Skip to content

feat: consume the published brand-numbers artifact - #14

Merged
VickyXAI merged 2 commits into
mainfrom
feat/brand-numbers-sync
Jul 27, 2026
Merged

feat: consume the published brand-numbers artifact#14
VickyXAI merged 2 commits into
mainfrom
feat/brand-numbers-sync

Conversation

@VickyXAI

Copy link
Copy Markdown
Contributor

Wires this repo into the brand-numbers system from blockrun#288.

What was stale

Claim Was Published
free tier (4 places) 7 models 8
model count (4 places) 50+ / 60+ 66
savings (3 places) 92% 87%, with a stated baseline

The skill was the worst of it

skills/xclawrouter/SKILL.md is loaded into an agent's context. It claimed "saves 67%" — a figure that appears nowhere else — next to a tier table describing a routing configuration that no longer exists:

Gemini Flash · DeepSeek Chat · Claude Opus ($75/M) · o3

Opus 5 is $5/$25. Rewritten from the router config and the live catalog. The per-tier percentages are dropped rather than restated: they were measured against a baseline nobody wrote down, and replacing one unverifiable number with another is not a fix.

docs/ is two things, treated as two

  • Evergreen posts carry live claims → bound to the artifact
  • Dated engineering posts are measurements → keep their numbers, gain a note

"We Benchmarked 39 AI Models" ran 39 models on March 16. Rewriting that to 66 would misrepresent it, not update it.

11-free-ai-models-zero-cost-blockrun.md has the count in its title and its URL. The tier is 8 now; renaming breaks every link, so it takes the same note — which also explains why it moved.

Left alone deliberately

  • 5 models for image generation — this repo's own support surface. I could not verify it against the gateway catalog, and guessing is the exact failure this system exists to stop. Worth a separate look.
  • up to 97% is compression ratio · 3 models is the fallback cap · 2 chains is L1/L2 · up to 30% is Anthropic's promo credit — same words, different subjects.

Separate from ClawRouter

Own commit, own PR, own content. Nothing was copied between the two repos — the identical corrections are to identical facts about the gateway, not shared marketing copy.

1bcMax added 2 commits July 27, 2026 16:07
Numbers here were hand-maintained and stale in every direction: a free tier of
7 (published: 8), 50+ and 60+ models against a catalog of 66, and 92% savings
against no stated baseline.

The skill was the worst of it. skills/xclawrouter/SKILL.md is loaded into an
agent's context and claimed "saves 67%" — a figure that appears nowhere else —
next to a tier table describing a routing configuration that no longer exists.
It named Gemini Flash, DeepSeek Chat, Claude Opus and o3 as the four primaries
and priced Opus at $75/M, which is wrong by any reading. Rewritten from the
router config and the live catalog; the per-tier percentages are dropped rather
than restated, because they were measured against a baseline nobody wrote down.

docs/ is two things and is treated as two. Evergreen posts carry live claims and
are bound to the artifact. Dated engineering posts are measurements — "We
Benchmarked 39 AI Models" ran 39 models on March 16, and rewriting that would
misrepresent it rather than update it. Those keep their numbers and gain a note
saying so.

Left alone deliberately: "5 models" for image generation is this repo's own
support surface and I could not verify it from the gateway catalog; guessing is
the exact failure this system exists to stop. "up to 97%" is compression ratio,
"3 models" is the fallback cap, "2 chains" is L1/L2, "up to 30%" is Anthropic's
promo credit — same words, different subjects.
The script is copied byte-for-byte from BlockRunAI/blockrun:brand/. Letting
prettier reformat it here would fork this copy from the source and from every
other repo carrying it — and blockrun CI compares them, so the fork would then
fail there.
@VickyXAI
VickyXAI merged commit e540b06 into main Jul 27, 2026
4 checks passed
@VickyXAI
VickyXAI deleted the feat/brand-numbers-sync branch July 27, 2026 23:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant