Project summary

As a Creator in the UK's Advanced Research and Invention Agency's Safeguarded AI programme, Ink & Switch has a wide variety of contributions. Over 30 months, we're working across all levels of the stack to deliver user experience prototypes, data infrastructure, and to support our fellow researchers.

What are this project's goals? How will you achieve them?

We currently have three top-line goals:

• GAIOS: to study & develop a collaborative operating system for humans and LLMs to team on tasks that are safety critical

• Geolog: to build on the work of automerge and research from our TA1.1 mathematical partners a model of zero-concurrency collaborative auditable data structures secured by an E2EE replication system with a capabilities model

• Clearcut: to partner with the mathematicians in the programme to accelerate their research by adapting elements of the above (and our other existing research) to provide them with new collaboration tools

How will this funding be used?

This funding is intended to provide working capital in support of our ongoing open source research.

Who is on your team? What's your track record on similar projects?

Ink & Switch's team of researchers combines production software engineering experience, design thinking, and academic rigor to produce industry-leading research results. Our team helped pioneer the local-first movement (including coining the term), and has other research tracks including relevant work in malleable software and universal version control.

We were enticed to participate in the project by programme director Davidad's recognition that collaboration with untrusted peers requires version control and history review, and our collaboration expanded from there.

With a decade of experience pushing the boundaries of what's possible in computing and inspiring both industry and academia, we're excited to bring our perspectives and skills to this project.

In addition to our research outputs, we also develop and maintain the automerge sync engine.

What are the most likely causes and outcomes if this project fails?

The overall ARIA programme relies contributions from a wide variety of performers across numerous academic fields and industries. The ambitious goals and timelines for the project reflect the urgency of the current global moment, and on some level, failure to achieve all of them is the likely outcome.

Whether or not the ultimate vision of a "safeguarded AI" can be fully realized within the project scope and timeline, Ink & Switch will be able to use these resources to deliver open source implementations of valuable building blocks (such as geolog, GAIOS, and clearcut), and to disseminate knowledge of the problems and opportunities in the space.