Sullivan Signature Homes
Rebuild progress · generated Oct 19, 2026 at 22:14 · internal, not for the client
This is mid-run, not a deliverable. Brand candidates are still looping and will keep changing. Codex has not yet preferred any direction over the reference. Mockups, the SEO package, the lead magnet and both site versions have not been built. Nothing here has been seen by Sullivan.
The three surviving directions
Each direction is shown as a website first screen at 1440 and 390 (the like-for-like test against the reference), then its brand board. Marks are vector SVG generated from code. Boards use substitute typefaces where the specified face is not installed; each board prints its substitution.
High-contrast serif wordmark LIGHT
Rebuilt as vector from the client-supplied raster. Ranked first of five in the blind cull. This is the direction allowed to depart from the bold wordmark weight the client asked to keep at kickoff, shown so the trade-off can be judged.
Maroon #6E1423 and navy #16233D BOLD
Previously presented to the client. Holds the kickoff spec: bold serif caps SULLIVAN, a double rule, and a SIGNATURE HOMES subline.
Black #1A1A18 and tan #A9824C BOLD
Matches his existing signs and cards. Holds the kickoff spec: bold serif caps SULLIVAN, a double rule, and a SIGNATURE HOMES subline.
Dropped in the cull
Both were rebuilt as vector first, then judged blind in one batched Codex call. The reasons are the critic's own words.
Script wordmark DROPPED
The script and gold swoosh introduce decorative flourish that reads more like a boutique service business than an elevated architectural homebuilder.
Burgundy SH crest DROPPED
The crest, laurels, and intertwined SH create Victorian kitsch while collapsing into an indistinct shape at the actual 16px size.
The bar they are judged against
mcalpinehouse.com first screen at 1440, recaptured after the first screenshot turned out to be a defective 800x479 crop.
Every round so far
A critic never sees the builder's reasoning. It opens the real rendered artifact, puts it beside the reference unlabeled in random order, and makes a forced binary pick.
| Piece | Round | Item | Verdict | Judge |
|---|---|---|---|---|
| 1. Brand directions | Cull | A to E | Ranked C, E, D, A, B. Drop A and B | Codex |
| 1. Brand directions | Round 1 | C (best now v1) | v1 replaced incumbent; vs mcalpine: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 1 | E (best now v1) | v1 replaced incumbent; vs mcalpine: ours preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 1 | D (best now v1) | v1 replaced incumbent; vs mcalpine: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 2 | C (best now v2) | v2 replaced incumbent; vs mcalpine: ours preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 2 | D (best now v1) | incumbent kept; vs mcalpine: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 3 | D (best now v1) | incumbent kept; vs mcalpine: reference preferred; RELEASED after two rounds with no replacement | Sonnet (fresh context) |
| 1. Brand directions | Final head-to-head | C (best v2) | reference preferred | Codex |
| 1. Brand directions | Final head-to-head | E (best v1) | reference preferred | Codex |
| 1. Brand directions | Final head-to-head | D (best v1, released) | reference preferred | Codex |
| 1. Brand directions | Lead ruling | C, D, E | Loop resumed from round 4. D reopened: its release was judged against the defective 800x479 crop. Reference recaptured as bar/mcalpine-1440.png. C is LIGHT, D and E are BOLD (client spec). Company NAP keeps 185. | Lead (Opus) per user ruling |
| 1. Brand directions | Round 4 | C LIGHT (best now v4) | v4 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 4 | E BOLD (best now v4) | v4 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 4 | D BOLD (best now v4) | v4 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 5 | C LIGHT (best now v5) | v5 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 5 | E BOLD (best now v5) | v5 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 5 | D BOLD (best now v5) | v5 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 6 | C LIGHT (best now v6) | v6 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 6 | E BOLD (best now v5) | incumbent kept; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 6 | D BOLD (best now v5) | incumbent kept; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 7 | C LIGHT (best now v7) | v7 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 7 | E BOLD (best now v5) | incumbent kept; vs mcalpine-1440: reference preferred; RELEASED after two rounds with no replacement | Sonnet (fresh context) |
| 1. Brand directions | Round 7 | D BOLD (best now v7) | v7 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 8 | C LIGHT (best now v8) | v8 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 8 | D BOLD (best now v8) | v8 replaced incumbent; vs mcalpine-1440: ours preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 9 | C LIGHT (best now v9) | v9 replaced incumbent; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 10 | C LIGHT (best now v9) | incumbent kept; vs mcalpine-1440: reference preferred | Sonnet (fresh context) |
