Product
AANI
AANI asks what useful AI looks like when reliable internet disappears.
- Started
- 4 March 2026
- Status
- Research
- Signal
- Offline AI
The question
What should AI look like when internet is not guaranteed?
Most AI products quietly assume the user has stable connectivity, a modern device, enough data, and the patience to wait for a cloud response. That is not always real life.
AANI is my long-term question around accessible AI, low-connectivity environments, and what happens when AI products are designed from constraint instead of comfort.
Context
I am interested in AI that works where infrastructure is uneven, because people still need useful tools when the network is slow, expensive, or unavailable. If AI only works well in perfect conditions, it leaves out many of the people who could benefit from it most.
Building it
AANI is an exploration of local-first workflows, small models, SMS access, offline storage, low-bandwidth interfaces, and hybrid edge-cloud patterns.
The direction is that useful intelligence should degrade in stages. On a strong connection it can use the cloud. On a weak one it should fall back to a local model and sync later. With no connection at all it should still answer from what is already on the device.
What surprised me
Connectivity is not just a technical variable, it drives most of the other product decisions. It sets the interface, the cost per answer, the data flow, what the user can trust, and even where the bar for a good enough response sits.
What I learned
Access has to be designed in from the start rather than added once the cloud version works. Small models, offline states, and SMS or low-bandwidth paths are the parts that decide whether the thing survives outside a demo.