@nickrackley

Founder Combat Vet. Banned by OpenAI.

[@nickrackley @nick_rackley]

@StephNass Best explanation I’ve seen as to why ideas are worthless!

[@nickrackley @nick_rackley]

I got one of these from him and said “yes I’d like a refund” and never heard a response or got a refund.

I thought he made an ai of himself and it was malfunctioning, either way it was a really really bad look. Being bs’d by a customer service thing, by a CEO it’s completely different. I still dislike cursor because of the audicty to basically lie to canceling customers (although it may have just been he was busy and made a mistake).

[@nickrackley @nick_rackley]

@ibocodes 100% correct. If you’re even using the word “ai slop” it indicates you aren’t serious about conversions.

[@nickrackley @nick_rackley]

Did you know if you dent a rental car

You don’t have to pay for it?

Hertz's problem = idea 4/18👇🏻

💡 Idea (4/18): "Dent Detector"

I learned that rental car companies basically have no way to force you to pay if you damage a car while you’re renting it

The pre/post inspections are all for show

If they take you to court all you have to do is show that the employee who conducted your inspection failed to notice a dent in ANY car over their ENTIRE career to have the case thrown out

I learned this in the Fed Tech Defense Innovation Accelerator, which assigned me a dent detecting grid technology originally developed for tanks to try to find a commercial use for

We did the “Lean Startup” methodology taught by I Corps, and after about 40 customer interviews, the rental car industry clearly had a big problem our tech could fix

My cofounders didn’t seem particularly interested however, and I had a gut feeling that the technology wasn’t really the best one to fix that particular problem

This was pre-AI, but camera tracking technology was already getting decent and seemed like it was probably a better fundamental approach

So we let it die on the vine, and that doesn't seem like a bad decision in hindsight

⛔ Why the idea failed:

This was a classic case of a “Solution In Search of a Problem” SISP, this time in the extreme form as it was by design

Usually, SISP’s fail because you invent a problem for them to solve. This was a rare case where the SISP actually found a legitimate problem that it could theoretically solve, but even then, it clearly wasn’t the best match

✅ What I learned:

1.Lean Startup talking to 40 customers makes a ton of sense for “hard tech” – we got pinpoint requirements on a MASSIVE problem for a big customer, which would have been critical if we set out to make a B2B hardware product

2.Lean Startup doesn’t make as much sense for software/easy to launch startups, as 40 interviews takes a ton of time to prospect and source. You could very easily (especially now with AI) could spend more time getting requirements than it would take to literally get an MVP in somebodies hands

3.I should have learned not to do SISP ideas and only do customer problem driven ideas, but as you will see in the rest of these ideas... I am not that smart...

👉 Takeaway: Don’t do SISP ideas. Just don’t. Start with a customer problem and find the right solution.

[@nickrackley @nick_rackley]

@BigDickBull69 Yes! F that Reddit “I’m smart but the world is stupid” nonsense!

[@nickrackley @nick_rackley]

@sflorimm You can build an mvp in 1 day for $34 with vibe otter fyi

[@nickrackley @nick_rackley]

@signulll Vibe Otter

[@nickrackley @nick_rackley]

Don't you wish you could know

How to always get that last parking spot, table, etc?

Here’s idea 3/18👇🏻

💡 Idea (3/18): "Scavenger Hunt"

I thought it would be cool to make an app that would be like "an Yber for information"

You could pay a small amount of money to get a short video clip of a parking lot, restaurant, gym, bar, wherever you want to know if there’s a spot remaining for you at a more granular level than google saying somewhere is “busier than normal” is able to provide

A user of the app that’s near the location could then just take a short video clip of what you ask them to and receive a small amount of money in payment

I made a mock website and drove traffic to it with google ads, and the idea seemed to do well, getting some attempted app downloads while advertising on obviously unrelated keywords (like gyms, restaurants near me)

It seemed promising enough I decided to try to build it, and quickly realized I wasn’t going to learn app development any time soon and I tried to convince some friends who knew how to code, but they lost interest pretty quick.

I considered using Fiverr to build it, but got sketched out by the cost and expense of future payments to modify it

I kind of just drifted away from the idea after that – even beyond the build, I had no clue how to actually start the idea in the sense you need to convince information purchasers and information providers to join and use the app at the same time

⛔ Why the idea failed:

Pretty simply, I couldn’t build it, and I couldn’t find anybody else who wanted to. I think a lot of startups die this way.

