Tiny Tech: Anyone Can Cook (software)
One thing about AI discourse is that it often feels so…big.
To be clear, AI IS big! But I feel like all that I ever hear about is AI doomsday, economy-distorting capex, frothy VC funding, catastrophic job losses, etc.
These headlines have an unintended consequence: extreme headlines lead to extreme positions. You are either AI-pilled and a zombie who can’t function without it, or you are anti-AI and a Luddite living in the Stone Age. In between these extremes is an area with tremendous opportunity: the small scale.
Back in 2024, I hypothesized that AI would lead to an explosion of “Tiny Tech” - niche technology only possible thanks to AI. In fact, that was why I quit my job to build something (and failed at it…though that’s a story for another time). I was too busy building the startup to write this article back then, but better late than never.
I think Tiny Tech is an incredibly exciting and underdiscussed consequence of the AI boom. We’ll walk through exactly what Tiny Tech is, why AI enables it, and who is poised to benefit.
The basics
First, let’s discuss a simple but critical question: How does a product (physical or digital, good or service) get made and used?
Fundamentally, there are 3 steps:
- Ideation - someone has to have an idea for a product
- Creation - the product must be made
- Distribution - a product must find customers and get into the hands of those customers
For software, all 3 have been bottlenecks in the past. Software creation in particular required:
- Capex - infrastructure required to “host” the software
- Opex - investments (engineers) required to build and maintain the software
Starting in the late 2000s, capex was diminished as a bottleneck thanks to the Hyperscalers and SaaS; you can create storage, databases, email services, payments, etc. very quickly thanks to their products and services. However, opex was still a limiting factor! Software engineers were required to use these services in order to build a software product.
But now, AI is diminishing opex as a bottleneck! Coding agents are (arguably) better than most human software engineers, bringing the cost of software creation close to 0. All you need is a repo on GitHub, an LLM to write code and leverage Hyperscaler/SaaS products, and you’ve got functional software.
So the question we must ask ourselves is: what happens in a world where software creation is no longer a bottleneck? What does it mean that only ideation and distribution remain as bottlenecks to software products? To understand how the world might change, we must know who benefits in this new world.
Who benefits?
The people who benefit are those who already have ideation and/or distribution. In my mind, there are 3 groups that benefit the most:
Let’s dive into who these groups are, and why I think they are positioned to thrive:
Small Business Owners
As a small business owner, you have both ideation and distribution. After all, you’re full of ideas from all your customer complaints. And if you have customers, you have distribution.
Historically, if you were a small business that depended on tech, you relied on monolithic, expensive, and cumbersome technology products designed not for your needs, but for the average needs of many businesses. You had no other choice because nothing else could solve these problems, so you begrudgingly used these products.
Today, you are no longer constrained by this! If there is a customer problem that you want to address, you can build something perfectly suited for that specific problem. As an example, Lam Nguyen, the owner of my favorite dance studio in Chicago, runs a dance battle every year called Windy City Throwdown. Dance battles are too niche for great technology to have been built for event organizers, so these competitions have usually used lots of pencils and paper.
Thanks to AI, this has changed! I worked with Lam to create a custom tech stack for Windy City Throwdown that perfectly addressed his exact needs. This was only possible because AI brought the cost of developing that software close to 0 by empowering me to do this on my own with little effort.
Of course, there are limitations: small business owners aren’t going to vibecode a payments platform, and AI isn’t addressing physical bottlenecks like logistics. Yet there are countless other problems that can now be addressed by small businesses directly. This will drive better customer experiences, allowing them to be more successful over time.
Content Creators
Content creators have a lock on distribution thanks to their followings. Historically, they monetized their distribution by:
- Being the modern billboard and monetizing via revenue shares with their platform or directly via ad partnerships
- Selling exclusive content (Patreon, YouTube memberships, etc.) or merch
Creators like MrBeast have monetized in other ways as well, but those required real-world investments so capital-intensive that only the largest creators can make them. AI could offer new revenue opportunities for all creators. Creators often become known for a particular niche, and now they can build any software they can imagine for that niche.
An example that I really love is Jorbs and Spirebird.com. Jorbs is a content creator who specializes in the Slay the Spire (StS) games - a niche and difficult deckbuilding game that I absolutely adore. The types of people who like StS tend to be analytical and nerdy, and Jorbs has built a perfect product for this audience: Spirebird.
Spirebird is a way for all StS players to aggregate and analyze their data. This enables the entire community to use large sample sizes to make better gameplay decisions. If that sentence didn’t make any sense, that’s okay! Spirebird is completely arcane and inscrutable for anyone outside of hardcore StS players, and I think that it is absolutely beautiful! Jorbs doesn’t monetize the platform directly, but it deepens his connection with his audience and introduces new players to his content. I’m so glad Spirebird exists, and I’m glad AI tools were available to help Jorbs build it.
(side note: you could argue that Content Creators are screwed due to having to compete with AI content. Personally, I don’t think that AI content will threaten niche content creators in the long term, but that’s a discussion/thesis for another day!)
