← The AI Hype Audit — all 314 verdicts
BS Straight BS: 'Claude Prompt That Should Be Banned #183' is a composited fake Claude window (two different models in one screenshot, an 'Adaptive' label that does not exist in claude.ai) wrapped around a recycled ChatGPT prompt, posted by a numbered follow-farm
The claim'Claude Prompt That Should Be Banned #183.' 'These Claude prompts are for anyone who wants to use AI to think at a level most people never reach, make better decisions faster, and create unfair leverage.' The prompt shown in a mocked-up Claude window: 'My biggest fear isn't failure. It's discovering twenty years from now that I never became the person I could have been. I have one skill, ChatGPT, £500, and unlimited willingness to learn. Build the shortest path toward ownership instead of employment. Reject every strategy that depends on luck or hype. Keep only those that compound through skill, assets, and leverage.' 'Follow for more AI prompts that feel illegal to know.'
The screenshot is not a Claude session. The mock-up's header reads 'Sonnet 4.6 Adaptive' while the composer chip in the same frame reads 'Sonnet 5 Medium'. Both model names are real: Anthropic shipped Sonnet 4.6 on Feb 17, 2026 and Sonnet 5 on Jun 30, 2026, and Sonnet 4.6 still sits under 'More models' in claude.ai. But a real window shows one selected model, once, next to the send button, in the format 'Model Effort' (Fable 5.1 Medium, Sonnet 4.6 Low). There is no model name in the chat header, and 'Adaptive' is not a label anywhere in the app; the effort options are Low, Medium, High and Max plus a 'Thinking' toggle. 'Adaptive' is an API-documentation term for adaptive thinking, which is the tell of someone typing over a template. Two pickers disagreeing inside one window means a composite, not a capture. The prompt itself says 'I have one skill, ChatGPT' inside a 'Claude prompt', which is a ChatGPT-era text re-skinned for the Claude hashtag. Nothing about it is banned or bannable; it is a career question with a budget and a preference for compounding strategies, and Claude will answer it as it would any career question. 'Prompts that feel illegal to know' and 'should be banned' are a documented engagement-bait format across Medium, Threads, TikTok and Facebook. The numbering (#183) tells you the business model: volume, follows, and whatever the profile links to. The reel is a login-walled Facebook page, so the exact link-in-bio could not be verified.
What holds up
- Frames read from the reel: header 'Sonnet 4.6 Adaptive', composer chip 'Sonnet 5 Medium', title card 'Claude Prompt That Should Be Banned #183', prompt text contains 'I have one skill, ChatGPT'. Two different selected models in one window.
- Live claude.ai checked Sep 18, 2026 on a Max account: the only model label is the chip next to the send button in 'Model Effort' format ('Fable 5.1 Medium'; switching to Sonnet 4.6 shows 'Sonnet 4.6 Low'). Effort options are Low (Default), Medium, High, Max plus a 'Thinking' toggle. No 'Adaptive' label exists; no model name appears in the chat header.
- Sonnet 4.6 is real (released Feb 17, 2026; Anthropic announcement and CNBC) and still selectable under 'More models'. Sonnet 5 is real (released Jun 30, 2026; TechCrunch, Anthropic) and replaced 4.6 as the Free/Pro default. 'Adaptive' is a term from Anthropic's API docs (adaptive thinking), not a claude.ai UI string.
- 'Prompts that feel illegal to know' / 'banned prompts' is a recognized engagement-bait format (Medium, Threads, TikTok, LinkedIn, Gumroad prompt packs); the phrasing markets ordinary prompts as forbidden. The exact prompt text does not surface in web search, so it could not be traced to a first source; the 'ChatGPT' line inside it is the recycling evidence.
- Facebook reel page is a login wall to a fetcher, so the page's link-in-bio and any paid product could not be confirmed; the caption's only stated ask is 'Follow'. Small poster, described generically.
What doesn’t
- Fabricated UI: two model pickers showing two different models in one 'screenshot', plus a label ('Adaptive') that does not exist in claude.ai.
- Recycled prompt: 'I have one skill, ChatGPT' inside a 'Claude prompt' is a copy-paste from an earlier ChatGPT post with the hashtag swapped.
- 'Should be banned' / 'feel illegal to know' is the hook; nothing in the prompt is restricted, and the answer is the same you get from asking plainly.
- Numbered series (#183) and 'Follow for more' is a follow-farm; the value is the follower count, not the prompt.
The catch
The prompt is not dangerous and it is not special. It is a decent opening line for a career conversation with an AI, dressed in a forged screenshot so it looks like a leak. If you paste it as written, you get a generic list (freelance, productize a service, buy a small asset) because the prompt gives Claude nothing real to work with: no actual skill, no location, no hours available, no risk tolerance, no deadline. The 'twenty years from now' emotional framing does nothing for the model; the constraints do.
How to actually do it
- Open claude.ai (Sonnet 5 or Opus 5 is fine), pick High effort, and write the real version: your actual skill (not 'ChatGPT'), your city, cash you can lose, hours per week outside your job, obligations, and the date by which you want the first paying customer. Ask Claude to interview you with 8 to 10 questions before it proposes anything.
- Then ask for three ownership paths ranked by how fast each compounds through skill, assets and leverage, each with the first 30 days broken into weekly actions, the cash it burns, and the one metric that tells you to kill it or double down.
- Take the winner and have Claude turn week one into tomorrow's to-do list, including the first five people to contact and the exact message to send them. Come back in a week with what happened and let it revise the plan against real results, not the prompt's fantasy.
Both model names in the mock-up are real, which is exactly why the frame gives itself away: claude.ai shows one model, once, as 'Model Effort' by the send button, never two, and never the word 'Adaptive'. The prompt underneath is a re-skinned ChatGPT post that Claude will answer about as well as any vague career question. The honest route is a specific, multi-turn interview with real numbers, and Claude does that well.
- Confidence
- High
- Posted by
- An AI-prompts page that numbers its posts (this one is #183); 269 reactions, 27 shares on the reel; sent to the self-thread Fri Sep 18, 2026, 12:37 AM
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.