Tighten mid-training text, fold VEP example into intro section, collapse data mixture, Safety and dual use section, drop What Carbon unlocks, inline BioReason citation, fix Kimi K3 flag
- Red G laid over the faded reference A (no arrow); tighter input box with the reference line inside
- Smaller main text, tighter cards and connector; less space around the figure
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Variant comparison: outcome icons for every reasoning effort in the model menu and the effort switch
- Right card mirrors Carbon-X: name (opens model menu) + effort, verdict badge, main text
- Main text is the model's answer or the end of its reasoning; bare <answer> blocks dropped
- Explanations under the main text; DNA strip fits on phones
- Table 1: 14.5px numbers, vertically centred
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Sections: cost, DNA and language in one model, openness, Carbon's DNA capabilities,
training recipe (mid- and post-training, collapsible boxes), what Carbon unlocks,
limitations, responsible release
- Figures that exist in carbon-x-blog-figures are copied into figures/ and embedded
same-origin (background matched, frame sized to content)
- Figures missing from the figures Space (Carbon -> Carbon-X animation, architecture
diagram, five-benchmark refusal chart) are draft screenshots with a TMP watermark
- Draft notes and number mismatches marked as placeholders
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>