LLM Imposter Syndrome

I’ve built many things since code agents emerged on the market in May 2025. With my product knowledge and ability to use Reddit, StackOverflow, and reasoning agents, I was able to create some applications. When DALL-E 2 was released in 2021, I utilized it to try to create anything under the sun. Same with Sora, Stable Diffusion… but it’s been four years since I first used all of these tools.

The Developer Divide

When I develop, I feel a few things: I go online and read about non-technical people building apps, and developers REAMING on them. I honestly side with the developers—a bit of me takes the mob mentality here—but also developers are kind of in the wrong, unless they’re compiling new code or going a layer deeper. I think they feel threatened, and really it comes down to this: there is now an intersection of product developers that no longer need to be very technical, and technical developers either deep-diving libraries and assisting in more technical work.

Here’s the thing though: these non-technical people are VERY non-technical. There is an abstraction layer I have where I’m actually able to understand the code LLMs write, where the code lives client or server side, why things work as they do, the idea of languages…

BLUF: I have 50+ credits in college courses but wouldn’t be able to code an application from scratch without heavily consulting StackOverflow and documentation.

The Weight of Inadequacy

I feel like I don’t deserve to be here, to be creating these products. I feel cheap in what I develop. The amount of work that the LLM can do on my behalf is pretty impressive now. The issue I have is I’m expanding my universe per usual and not finishing what I started. I am too far across multiple maps and not tracking singular progress.

The All-In guys (outside of my disagreement with their change in views and heavy Trump lobbying, as well as their inability to level set with people—elitist tendencies—not in scope for this writing) predicted the move from SaaS to “Models As a Service,” which actually happened. But what they didn’t predict was that models would become a bit of a utility, and the bottom lateral—meaning all the apps at the base level and digital footprints from scratch—would be much easier.

I was able to set up a Next.js site using Vercel and even get the SEO up to 5 sites on Google for Washington Rugby. It’s kind of crazy.

The Crowded Marketplace of Ideas

We are truly able to generate information, product, and ideation in physical form. Because of that, you have millions of other creators now able to ideate. People are creating ideas that they think are great, but really there is such a crowded market for it.

What does the marketplace look like at that scale? Is it really sustainable?

At some point, again, the providers for the amount of services are the 10x, UNLESS we go back to local-first principles and the user brings their own compute, storage… Anyways, I feel like whatever I develop is not worth it, crushed in a sea of obsolescence.

Building on Stolen Knowledge

The imposter syndrome I find also revolves around the fact that LLMs themselves are inherently “thieves” of ideas. This is why I mentioned DALL-E. I quickly found myself disenchanted by creations—not what I created but what others created on LLMs. Some prompts were extraordinary, were very interesting, but these ideas are synthetic. They are too high velocity and they don’t mesh with our primate brains.

We are getting fried by screen-maxing, 6-second videos, and general content degradation, as well as enshittification by large behemoth vendors (Netflix, Amazon, Facebook). We are building on stolen knowledge.

Enter Thoughtful App Co (TACO)

Hence the idea of Thoughtful App Co came up with a different approach to building apps: using local-first principles and useful ideas that people could use all the time. We search for user journeys and archetypes that any adult experiences or could use—any concept that can be properly optimized to reduce screen time, allow users to not be on the app (through integration with omnichannel), and provide some brick and mortar (even user-supplied) to the digital experience.

A Different Approach to Ethical AI

A TACO idea is one:

Create a platform where we have a Reddit-like data provider that allows contributors to make money off their data being sold in packages. Think of this: tokenizing contribution value to every dividend. We can have people also validate other contributors as another function (think how blockchains have different roles and mechanisms served by different users). Having a proof-of-custody for data and data rights.

Create an ethically sourced LLM that truly uses open-source data or looks to vendors to provide information from ethically sourced places. Maybe it sucks, but there are zero legal issues downstream.

As long as inference can be contained in costs reasonable to a corporate user, then we’ll be okay in gradually taking market share from the top 10% of current companies. It’s just interesting where capital will flow, what companies will do to pivot, and how people will think. I try not to be too bearish and doomer on my thoughts, but I can’t help it when our current market DOES NOT favor an open market. In the same way Verizons and AT&Ts were made in their ability to establish on the internet, you may see this in AI. The emergence of P2P and other mechanisms have emerged to counteract this at some point.

The Path Forward

I think the issue is I need to create these products. If it’s one of those things where I need to create 2-3 products and then farm them, then I can. Part of me believes if I could have a product lead and technical lead, and between both of them they can come up with a product, then they have ownership for that product. Across 3-6 apps I’d have a team of 12 people: 2 overhead, myself, a designer who contracts with artists, and a marketing team.

I was also thinking of using micro-marketing on existing platforms—like the ones I consistently bash. I could potentially take users off the brain rot and loneliness platforms and invite them to be more grounded in reality.

Summary

The imposter syndrome is real. The feeling that I’m building on foundations I don’t fully understand, using tools that do the heavy lifting, in a market that’s oversaturated with synthetic ideas—it’s all there. But maybe that’s the point. Maybe the real value isn’t in whether you can code everything from scratch, but in understanding what needs to be built, for whom, and why.

The question isn’t whether I deserve to be here. The question is: can I build things that matter, that ground people in reality rather than fry their brains, that respect data rights and human dignity?

That’s the challenge worth tackling, imposter syndrome and all.