_PROMPT_OPUS.md / CHAIN_PROMPT_SONNET.md) xhigh = Opus-4.7-exclusive effort level (deepest bounded)

Live mock URLs (open in browser)

Opus Generic

Opus 4.7 CHAIN_PROMPT.md (generic) --effort default (high)
✓ ready
mobile 390opus-generic mobile
desktop 1440opus-generic desktop
HTML size15,744 bytes
<img> refs3
<section>5
Picked variantEditorial Owner Portrait (score 9.5)
Other scores7.7
OPEN LIVE MOCK →https://lead-to-mock-compare.pages.dev/mocks/opus-generic/index.html

Sonnet Generic

Sonnet 4.6 CHAIN_PROMPT.md (generic) --effort high
✗ FAILED at step 1
completed: none
mobile 390
desktop 1440

Opus Tuned

Opus 4.7 CHAIN_PROMPT_OPUS.md --effort xhigh
✓ ready
mobile 390opus-tuned mobile
desktop 1440opus-tuned desktop
HTML size31,211 bytes
<img> refs4
<section>7
Picked variantBillboard Diagnostic (score 9.8)
Other scores9.2
OPEN LIVE MOCK →https://lead-to-mock-compare.pages.dev/mocks/opus-tuned/index.html

Sonnet Tuned

Sonnet 4.6 CHAIN_PROMPT_SONNET.md --effort high
✓ ready
mobile 390sonnet-tuned mobile
desktop 1440sonnet-tuned desktop
HTML size15,343 bytes
<img> refs3
<section>5
Picked variantWest Texas Trade Shop (score 8.0)
Other scores7.0
OPEN LIVE MOCK →https://lead-to-mock-compare.pages.dev/mocks/sonnet-tuned/index.html