Sign in Subscribe

Is a bootstrapped AI wrapper startup possible ?

The short answer is still Yes. In May, I wrote in this blog about the uphill task in creating a 0-1 coding startup in this AI age. It is much easier for a bigger company to vibe-copy your code and do well due to its existing network and distribution effects. Here, we delve into the economics / viability of AI wrappers companies

Are AI harness companies worthless ?

Since the main brain is the AI model, the valuation of wrapper companies is questionable.

Quoting from longerbook

There is a specific category of AI startup that looks like a rocket ship right now and is, in fact, a sandcastle at low tide. Lovable is the cleanest example, but the argument applies equally to Bolt, v0, Replit Agent, and the dozen other “describe an app, get an app” products that raised at unicorn valuations in 2024 and 2025. They will not exist as standalone businesses in three years

Their main argument is that these startups are lacking any moat and will ultimately be made obsolete by AI model companies who are looking to move up the value chain. Claude code and Codex is a very good example of this. However, AI wrappers that have distribution moat will prosper. Harvey has this for legal. Hippocratic AI has this for hospital systems. Tessl has it for enterprise software development. Anduril has it for defence.

Neolemon - A rare gem

Neolemon is an AI-powered cartoon generator and character consistency system designed for creating illustrated content like children's books, comics, and animations. When creating images for content using tools like Midjourney, users have frequently reported inconsistency across characters in images. Neolemon addresses this by giving a single interface where users can plan / edit characters keeping them consistent across the storyline. The founder frequently engages with a community of users addressing their needs to update the product user interface making it more accessible.

Neolemon provides an editor to help user tweak the character (Source: neolemon.com)


Neolemon was bootstrapped by founder Sachin Kamath and a two-person team starting in a small flat. Operating with zero ad spend, the AI-powered cartoon storytelling platform scaled rapidly from its initial launch (formerly rebranded from ConsistentCharacter.ai) to over $35,000 to $50,000+ MRR within roughly a year. Comparing to companies like Cursor, Midjournery and Harvey which have millions of dollars to create network/distribution effects, this is a rare gem that has bootsrapped its way to revenue. I wish the best of luck to Sachin Kamath and his team.

Future in progress - Self improving AI

There is extensive research ongoing in a self improving AI harness. The feedback loop in modern AI may indicate the model rewriting its own weights directly, or more broadly the model improves the training pipeline and the deployment system, which in turn enables a better successor model with improved performance across economically valuable tasks. A popular example is autoresearch code by Karpathy . Below are extracts from a self taught optimizer

Self Taught Optimizer(Image source: Zelikman et al. 2023)
Examples of self-improvement strategies discovered by STOP. (Image source: Zelikman et al. 2023)

Harness Layer vs Core Intelligence?

Lilian weng has a prediction on future of RSI (recursive self improvement):

  1. Harness engineering will evolve in the direction of meta-methodology (i.e. improving the machinery for getting better answers, not just improving the answer itself). The harness system itself becomes an optimization target, with fewer heuristic rules and more general mechanisms.
  2. In turn, mature harnesses enable auto-research for model self-improvement loop and smarter models prevents harnesses from overengineering and keep the system sustainable.

Eventually it is possible that many harness improvements will be internalized into core model behavior, but the interface with external context and tools should remain.

Takeway for future startup founders

AI remains a fluid and fast evolving field and the future is quite uncertain. From the above discussion, I can suggest a few takeaways for future AI startup founders:

  • Product differentiation and UI matters
  • Listen to customers/market and fine tune the offering
  • Focus on human workflow that is a real value add
  • Distribution / data remains an invaluable moat

Comments ()