There’s a growing contingent of AI-native companies and services, some pushing the bounds of creativity and expression.
Sora is releasing mind-bending, production-ready clips from a simple prompt:
Suno is dropping bangers, including my recent hit prompted by “chill couch vibes as I sit back and eat potato chips while watching TV. edm chillout”:
I laughed at the idea of a prompt engineer becoming the ‘Hot, New High-Paying’ job in tech, but I think there to be a broader truth behind it. Building, writing, creating, and doing have been resolved into a simple series of prompts. Explicit, well-defined work tasks can and often should be outsourced to LLMs. No longer should a software engineer be optimizing a loop, a writer cleaning up syntax, a translator having to write/annotate texts manually, or a copywriter creating and testing copy. Crafting work will translate to crafting asks.
Soon enough, we’ll see LLMs taking over longer-form, complex work, including rewriting books with a certain flare and creative liberty or sequels to hit movies. The range of motion will expand and quickly corrupt the long-developed skills of standard craftsmen (designers, artists, writers).
There are obvious hints that the next generation of companies will be AI-native or AI-augmented, though I expect quick shifts in the nature and aspects of work at the present moment. For the millions of craftsmen who’ve built and shared that work publicly, there’s likely a growing proportion that could be easily redone by a well-prompted AI (think research papers, accounting decks, company designs, musical beats). Quickly, we’ll lose the necessity of craft, friction, and the toils of defining personal feel before releasing work out in the open. Creative strategy work will continue to exist - determining how companies want a brand to look and feel, the general strategy for generating sales pipeline, or learning and creating a song. Just a few months ago I wrote about the want to produce music. The desire is still there, but the activation energy to start from scratch when I can get to 90% in a few minutes makes it a true personal undertaking.
It’s a sad reality, and I may be wrong. But in a couple of years, when research arms turn into research individuals orchestrating AI agents, and designs are simply left to a prompted UI, we’ll think back to when such obviously ‘automatable’ tasks and jobs were meticulously worked on by such talented creatives.
Every released a nice read on the one-person billion-dollar company. It is highly resonant, and a north star for what can be done (potentially not should be done). Crafting will become orchestrating, thinking through each arm of the business as a system with inputs and outputs, and building or buying AI tools to work on those tasks. Agents have proven capability of iterating on less defined tasks, so tasks like figuring out a way to deploy a certain software can likely be achieved without needing a developer resource.
With that, I don’t think the goal should be billion-dollar outcomes. The goal should be on well-crafted, self-sustaining businesses that solve specific user needs.
Why This is Exciting
Entrepreneurship should be about solving problems efficiently and resourcefully. Most venture firms require billion-dollar + valuations to justify their investments, meaning many companies, even those that have solved a problem for customers, simply cannot monetize to the scale worthy of a venture return. This is a sad reality for many worthy companies solving real problems. AI tools expedite workflows and ideally address the pressing use cases for a small set of customers sustainably.
Solopreneurship is taking new meaning, and the spaces and wedges to operate are expanding. Corollary to this - unbundling has its problems.
Update (April 2024)
Every’s more recent article touches on a similar topic. Replace technical with ‘Design,’ ‘GTM,’ ‘Marketing’ and it fits nicely. The craft is now articulation and definition, alongside a recognition of the gaps. Also exposes what a concern around learned hard skills. Much of my undergrad CS work revolved around work that could be easily automated (debugging, implementation, etc.). Why learn to wade through the muddy waters of learning difficult coding concepts (including working through errors, finding bugs) when there is a free partner who can do it for you?