phoenixscore

Drop your post in the box — like you would on X — and see if it flies. Real xAI Phoenix model + an algorithm-calibrated judge.

You get a 0–100 grade, the 19 predicted actions, and one rewrite — re-scored to prove it actually helps.

· @
280

Credit packs

One key. No account. Pay once, score instantly.

$5

10 runs

$0.50 / run

$20

50 runs

$0.40 / run

$50

150 runs

$0.33 / run

Key shown once after payment. Lost it? Recover.

Honest FAQ

Does this run X's real algorithm?

It runs the open-source xAI Phoenix mini checkpoint — not X's production weights. Phoenix is collaborative: it scores on hashed post/ author IDs & engagement history, so it does not read your text. We run the real model for the behavioral prior and an LLM (calibrated to X's published algorithm) to estimate the text-driven part. Predictions are directional, not exact.

What does the @handle do?

It's an optional, coarse author signal — Phoenix keys on hashed account identity, heavily collided across millions of accounts. It nudges the behavioral prior; it is not "we analyzed your reputation". Leave it blank for a generic author.

What's the "grox" panel?

xAI's grox content classifiers ship as LLM prompts with no weights. We faithfully re-host that design on an LLM. It's a grox-style screen, not grox output.

Are the hashtag/link warnings "the algorithm"?

No. Those are community heuristics — no such rule exists in the 2026 algorithm code. They're flagged as guidance, not law.