Video wird geladen...
Video konnte nicht geladen werden
CODEX SKILL TO BRUTALLY TEST ANY STARTUP IDEA! Most startup ideas sound good. This Codex skill tells you why they probably won’t work. Just give Codex your idea and it pressure-tests it for you -> finds the core assumption -> exposes fatal flaws -> checks if the problem is... show more
513,769 Aufrufe • vor 4 Monaten •via X (Twitter)
74 Kommentare

this is the repo👇

matbe

Bookmarked it. This is actually a good way to stop wating time on something which has low probabilities of working out. What i already do is use my manual frameworks to filter out ideas. This would be an interesting addition to that.

@nishancodes Thank you, Nishan. Yes, I saw it exactly the same way, which is why I created it. If you use it, let me know how it goes!

@nishancodes Are u a founder yourself?

@nishancodes yes

the problem is that codex (or any ai) will most likely make up the “fatal flaws” just because skill says it

nahhh

Best way to get other people to give you their best ideas 🤣

no?

ha just kidding - but that would be an epic move

maybe time to build a codex skill marketplace :)

why not..

Is there a back door that sends you all the good ideas ?

nope

That’s good. I’ll check it out.

installing that npx command just to get roasted by an open source script finally a tool that admits my startup idea is garbage before i

AHHAHAHAHA if u tty it let me know what you think!

Using it to come up with ideas is the real hack

reallll

Very

I love these pressure tests. So much more useful than the classic llm bias « yes sir you’re totally right »

I was tired of him always saying I was right, so why not strain my mind for better ideas?

I can confirm that it works with claude code as well :)

really good!

Honest note on the pressure-test skill: it's designed to find the narrowest defensible wedge for pre-seed. That's useful. But it doesn't tell you how to keep the bigger vision while pitching the wedge - that's a structural blind spot of the skill, not a fact about your company.

Okay, I'll try to improve. Thanks!

Thank you! this is a great skill - lots of amazing potential that will emerge from here

I love asking our Overlords to roast us. It burns, but in a good way 😅

AHAHHAHAH

This is really good. I got an 8/10. Codex: "Why not higher yet? Because this is still doctrine unless it becomes product..." I wont list the what Codex bulleted for now 😊

Aoh nice

Interesting. But it seems unlikely that general prompting will give the LLM more info to do proper analysis. It might tell you things that were already available out there but in the new ideas era, a lot of startups are proposing ideas that haven’t been tested before. So, interesting to see how often LLMs will give good analysis for a brand new idea

real

cooked

thanks broda

strunz rispunnm ngopp a whatsapp

That's the Mom Test applied to an LLM’s interpretation of founders’ fantasies. A triple layer of epistemic noise. It generates highly confident prescriptive outputs with zero structural verification. This is not a pressure test of the idea — it’s a pressure test of its narrative.

ok!

I would give more details of my analysis, but sorry—lately many people have been releasing products based on my ideas and passing them off as their own. GPT often presents wishful thinking as fact.

Bro this is just their way to scan for new feature and product ideas, you guys don’t get it yet

real

Awesome! This is great

thabks bro

This is actually very useful - saves time by showing what’s wrong before you start. Every startup idea should go through this kind of real test.

💪🏻

you da life saver man

AHHAAH

Does it work for Claude?

maybe idk

gotta give this a try

let me know what u thinks looks so cool

yes

half of startup twitter needs this immediately

real

Thanks for sharing this!

Bro Just killed my startup idea with a codex skill 😂

Codex as a devil's advocate is underrated. Best founders I've seen already know their fatal flaw—this just forces honesty faster.

Too much obsession on the idea and too little on execution. Look at literally all the AI models out there (Claude, Kimi, Deepseek, etc). None of them started with "I have this crazy idea". They all started with "I can do this better/cheaper than OpenAI".

I'll take that, thank you very much!

let me know what u think?

Ok yeah got a couple big angles here in Houston I’ll pretend I’ve finally learned my lesson (I have not)

Wow

do u like it?

codex finding the fatal flaw is useful but founders usually already know the flaw and hope no one notices. the real test is whether you can explain why it won't kill you

Thanks for the sharing. Will help?

cool demo but without real market data it just hallucinates confidence, not insight

mmh...I know but in the end it's just a basic assessment

This is sweet. Building something similar to this that goes beyond the skill, would be interested if you had any thoughts on where this still falls short? An issue I find is that everything is infinitely critique-able and the model will find flaws that exist only in prose, not reality, after enough loops.

@llmgram

Does this work in a retroactive way, in terms of projects already in flight not just in ideation?

nope

How about another codex skill for investing, generating leads, etc., so you fully found a startup and codex would be the co-founder.

👀👀
