Start
In 2011, Marc Andreessen wrote the then-prescient, now-famous article Why Software is Eating the World. In that article, Andreessen argued that, after many decades of development, a foundation was finally in place for software to meaningfully transform every industry.
If there is one theme within software that defined the next-decade, it was B2B SaaS. The formula was almost too easy: pair the malleability and zero-marginal costs of software with domain expertise in any industry to deliver immense value and productivity gains.
The structure for the B2B SaaS startups looked something like this:
Today we're at the cusp of a new upheaval in technology, one enabled by AI, or, as I believe it should've been called, Indeterministic Computing.
AI is not just a buzzword. It promises to fundamentally transform our relationship with technology. We spent the last decade building or attempting B2B SaaS products. Our attempts included helping an oil and gas plant monitor its equipment, helping university education departments keep track of evaluations of their student-teachers, and, the most notable, helping Canadian doctors streamline their medical billing.
In every instance, there was an impedance mismatch at the user interface layer, where unstructured and chaotic real-world data had to be transformed into structured data that software could work with. The user had to conform to the constraints imposed by the software, instead of the other way around. AI promises to change that, and therefore, AI promises to help software eat even more of the world than previously possible.
Over the last decade, as software ate the world, software developers were in hot demand, as incumbents and startups in every industry tried to capitalize on the efficiency gains it offered. Will the AI revolution do the same? Will there be a new-found demand for software development expertise as companies try to capitalize on AI gains?
Our answer is “no”.
The skills required to build solutions with indeterministic computing will be fundamentally different from those involved in a deterministic one. Building end products with AI will be far more of an art than a science. Over the last two years we've worked with several teams to integrate AI in their work. We've seen that building compelling AI pipelines that can be integrated in software requires craftsmanship, experimentation, and taste in a way that writing deterministic software does not.
We believe the traits that make a successful AI craftsman are fundamentally different from those that make a successful software developer in the classical sense. And we believe that the future of building software will look more like this:
Today, in 2024, this transition is still in its nascent stages. While today's pioneering AI craftsmen are producing some incredible work using open source models and tools, they're not building products. The industry is still in the hobbyist phase. But this will change. A whole new generation of kids is growing up learning how to wrangle AI models with the same passion that our generation had for software development.
Our goal is to provide tools to the AI craftsman that will allow him to productize his talent. We want to play a small part in transforming this craft from a hobby to a profession.
We're starting with image generation because of its well-developed open source ecosystem and its immediate path to monetization. Our product plan revolves around three pillars:
Over the last decade, “learn to code” was seen as a failsafe career advice. Over the coming decade, it will be “learn to create (with AI)”. The rise of the AI craftsman is just around the corner.
Stay tuned.
The team at InstaSD