The prospect of standing up a “two-sided network” product is really intimidating to most people

✅ What I learned:

I think this experience made me realize it may be just as hard to convince people of your ideas as it would be to learn to build them yourself. This probably wasn’t true back then, but I think with AI it really is true now

Build it yourself, or find a friend, im REALLY glad I didnt use gig workers. You need to change your app every day in a startup, hiring it out is completely unacceptable it turns out

In case you’re stuck on a “two-sided network idea”, a really good book I’ve come across that explains how to start “two-sided network” ideas is “The Cold Start Probl

👉 Takeaway: If you don’t have the people to come up with, build, and sell an idea, you clearly aren’t going anywhere

Any other thoughts on this idea?

[@nickrackley @nick_rackley]

@pcshipp I really think a lot of people lie at this point.

[@nickrackley @nick_rackley]

@marcrandolph I always wondered if you used AI to write, and now I’m beginning to wonder if it was the AI who was using you to write.

[@nickrackley @nick_rackley]

@yrubnab @Novarcharesk3 @DudeWhoInvests The above + the ability to back out of a deep analysis and ask “does this make sense”. The ability to see the forest once you’ve walked through a bunch of trees.

[@nickrackley @nick_rackley]

Dude it’s not my product and I don’t care what a chart says. I care about how effectively I can make complex software with them, and the difference between frontier to open is lambo to civic. If you’re citing a chart you clearly don’t even use these things so why are you arguing lol?

[@nickrackley @nick_rackley]

That’s predicated on the notion people go to college to learn, which we all know is a very small minority of students.

Degrees are primarily a source of massive cultural capital well in excess of their tuitions. The fact people learn things is primarily to create a facade of legitimacy in the degree beyond balatent buying of cultural capital which would diminish the capital itself.

This is why so few people work in fields relevant to their degrees, why so many people cheat (cheating is idiotic if you want to learn your subject).

[@nickrackley @nick_rackley]

Albert Einstein and Thomas Edison were incredibly smart by my definition.

Edison especially - he used common sense to figure out a invention process that was way more effective by accepting the limits of his abilities to reason about things a priori. The phonograph sparked a movie and music revolution because he had the common sense to realize how incredible it was he could record his own voice , even though his main project was working on a telephone. Many intelligent people would have simply remain focused on the telephone. He has about 1000 (literally) examples like this.

Einstein used common sense to question basic physics notions at the time. He realized the michaelson morely experiment was well constructed and thus very likely to be true at a time where the intelligent physics community was trying to explain away the result that speed of light doesn’t change in an imaginary ether.

[@nickrackley @nick_rackley]

But if it’s a U.S. provider of a harness using a chinese model on APIs then youd have to pay a price similar to OpenAI and Anthropic. A U.S. provider using Chinese APIs is way cheaper but you have ip issues.

I think your comparing the current cost of us glm and Chinese api providers to the theoretical unsubsidized cost of OpenAI and Anthropic. That completely ignores that us providers and Chinese APIs are also heavily subsidized. The Chinese prices make no sense if you calculate how much it would cost you to run their models on your own hardware so it’s obvious they are also subsidized, likely by their own govt.

[@nickrackley @nick_rackley]

I’ve looked at if I stood up my own GLM 5.2 H200 cluster and come to the conclusion is probably roughly similar to paying frontier api costs (10-20k /month depending on if you turn it off when your not working). It would only make sense if I inference my product on it also, and I’d have to make my own coding harness. I may do it eventually to avoid ip sharing with anthropic and OpenAI, but definitely not for cost savings.

If your a startup and talking about using the Chinese APIs to code, you are saying you are cool mainlining your IP to a company that are are using because they stole the IP from frontier labs.

Probably not a good business strategy.

[@nickrackley @nick_rackley]

@Chris_Somewhere @Ross__Hendricks What overconfidence, I’m literally describing my own personal product preferences and people I know.

That’s like saying a Lamborghini club is overconfident their cars are faster than civics and that they are willing to pay for it.

[@nickrackley @nick_rackley]

@hoangbrian @Ross__Hendricks https://t.co/PQ06fJv2VQ , I’ve completely replaced income from two high paying jobs for me and my wife. The narrative nobody is making money is just nonsense.

[@nickrackley @nick_rackley]

@ivan_bezdomny This is why I haven’t wanted to raise or hire, the time either would take me quite possibly would approach the time to personally build whatever features I was raising or hiring an engineer to build in the first place.

[@nickrackley @nick_rackley]

@omarvvvr AI already writes better code then me, so I use it to write code.