
<aside>
💬
"One way to see courage is acting before necessarily knowing all of the answers."
</aside>
Friday October 30, 2026 · 10:30 AM · Conversation
Key ideas from the talk
- Ana proposes treating AI systems as a public utility, built as good enough federated public AI stacks. A fully sovereign stack is not realistic yet.
- The first targets are small arts and culture organizations. The aim is to cut administrative load so a four-person shop can do more creative and strategic work.
- Her main example is an open source, Canada-specific grant agent. It would teach applicants a funder's criteria and past patterns instead of writing the grant for them.
- She says tools that write grants for everyone make every application sound the same.
- She sorts student worries into three buckets: copyright and compensation, career paths, and how teaching changes. She says educators should have agency over the tools.
- On federal policy she names the $1.6 billion investment in Cohere, a $160 million Mitacs program for AI adoption and a $50 million creative technology fund. She links adoption to trust.
- Prem says peer juries keep a human step in grant decisions. An audience member described a funder lottery that screened out applications before any review.
Quotes
- "We really need to start building federated public AI stacks."
- "We're faking it till you make it."
- "They are largely made up of small, medium enterprises and organizations who are doing ten times more work with less people than other sectors."
- "It's a capacity building kind of tool as opposed to necessarily a grant writing tool."
- "People are using AI to write grants. Granting agencies are using AI to process the grants. So you've got these tools talking to tools."
- "Innovation happens at the speed of trust." Ana, quoting Minister Solomon.
- "You need deep understanding of practice to be able to figure out how to teach these tools properly."
About the speakers