Map Notes // Current Reading
As LLMs get smarter and we build richer harnesses around them, the economic center of software starts to change. The first things affected are the products where the moat was mostly implementation effort.
Simple CRUD sits at the center. Nearby are marketing sites, internal tools, personal finance apps, bookkeeping software, CRM, e-commerce, and other categories where current models can already compress a lot of implementation effort.
Farther out are vertical SaaS, regulated SaaS, legal tools, healthcare records, design tools, games, cybersecurity, databases, infrastructure, embedded software, CAD, robotics, bioinformatics, energy systems, semiconductor software, aerospace, operating systems, and expert software.
The farther from the center, the more the moat depends on domain judgment, safety, latency, hardware, distribution, regulation, proprietary data, operational trust, or deep expert context.
This map is meant to be updated over time. As models improve and new harnesses make them more capable, the value boundary moves and more categories need a clearer reason to exist beyond implementation.
The useful part of the map is not pretending the radius is exact. It is making the argument inspectable: which kinds of software are being repriced because implementation is becoming cheaper, and which still have a moat because the hard part lives outside the prompt.
Article FAQ
Article takeaways
- What is the core argument in this post?
- The map shows where software categories are becoming easier to build as AI improves, and where durable value still comes from judgment, safety, context, and hard-to-automate operations.
- Who should use this map?
- Founders, technical leaders, and builders deciding where to compete in software categories with changing AI leverage.
- How should I use this framing in planning decisions?
- Use it as a quick filter for project risk and moat strength before allocating budget to feature work.