Tinkerers
I’m slow in this video because…I’m dumb
This last group needs a definition - tinkerers are people who like to build stuff for themselves. It could be for fun or to solve an inconvenience. Tinkerers are overflowing with ideas, and distribution isn’t an issue because the audience is just themselves or maybe immediate family members and friends.
I’d consider myself to be a tinkerer, so I have many personal examples. An earlier article dove into a cooking app I built, but another example is an app I built to learn Sinhala! Sinhala, and other rare languages, aren’t on services like Duolingo because so few people care to learn them.
So, fuck Duolingo! I just created a tool for myself.
A detailed write-up on the Sinhala app will come soon. The important thing for this article is that it is a tool that I only created because AI removed the bottlenecks in my way. The past year has been unbelievably exciting because I’ve now realized that most of my barriers to creation have been removed, letting me build anything and everything I’ve ever wanted to build. It truly has been a Tinkerer’s dream!
Smallness is an advantage
You may have noticed that all the groups mentioned above are SMALL. Tinkerers or content creators are often working solo, and small businesses have small teams. A nuance worth highlighting is that the smallness of a niche or the team working on that niche is actually an ADVANTAGE with AI! This is true for 3 reasons.
First, small teams don’t have to deal with coordination taxes. Everyone on a small team is close to the customer, so it’s very easy to know exactly what the customer wants, and what to build to meet that need. Compare that to large companies, where teams need UX researchers, PMs, designers, etc. to address customer needs because the builders (engineers) are too far from customers.
Second, smaller codebases work better within the constraint of LLM context windows. For a small team/company, the entirety of an app can be found in a single codebase. And with good repo hygiene, it can be very easy for the model to navigate and understand the entire repo. If you’re at a large company with various systems and interlocking APIs, models need a lot of guidance to understand these interconnections without breaking things.
Third, because these projects are lower-stakes (not VC-funded, win-at-all-costs products), maintenance of the software is much less costly. If you create a tool for just you and your family, you don’t need to do any maintenance at all except for when someone makes a feature request. In addition, you don’t need to invest much in cybersecurity since your project is small and not worth attacking - you can trust the models to implement basic cybersecurity defenses.
Anyone can cook
In highlighting the groups who will benefit the most from AI, I am reminded of one of my favorite scenes in all of cinema: Anton Ego’s review at the end of Ratatouille.
Anyone can become a small business owner, a content creator, or a tinkerer, and now, AI lets anyone cook (software).
If you’ve met any software developers, you know they tend to be a certain type of person. This is not a bad thing; it’s just that certain personalities, idiosyncrasies, and thought processes make for better software developers. As a result, most software tends to come from a specific worldview.
Now that AI is democratizing knowledge so that anyone can be a software developer, many new ideas will be brought to life by individuals with experiences and knowledge well outside the norm of Silicon Valley and the average software engineer.
Tiny Tech’s Cambrian Explosion
If anyone from anywhere is able to build software, what results is a Cambrian explosion of Tiny Tech. And like the Cambrian explosion, it’s gonna get real weird. As a reminder, look at Spirebird.com! That thing is fucking weird, yet that’s what makes it magical! Spirebird is truly the Opabinia of today’s world:
isn't this just the cutest Lovecraftian nightmare you've ever seen?
To be clear, most of what gets built will be shit. If you’ve looked at random apps on the App Store recently, most of those are shit. Honestly, most of what I build is probably shit too.
BUT this is exactly what happened with social media: 99.9% of what gets uploaded to YouTube/TikTok/Instagram is shit. What matters isn’t the frequency of quality, but the overall volume of quality things being created. The amount of tech that will be created in this Cambrian explosion will be staggering; if just 0.1% of it is good, we’ll be swimming in incredible products and tools that otherwise never would have existed.
Jump in! The water’s lovely
Admittedly, the Tiny Tech dream is not for most people. There just aren’t that many people with the initiative and vision to be a small business owner, content creator, or tinkerer. Returning to Ratatouille: “anyone can cook” doesn’t mean everyone can cook - instead, a great cook can come from anywhere.
That being said - I do think that this represents an opportunity for you, dear reader, if you are a tinkerer! You probably have a small problem in your life that you’d like to address. Please go build something to solve it! If you haven’t been using AI tools to build software recently, you will be absolutely floored by their capabilities. Once you start building, you’ll start spotting small problems everywhere you look.
I’d like to end with an article I discovered after completing my final draft: An app can be a home-cooked meal by Robin Sloan. This article really tilted me because Robin captured portions of my thesis, but did it all the way back in 2020!
I finally understand your pain, Butters...I finally understand
Alas, no ideas are truly original, so I’d like to just reiterate his message here: build small, delightful, bespoke software about things that you love for the people that you love. What you create will be for you, by you, without any ulterior motives. Now that AI has completely democratized software development capabilities and knowledge, anyone can jump in the water and get started. You don’t wanna miss out on this Cambrian explosion, do you?