Carbon Commitment
Last Updated: September 2026
Photo: NASA, Blue Marble 2012, public domain
Using Model Communicator to communicate does not depend on any AI model in the cloud.
Model Communicator is software, and software has a carbon footprint. Every server, every download and every AI model we use takes electricity to run. We are a very small team, and our footprint is small, but we do not think that lets us off the hook. This page explains where our footprint comes from, what we already do to keep it low, and what we commit to going forward.
This page is a companion to our Use of AI page. AI is the largest and fastest-growing part of our footprint, and the two pages are meant to be read together.
It also grows out of our Frictionless Philosophy. A tool that runs offline on the device you already own is both easier to adopt and lighter on the planet.
The short version: Using Model Communicator to communicate does not depend on any AI model in the cloud. The apps run on devices people already own, work offline, and do their speaking and predicting on the device. Our footprint comes mostly from how we build the apps, not from how you use them, and we plan to shrink that too.
Where Our Footprint Comes From
We can account for our footprint in three places.
Building the apps
As we explain on our Use of AI page, we used AI tools to help write our software and draw our symbols. Those tools run on large models in data centers, and they are the biggest part of our footprint.
Hosting
Our website, the web app and the small database that stores accounts and custom boards run on Microsoft Azure. This is a light load: the website and the app are delivered as files, not generated by servers on each visit.
Your device
When you use Model Communicator, the work happens on your iPad, Chromebook, tablet or computer. We do not run AI models or speech services in the cloud on your behalf, so using the app adds almost nothing beyond the power your device already uses.
What We Do Today
No cloud AI when you communicate
Word prediction and speech run on your device. Every tap on a communication board would otherwise be a request to a data center; instead, it is handled locally.
Offline first
Once loaded, the web app and the iPad app work without a network. Symbols and boards are stored on the device, so they are not downloaded again on every visit.
Hardware people already own
Model Communicator runs on iPads, Chromebooks, Android tablets, Fire tablets and ordinary computers. Nobody needs to buy a dedicated device to use it, and a device that is already in the home or classroom does not need to be manufactured again.
Static, not generated
Our website is built once and served as plain files, which takes far less computing than generating pages for each visitor.
Our Commitments
What we promise
- Using Model Communicator to communicate will never depend on an AI model in the cloud. Speaking and predicting will stay on your device. This is also a privacy commitment, and we make it on our Use of AI page too.
- We will keep the apps offline-first and lightweight, and we will keep supporting older devices for as long as we reasonably can, so that nobody has to replace hardware to keep communicating.
- We will use the smallest tool that does the job. Where a fixed word list, a template or a simple program will do, we will not use an AI model.
- When we are able to purchase our own capable hardware, we will move our ongoing development work to lightweight, open-source models running on it, and away from large cloud models.
- We will be honest about our footprint. We will not make claims we cannot back up, and we will update this page when what we do changes.
Our Plan
The biggest change we can make is in how we build. Large AI models in data centers use a great deal of electricity and water. Small, open-source models running on a single computer use a fraction of that, and they keep our work independent of any one company. When we can afford our own capable hardware, that is where our development will go.
We cannot yet afford that hardware. Until we can, we will keep our use of large models to what a team of our size needs to keep Model Communicator free and working, and we will say so plainly rather than pretend otherwise. We will update this page as that changes.
earth 
tree 
recycle 
battery
Smaller models, local hardware, less electricity.
Ideas for shrinking our footprint?
If you have questions about our carbon footprint, or ideas for reducing it, we would like to hear from you.