Hector has not found who is the best builder, everyone is doing it differently. The Agent UI builder is in pre-alpha There is a still a core debate on how we are building it
What we call an agent and the rest of the world are different Our agents don’t run automatically in the background Building the agents are getting easier The issue is when to build an agent
Identify the workflow and problem Building it is getting easier The feedback loop is the more intense component
Understand how to read SDK
How to connect and MCP
Identify when to build it, how to build it, test is and now it’s working as intended
Connectors are Coda packs that have been renamed - we have a good idea of what a Coda pack looks like. We might have one accred for that.
The engineering team thinks our partners are going to be built by partners more complex - hundreds of lines of code pack with a sync table or MCP or both - submit a 5 min loom and walk us through what you built and why Choose your tool of choice Agents - the patterns are less known and murky now
Engineering thinks that agents will be built by the user because of the agent builder - level 100 Eric doesn’t see us building a complex agent for a year our agents are not going to be as robust as Zapier agents identify 1 use case that is a good fit - build an agent with 1 skill what they can’t and can do.
Resourcing?
Who can grade them?
TA or SA
Doctorate
Resources