The Thesis
Software is becoming the work.
For thirty years, technology helped humans do work. Systems captured information, organized it, and showed someone what happened. Then a person took over.
AI erases that boundary. And once software can act, the question stops being what your company knows. It becomes how fast and how reliably your company can turn what it knows into a completed outcome.
The shift
Six assumptions moving at the same time.
Software informs work
Software performs work
Humans integrate systems
Intelligence integrates systems
Headcount creates capacity
Systems create capacity
Features are the moat
Execution is the moat
Productivity is the metric
Execution throughput is the metric
Systems of record
Systems of execution
What this is not
“AI replaces enterprise software” is the wrong argument.
That framing is too crude. The systems underneath the enterprise remain essential. Records, rules, permissions, transactions, and audit trails do not disappear because an agent can reason.
What changes is their strategic role. A system of record can be immensely valuable without being the primary interface. A transaction engine can be indispensable without determining the workflow.
AI rearranges where value lives inside the stack. Some parts become less visible. Others become much more important.
The three formats
- Arguments
- Longer pieces that build the case.
- Dead Ideas
- A named assumption, what dies, what survives, where the value moves.
- Signals
- What happened, why it matters, what to watch.
The arguments that build it
The Screen Can Shrink. Governance Cannot.
The strongest objection to the disappearing application is also the reason the execution layer becomes valuable.
The Workflow Escaped the Application
Once intelligence can coordinate across systems, the boundaries of the application stop being the boundaries of the work.
Software Is Becoming the Work
The important shift is not that software got smarter. It is that software can increasingly act.
A Violent Repricing of Tech Has Started
AI is attacking the two economic assumptions that powered technology for the last twenty years: expensive labor and differentiated features.
The Execution Layer Is Where the Value Moves
Everyone is focused on better models. I think the bigger shift is happening one layer lower, where decisions become actions.
Under observation
What I am watching next.
- AI interaction layers moving above SaaS applications
- MCP and open agent interoperability
- Cross-application agent execution
- Outcome pricing
- Software companies moving away from seat economics
- Services companies scaling revenue without proportional headcount
- Coding agents changing engineering team design
- Enterprise agent governance
- AI making management coordination less necessary
- Generated interfaces
- Headless enterprise software
- Context and permissions becoming strategic layers
- Authoritative state management between agents and systems
- AI-native systems integrators
- Changes in private-equity valuation models for services companies