good model behavior requires care for craft from the top
craft is hard at bigger companies with so many cooks in the kitchen, where product execs obsess over a/b test metrics and research execs over eval numbers. it requires a ton of work & political capital to make & advocate for something that feels invisible when good, and push back against efforts that ruin it -- and eventually people burn out
so then you end up with bad model behavior, and maybe a few people complain internally but then you launch anyway and then twitter blows up and then oh no we need a code red on model behavior but none of the structural problems go away
granola is the best you can feed your agents!
transcripts are a misleading source of truth. every name, jargon, internal company project is an opportunity for an agent to rabbit hole in the entirely wrong direction.
so we added these little hints so your agent always knows which Sam was in the room and that "UPS" is not always the postal service đ
Over 1 million Granola notes are read by agents. Every day.
So we built a version of Granola for them too.
Agents now get expanded notes via MCP. With extra context â like explaining who Sam is or the latest news on Project Sprout.
So they're always up-to-date with your work.
spent some time with the @SierraPlatform APM folks last week :)
was very very nostalgic to experience the energy and joy of being a fresh grad
also very happy to see companies continue to invest and nurture young talent đ±
The principle of semantic zoom is awesome and can be applied on many more places. This is a quick mockup I did on the Grok Bot app where zooming out groups by topic and allows you to find an earlier place in the chat easier
That said, I have a pretty radical vision/idea for how these interfaces should evolve, and it's nothing like this