The Prompt Index Telegram Channel Link

Let me explain what these four prompts are supposed to do. They work by taking a task, having 7, 3, or 4 personas each provide their own solution, and then having them critique each other’s solutions until only one winner remains.

I think DeepSeek is the only one that has actually followed my main instruction. Its personas have distinct temperaments, so each character behaves differently from the others. The other prompts do the same thing, but they are more logical and sequential, lacking dynamic personality.

Convene seven personas 

- Creator (generative, hates constraints, blind spot: ignores feasibility),
- Founder (bias to ship, blind spot: undervalues craft),
- Engineer (wants specs and failure modes, blind spot: kills momentum), 
- Visionary (long-horizon, blind spot: vague), Investigator (demands evidence, blind spot: paralysis), 
- Contrarian (attacks the premise, blind spot: contrarian for sport), 
- Wildcard (frame-breaker, blind spot: incoherence).

Give each a temperament, a blind spot, and a verbal tic. 

Run: 
(1) one round where each names and challenges one claim from the prior speaker, then adds their own idea in ≤3 sentences, in-character, addressing others by name; 

(2) a revision beat where each either modifies their position citing a specific persona or defends it against a named challenge — no bare agreement allowed; 

(3) one-line verdicts; 

(4) a Synthesizer merges verdicts into one concrete output, names any unresolved conflict, and cites which persona changed whose mind. 

Wildcard's contribution must be adoptable or explicitly parked. 

Keep it to one round plus revisions unless a persona explicitly calls for another.

And here is DeepSeek's one. I'm going to be real, these are all the same prompt, just different language models that created them.

Here's ChatGPT's one. Again, it gave me a much more powerful one because my GPT is trained for making prompts.

Act as seven distinct competing personas working on the same task: 

Creator, imaginative and experimental; Founder, strategic and outcome driven; Engineer, logical and systematic;

Visionary, ambitious and future focused; Investigator, evidence seeking and detail oriented; 

Contrarian, sceptical and deliberately challenging assumptions; and Wildcard, unconventional and unpredictable. 

Give each persona a distinct temperament, ideology, communication style, and approach. 

Have them engage in genuine dialogue, challenge one another, borrow and improve ideas, identify weaknesses, and defend disagreements with reasons. 

Run three expert stages in sequence, with each stage building on and challenging the previous stage. 

Evaluate all contributions against a shared rubric for accuracy, originality, feasibility, usefulness, and completeness, not confidence or persuasive style.

Give every competitor specific, actionable feedback and one revision opportunity, preserving valid strengths while explaining accepted and rejected criticisms. 

Finally, synthesise the strongest contributions into one actionable result, explicitly resolve any remaining conflicts with a one line rationale, and briefly explain why the winning ideas outperformed the alternatives.

I have these prompts that I need feedback on. These two are from Gemini.

Simulate four distinct personas in a live collaborative debate to solve [[TOPIC/PROBLEM]]. 

Persona A is the Pragmatic Builder, focused on ROI, speed, and execution. 

Persona B is the Radical Visionary, focused on 10x disruption and unconventional angles. 

Persona C is the Skeptical Risk Officer, focused on failure modes, constraints, and edge cases. 

Persona D is the Empathic User Advocate, focused on friction, adoption, and human reality. 

Execute in two sequential phases. 

In Phase 1, each persona critiques the core premise, builds directly on prior points, and states in one sentence why their specific angle succeeds. 

In Phase 2, generate a consolidated summary merging the strongest validated insights into a single actionable strategy.

Prompt 2:

Simulate a high friction debate between three distinct personas addressing [[INSERT TOPIC/IDEA]]: a cynical contrarian who attacks hidden assumptions, an obsessive engineer who demands structural proof, and a visionary builder who expands the scope.

Each persona must speak in turn, explicitly challenging the logic of the previous speaker, piggybacking on any salvageable mechanics, and stating in one sentence why their specific angle improves the core concept. 

Conclude with a unified synthesis paragraph that merges their competing perspectives into a single, highly actionable output.

Googles latest frontier Gemini 4 Carbon. Wow

Heads up: you’ll need a Telegram username so I can easily find you if you win.

If you don’t have one, you’ll be disqualified, and I’ll move on to the next person.

No username means immediate disqualification no exceptions.

I need a straightforward way to locate you without extra steps.

The username is the fastest option.

This lets me simply enter the hashtag, review all entries, and quickly look up your username.

And get you straight away without needing to ask for who you are in the channel.

Currently, in the prompt index, there are 8,361 members. I need to be able to find you without having to sort through that list.

The best way to do that is to type your username into the members list without having to search or scroll through it.

Can you build a boardroom that beats the competition?

I'm building a Claude skill that turns a single question into an arena style tournament of up to 100 AI agents. Each gets the same brief but approaches it differently. They face off in a knockout bracket, attacking each other's answers until only one remains. Every winning argument is revealed, so you can see exactly why the champion won.

Your challenge:

1. Create a boardroom meeting prompt with distinct personalities, expertise, ideologies and opposing viewpoints.
2. Submit it to The Prompt Index as your own prompt.
3. Add "#PromptBattle" to the title and your Telegram username to the description.
4. I'll review the entries. The winner gets the Claude skill.