| 1. Brand directions | Round 11 | C LIGHT (best now v9) | incumbent kept; vs mcalpine-1440: reference preferred; RELEASED after two rounds with no replacement | Sonnet (fresh context) |
| 1. Brand directions | Final head-to-head 2 | C LIGHT (best v9, released) | reference preferred | Codex |
| 1. Brand directions | Final head-to-head 2 | E BOLD (best v5, released) | reference preferred | Codex |
| 1. Brand directions | Final head-to-head 2 | D BOLD (best v8) | reference preferred (Sonnet had preferred ours in round 8) | Codex |
| 3. Brand copy | Round 1 | deck v1 (best v1) | replaced incumbent; 6 untraceable of 165, 0 em-dashes, 0 missing markers, 7 rule breaks | Sonnet (fresh context) |
| 3. Brand copy | Round 2 | deck v2 (best v2) | replaced incumbent; 3 untraceable of 112, 0 em-dashes, 0 missing markers, 7 rule breaks | Sonnet (fresh context) |
| 3. Brand copy | Round 3 | deck v3 (best v2) | incumbent kept; 3 untraceable of 120, 0 em-dashes, 0 missing markers, 7 rule breaks | Sonnet (fresh context) |
| 1. Brand directions (first screen) | Hero round 1 | C LIGHT (best v1) | first version; vs mcalpine: 1440px ours, 390px ours | Sonnet (fresh context) |
| 1. Brand directions (first screen) | Hero round 1 | D BOLD (best v1) | first version; vs mcalpine: 1440px reference, 390px ours | Sonnet (fresh context) |
| 1. Brand directions (first screen) | Hero round 1 | E BOLD (best v1) | first version; vs mcalpine: 1440px ours, 390px ours | Sonnet (fresh context) |
| 3. Brand copy | Round 4 | deck v4 (best v4) | replaced incumbent; 0 untraceable of 135, 0 em-dashes, 0 missing markers, 5 rule breaks | Sonnet (fresh context) |
| 3. Brand copy | Round 5 | deck v5 (best v5) | replaced incumbent; 3 untraceable of 125, 0 em-dashes, 0 missing markers, 0 rule breaks | Sonnet (fresh context) |
| 1. Brand directions (first screen) | Hero round 2 | D BOLD (best v2) | v2 replaced incumbent (won both widths); vs mcalpine: 1440px ours, 390px ours | Sonnet (fresh context) |
| 3. Brand copy | Round 6 | deck v6 (best v5) | incumbent kept; 2 untraceable of 150, 0 em-dashes, 0 missing markers, 2 rule breaks | Sonnet (fresh context) |
| 1. Brand directions (first screen) | Codex final (first screen) | C at 1440px | reference preferred | Codex |
| 1. Brand directions (first screen) | Codex final (first screen) | C at 390px | reference preferred | Codex |
| 1. Brand directions (first screen) | Codex final (first screen) | D at 1440px | reference preferred | Codex |
| 1. Brand directions (first screen) | Codex final (first screen) | D at 390px | reference preferred | Codex |
| 1. Brand directions (first screen) | Codex final (first screen) | E at 1440px | reference preferred | Codex |
| 1. Brand directions (first screen) | Codex final (first screen) | E at 390px | reference preferred | Codex |
| 1. Brand directions (first screen) | Lead note | C, D, E | Sonnet preferred all three by round 2; Codex preferred reference on all six pairs. Next run: shared quiet-photo scout, stricter critic, two consecutive Sonnet wins required before Codex. | Lead (Opus) |
| 3. Brand copy | Codex trace audit | deck v5 | FAIL on 3 sentences, 0 em-dashes. Same 3 the Sonnet critic named. | Codex |
| 3. Brand copy | Mechanical fix | copy/final-deck (from v5) | The 3 Codex-flagged sentences removed (4 lines, one heading appeared twice). 0 em-dashes, 10 CLIENT-BLOCKED markers. Every remaining sentence was traced by the Codex audit. Released: no further rounds. | Haiku (mechanical), verified by lead diff |
Done
- Brand copy. Copy deck extracted only from permitted sources; every sentence traced by a blind Codex audit; 10 visible CLIENT-BLOCKED markers where content is missing.
- Testimonials recovered verbatim from the live old site. The versions currently in Sanity are rewritten paraphrases and should not be published.
Not started or blocked
- Mockups and collateral: blocked, no fal key
- SEO package: blocked, no DataForSEO credentials
- Lake Conroe lead magnet: waits on the brand winner
- Both site versions on Astro, Sanity and Cloudflare: wait on the brand winner
- Form delivery proof: blocked, GHL_MARATHON_PIT and GHL_MARATHON_LOCATION_ID unset
Open, needs you or the client
- Folders 3 and 5. Folder 5 is Holly Hollow and folder 3 is Southshore, from the photographer's file sequence. A visual exterior match against a labeled photo is still outstanding.
- Sanity data. Testimonials are paraphrased, siteSettings address reads '185 Superior Road' instead of '185 Superior Rd', and there is a placeholder team entry. Needs a token and a go-ahead to correct.
- Client-blocked content: family photo, individual bios, process detail, exact Master Electrician wording, his outline for the Lake Conroe guide, whether a Google Business Profile exists.
Marathon Creative · builders on Sonnet, critics blind on Codex and Sonnet, lead on Opus