← The AI Hype Audit — all 645 verdicts

PARTLY

Partly true: giving three AI models the same '3D Earth dashboard' prompt is a fair way to compare them, but one run of one prompt proves little, and 'GPT-6.2 Sol' had no official release we could find

Facebook · Nov 2026

The claim'I gave 3 AI models the exact same prompt: build an interactive 3D Earth dashboard. Gemini 4 Hydrogen. GPT-6.2 Sol. Claude Fable 5.2. Same task. Three completely different results... This is exactly why testing AI models on real builds matters more than reading benchmarks.' Asks viewers to vote 1, 2 or 3 and to subscribe to the page's paid content.

The idea is sound and the reel does show three different working globes. One has a dark mission-control look, one is a polished atlas with a big title, one goes for realism with clouds, glow and city detail. Testing models on a real task instead of trusting a leaderboard is good advice, and the caption does not crown a winner, which is to its credit. What the reel cannot support is any conclusion about which model is better. It is one prompt, run once per model, with no prompt text shown, no settings, no time or cost, and no word on how many attempts were thrown away. These models give different output every run, so a second try could reverse the ranking. The labels need care too. Gemini 4 Hydrogen is real; Google released it on Nov 12, 2026. Claude Fable 5.2 is real; Anthropic released it Nov 9. But reports dated Nov 9 and 10 said OpenAI had published no announcement, model card or API listing for a 'GPT-6.2 Sol'; the latest confirmed model was GPT-6.1 Sol. So the middle panel is either mislabeled or an unannounced preview we cannot confirm. The vote request and the paid subscription link are what the post is for.

What holds up

  • Contact sheet viewed: three labeled globe dashboards shown side by side for the whole 43 seconds; no narration.
  • Gemini 4 Hydrogen confirmed released Nov 12, 2026 (Google's blog post and press coverage).
  • Searched for GPT-6.2 Sol: no official release found; Nov 9 to 10 reports describe it as rumor and screenshots only.
  • Claude Fable 5.2 release (Nov 9, 2026) was verified from Anthropic's docs in last night's batch.

What doesn’t

  • One prompt, one run each, no prompt or settings shown.
  • 'GPT-6.2 Sol' could not be confirmed as a released model.
  • No cost or time figures, so 'the gap' is a matter of taste.
  • Ends on a vote-in-comments ask and a paid page subscription.

The catch

A nice-looking demo and a fair point about testing on real work, but a single unshown prompt cannot rank models, and one of the three names does not check out.

How to actually do it

  • Pick one task you actually need done, such as a quote calculator or a landing page.
  • Give the identical prompt to two or three models, three runs each, and note time, cost and how much fixing each needed.
  • Choose on your own results, and re-test every few months because the rankings move.

Good method, thin evidence. Run the same test on your own work before believing any side-by-side reel, including this one.

Confidence
Medium
Posted by
an AI tips page (about 3,000 plays); reel sent to Buddy Sun Nov 15, 2026, 7:38 AM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.

© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.