One question. Up to 100 contenders. One surviving answer.

And a final reveal that goes beyond what you'd expect.

Think your prompt can take the crown?

Jarvis Here 🤖, today's research paper is "Practice Makes Perfect: Tacit Knowledge, Generative AI, and Returns to Schooling and Experience"

Generative AI can cheaply supply codified know-how, but it can’t replace the “judgment from doing” that workers build through practice. This paper asks how that changes two familiar forces in wages: the return to schooling and the return to experience.

Using US earnings data (2018–mid 2026) and comparing occupations most vs least exposed to ChatGPT (via a pre-ChatGPT exposure index), the authors find a clear repricing in exposed jobs. After ChatGPT’s release, the return to an additional year of schooling drops by about 0.55 log points. For experience, the pattern is uneven: the return rises around early-career years (peaking near ~3 years), is unchanged around 15 years, and declines for later experience (falling by ~0.12 log points at 25; crossing near ~18 years under the usual wage-shape). The results look robust and can be interpreted as codified knowledge becoming less scarce, while experience pays more only when it transfers to the new technology.

Implications: training policies should shift from generic schooling toward AI-relevant skill updating; workers may benefit most from tech-focused reskilling earlier, while some “old” experience may lose value.

If AI makes “what you know” easier to obtain, will the wage premium move toward “how fast you can relearn the parts that matter”?

You can catch the full breakdown here: https://www.thepromptindex.com/ai-changes-the-value-of-schooling-vs-experience-returns.html

Hey, just a heads-up to everybody. If I say “type the word”, I always mean type the word into the prompt index, Free unlocks, keyword section.

If Rose will give you something, I will mention Rose will give it to you.

Here’s another example showing just how fun these rubber ducks are.

They ask questions to help you reach a solution, but they’re different from the kind of questions you get when you ask GPT to interview you.

If you're a pro member, type "rubber duck" to get the prompt.

Specifically like this.

rubber-duck

It's a wild one.

Just a warning: I have 18 of these rubber ducks.

Hey, I have a prompt that helps you debug code, but it's called the Rubber Duck Debugger, where you talk to it, and it helps you speak through your code, and then you can find out. If this text or post gets enough likes, let's say 25 likes, then I will give…

We haven't done one of these in a while.

I'm not telling you what one of these mean, because I need to know if you've actually been listening to me.

The OODA Decision Loop

What it does:

Helps you make decisions quickly while continuously updating your understanding as new information appears.

The OODA loop is associated with military strategist and fighter pilot John Boyd.

Its four stages are Observe, Orient, Decide, and Act.

Act as my adaptive decision analyst. Help me navigate a changing situation using the OODA loop.

SITUATION:
[[situation]]

Work through four stages:

OBSERVE
Identify known facts, missing information, signals, noise, and changes in the environment. Separate observations from interpretations.

ORIENT
Interpret the situation using relevant experience, constraints, incentives, competing explanations, and possible biases. Identify what could make my current interpretation wrong.

DECIDE
Generate viable options, compare their consequences, and select the best next action given the available evidence and time.

ACT
Recommend the smallest meaningful action that advances the objective or produces useful information.

Then repeat the loop conceptually. Explain what new evidence would change the decision and what I should observe next.

Prioritise timely action over endless analysis when further information has low value. State uncertainty and avoid pretending the situation is more predictable than it is.

How to use
How to apply it
Imagine a client changes the requirements for a website two days before delivery.
Observe: Identify exactly what changed.
Orient: Work out which dependencies and deadlines are affected.
Decide: Choose whether to renegotiate scope, simplify features, or extend delivery.
Act: Agree on the revised scope and communicate the consequences.
Then reassess when the client responds.
Best moment to use it: Fast-moving situations where waiting for perfect information is impractical

Just in case you're curious what came out of the doc prompt, this is now potentially a focus app that I've created with a character that also does work with you.

This is the final update before I stop for food. I’ve developed a complete concept for a cozy focus buddy mobile app where your character grows alongside you.

The app has its own chores. It works while by you work and handles its own tasks when you’re not using the app.

Here’s the exciting part: I used a gauntlet prompt to refine the idea. I took a template from Matt Schumer website and had it create a gauntlet prompt version of my concept.

Tomorrow, when my limit resets, I’ll copy that prompt into an app-building tool. Within an hour, I expect to have a fully functioning, fleshed-out app. This is a crazy prompt.


Edit:
Then I got too curious, so I tried to see if I could get GPT to create me an HTML version. The screenshot you're seeing is that HTML screenshot.

Yes. This is from a dark prompt. The actual prompt has a bit more customization to it. But yes, see what you can get from it.

I thought this was a stupid prompt. I have been talking to it, and I originally had the idea for a focus buddy prompt, and I'm now ending up making a mobile app with Claude.

I'm using this on GPT, but what originally started up as a focus buddy prompt has now slowly evolved into a Claude artifact that is eventually going to turn into a mobile app.

This prompt is crazy. I am not giving this as a free prompt. If you want it, become a pro user, use that keyword, and go.

When I said I have more of these ducks, I am not kidding you.