AI-assisted, AI-first, and AI-native
| AI-assisted | AI-first | AI-native | |
|---|---|---|---|
| Where AI sits | A feature in an existing product | The main interface is AI | The product is built for AI to inspect and change |
| What AI can see | The prompt and a few context fields | The prompt and selected data | The interface, data, logic, and capabilities |
| Can AI change the product? | No, or only through narrow settings | Sometimes, through tools | Yes, in the same structure that runs it |
| Who can use the result | People | People, through AI | People and agents |
The labels are not a maturity ladder. A good AI-assisted tool can be more useful than a weak AI-native one. The distinction matters when the work needs to keep changing, or when an agent needs to operate the software rather than just answer around it.
What an AI Native Framework changes
A normal framework is optimized for humans to write code. An AI Native Framework is optimized for AI to read, change, run, and verify the application. That changes what the code has to be:
- The interface, data, logic, and capabilities should live in one readable structure.
- The code should explain what the app does and where its state lives.
- Changes should be verifiable in the running application, not only in a diff.
- A person and an agent should be able to use the same result.
This is the design goal behind MCL, Sigla's AI Native Framework. MCL expresses the Space as one body of code. That code is also the context an agent reads. The agent can inspect the current application, make a change, run it, and compare the result.
Code as context
"Code as context" means the application code is the description an agent reads, rather than a separate document that can drift out of date. When the code is the context:
- There is one source of truth for the interface, data, and logic.
- The agent can see what actions are available and what state exists.
- A change is made against the current application, not a summary of it.
This is why Sigla can keep the result usable. The app is not generated once and then frozen; the same structure that runs it can be changed again.
How Sigla applies it
In Sigla, an AI Agent Space is the working environment. It holds the interface, data, files, logic, and connected capabilities. MCL describes it. The agent reads the code, changes the Space, and verifies the result in the render view. You can then use the Space yourself, or ask the agent to operate it for you.
That is also the link to App for Agent: because the interface and state are readable, an agent can use the app without a human clicking through every screen.
FAQ
Is AI Native the same as AI-first?
No. AI-first usually means AI is the main interface or the main value. AI Native means the software is built so AI can inspect and change it. A product can be AI-first without being AI Native.
Is AI Native only for developers?
No. The point is that you do not have to be the one writing the code. In Sigla, AI writes and changes the Space; you describe the result and check it in the interface. Developers can still inspect the code in dev mode.
Is MCL open source?
MCL is already used in Sigla. The open-source release is in preparation and will be announced on the MCL page.