My Prompt to AI chat buddy:
Riffing off “AI and the Next Economy“, I want to think about what might change in society as it becomes cheaper to make ad hoc apps using AI tools. What desires and capabilities might change as the benefits of easy to make software are felt by AI users and those they help in society? What opportunities to create elements of a new circulatory economy might there be?
Opus 4.5 reply:
O’Reilly’s central insight—that an economy is circulation, not just production—becomes especially interesting when applied to the specific case of ad hoc software creation. Let me think through this with you.
The Current State: Software as Choke Point
Right now, software is itself one of the “choke points” O’Reilly describes. There’s a massive gap between what people need and what they can get. You either:
- Use mass-market software that’s 80% features you don’t want and missing the 20% you need
- Pay significant money to hire developers for custom solutions
- Cobble together workflows with duct tape and spreadsheets
- Simply go without
This gap represents enormous latent demand that never becomes economic activity. The person who needs a specific inventory system for their small pottery business, the neighborhood group that wants to coordinate tool-sharing, the teacher who wants a particular kind of quiz generator—these needs mostly go unmet or are met badly.
What Changes When Software Becomes Cheap to Make
The desires that might shift:
From adaptation to articulation. Currently we adapt ourselves to software. We learn the app’s logic, its vocabulary, its constraints. When software becomes cheap to create, the valuable skill becomes articulating what you actually need. “I want to track my sourdough starters, when I fed them, how they behaved, and see patterns over time” becomes a software specification rather than a search query for an app that almost does this.
From features to fit. Mass-market software competes on feature count because it’s trying to serve everyone. Ad hoc software competes on fit—does it solve your actual problem? This could rehabilitate the idea of “good enough” as genuinely good. Your bespoke tool that does one thing perfectly may be more valuable than a polished app that does a hundred things you don’t need.
From consumption to commissioning. There’s a qualitative difference between being a user and being someone who brings software into existence. Even if AI does the actual coding, the person who says “I need this” and iterates on it until it works is engaged in a kind of creation. This might shift how people relate to their tools—more ownership, more willingness to change them.
The capabilities that might emerge:
Problem articulation as a skill. If anyone can make software but the bottleneck is describing what you need, then people who are good at understanding and articulating problems become valuable. This isn’t quite programming, and it isn’t quite traditional “business analysis”—it’s something like being a good commissioner of solutions.
Domain experts as direct tool-makers. The nurse who understands exactly what’s wrong with the patient handoff process can create the tool to fix it, without translating through layers of product managers and developers. The gap between “knows what’s needed” and “can make it” collapses.
Rapid local iteration. Small organizations could try things, learn, discard, try again—a kind of evolutionary approach to tooling that was previously too expensive. The pottery studio tries five different inventory approaches in a month rather than committing to one system for years.
Circulatory Opportunities
Here’s where it gets interesting from O’Reilly’s framework. What new kinds of value flow might emerge?
The long tail of software needs becomes economically real. Right now, there’s a power law in software: a few products serve millions of users, and countless specific needs go unmet. Ad hoc creation could make the long tail viable. Every small business, every hobby community, every neighborhood could have tools that fit. This is a vast expansion of the “devouring” capacity O’Reilly discusses—more demand satisfied, more value flowing.
New roles in the circulation:
- Articulators: People who help others figure out what they actually need. Part therapist, part consultant, part translator.
- Curators: People who notice that the tool someone made for their situation might help others with similar situations—a kind of pattern-matching across the long tail.
- Maintainers and adapters: Taking ad hoc tools and evolving them as needs change.
Local solution economies. A person who’s good at commissioning AI to make tools could serve their community—the person in the neighborhood who helps local businesses get custom solutions. This is labor-intensive, relationship-based work that circulates value locally rather than concentrating it in platform companies.
The “complement” question. O’Reilly notes that Ford’s cars needed roads, gas stations, hotels. What are the complements for ad hoc software?
- Trust and verification systems (does this software do what it claims? Is it safe?)
- Sharing mechanisms (how does a good local solution find others who might benefit?)
- Integration layers (how do ad hoc tools connect to each other and to existing systems?)
- Teaching and support (who helps people learn to articulate, iterate, maintain?)
The Risk: New Choke Points
The circulatory potential is real, but so is the risk of new concentration. If ad hoc app creation requires access to powerful AI models controlled by a few companies, we might just be moving the choke point rather than eliminating it. The person can create the software, but only by paying rent to the model provider. The value circulates locally but a significant portion flows upstream to the AI platforms.
O’Reilly’s point about decentralized architectures is crucial here. Open models, local inference, tools that work without constant cloud connection—these become infrastructure for a genuinely circulatory economy rather than a new form of extraction.
A Thought Experiment
Imagine a future where creating a custom app is as easy as describing what you want in a conversation. What happens to:
The person who currently uses spreadsheets for everything? They might become prolific creators of small tools, each fit to purpose. Their skills in understanding workflows and data relationships become more valuable, not less.
The small business that can’t afford custom software? They get tools that fit their actual operations. The plumber has software that understands how plumbers actually schedule. The productivity gains that large enterprises have captured through custom tooling become available broadly.
The community group? They can have tools for coordination, memory, decision-making. The neighborhood tool library has an actual system for tracking who has what. The capacity for collective action increases.
The relationship between “technical” and “non-technical” people? It blurs. The valuable distinction becomes something like “good at articulating problems and iterating on solutions” vs. “not good at that”—which is a different skill than coding, and more widely distributed.
What’s striking to me about your question is that it takes O’Reilly’s macro-economic argument and asks what it looks like at human scale. The “circulation” he’s describing isn’t just about GDP and wage shares—it’s about whether people experience themselves as participants in creating value or merely as consumers of what’s been created for them. Ad hoc software creation could be genuinely democratizing in this sense: it makes more people capable of solving problems, which creates more value, which circulates through relationships and communities in ways that mass-market software never could.
What aspects of this are you most interested in exploring further?