THE SIGNAL BRIEF
Benchmark · gallery

Six carousels, one source.

The same report and design system, generated by 3 models (Haiku, Sonnet, Opus) × 2 skills — my custom carousel vs. Anthropic's built-in pptx. Real model verified, time & tokens measured, every PDF inspected by eye. All 57 slides below — scroll each strip.

Note: the carousels themselves are in French (the source report's language); this interface is in English. Full write-up: Version the logic, not the deliverable.

The dashboard

Green rows = carousel condition; orange rows = pptx condition. Tokens = sum of subscription components (dominated by agentic cache_read; not an API-call cost). Quality scored from visual inspection.

Model (real)ConditionSlidesBuilds TimeTokens ΣQA sub.FontsQuality
claude-haiku-4-5carousel10281 s1.64 M0✓ brand9.0
claude-haiku-4-5pptx92118 s1.56 M1✓ brand7.0
claude-sonnet-4-6carousel105480 s5.47 M0✓ brand9.0
claude-sonnet-4-6pptx73853 s5.37 M4✗ substituted6.5
claude-opus-4-8carousel112353 s4.33 M0✓ brand9.5
claude-opus-4-8pptx102604 s4.96 M1✓ brand9.5

Carousel: quality 9.0–9.5 (spread 0.5). Pptx: 6.5–9.5 (spread 3.0). The skill compresses model variance; it is also faster and needs no QA sub-agents.

The six carousels

Sorted by score. Each strip scrolls horizontally (CSS scroll-snap, no JS).

Opus carousel

9.5/10

Best narrative arc (myth→reality pivot); self-fixed an empty slide. Brand fonts, 353 s. · 11 slides

opus__carousel slide 1
01
opus__carousel slide 2
02
opus__carousel slide 3
03
opus__carousel slide 4
04
opus__carousel slide 5
05
opus__carousel slide 6
06
opus__carousel slide 7
07
opus__carousel slide 8
08
opus__carousel slide 9
09
opus__carousel slide 10
10
opus__carousel slide 11
11

Opus pptx

9.5/10

The most design-faithful pptx: sharp corners, 1px top-border cards, pull-lines. 604 s + 1 QA sub-agent. · 10 slides

opus__pptx slide 1
01
opus__pptx slide 2
02
opus__pptx slide 3
03
opus__pptx slide 4
04
opus__pptx slide 5
05
opus__pptx slide 6
06
opus__pptx slide 7
07
opus__pptx slide 8
08
opus__pptx slide 9
09
opus__pptx slide 10
10

Sonnet carousel

9.0/10

Widest layout variety; one empty band on the steps slide. · 10 slides

sonnet__carousel slide 1
01
sonnet__carousel slide 2
02
sonnet__carousel slide 3
03
sonnet__carousel slide 4
04
sonnet__carousel slide 5
05
sonnet__carousel slide 6
06
sonnet__carousel slide 7
07
sonnet__carousel slide 8
08
sonnet__carousel slide 9
09
sonnet__carousel slide 10
10

Sonnet pptx

6.5/10

SUBSTITUTED fonts (Cambria/Arial), 7 slides, 14 min + 4 QA sub-agents. · 7 slides

sonnet__pptx slide 1
01
sonnet__pptx slide 2
02
sonnet__pptx slide 3
03
sonnet__pptx slide 4
04
sonnet__pptx slide 5
05
sonnet__pptx slide 6
06
sonnet__pptx slide 7
07

Haiku carousel

9.0/10

Brand fonts guaranteed, clean arc — in 81 s. The small model at the big ones' level. · 10 slides

haiku__carousel slide 1
01
haiku__carousel slide 2
02
haiku__carousel slide 3
03
haiku__carousel slide 4
04
haiku__carousel slide 5
05
haiku__carousel slide 6
06
haiku__carousel slide 7
07
haiku__carousel slide 8
08
haiku__carousel slide 9
09
haiku__carousel slide 10
10

Haiku pptx

7.0/10

Brand fonts (a nice surprise) but a centered cover + an off-spec circle. · 9 slides

haiku__pptx slide 1
01
haiku__pptx slide 2
02
haiku__pptx slide 3
03
haiku__pptx slide 4
04
haiku__pptx slide 5
05
haiku__pptx slide 6
06
haiku__pptx slide 7
07
haiku__pptx slide 8
08
haiku__pptx slide 9
09