"For the breakouts, we spent a bunch of time ideating on what the best prompts should be for the teams to go deep on. We framed two questions:
  1. Tooling question: What's needed to make agents more productive?
  2. People question: What new behaviors do we need to adopt as a team?"
— Augment Code
"As agents get more capable, the bottleneck becomes less about writing code and more about two things:
  1. making it easier for humans to understand, plan, and ask questions.
  2. making it easier for agents to ingest all the real-world context surrounding your task."
— Scott Wu, Cognition

Providing async background agents real-world context (often cross-system/cross-service) w/o introducing context bloat is a challenge the whole industry is trying to solve. Many say the source of truth will be in the code, and that's why this service caught my eye: Sourcegraph. Growing context windows + smarter models will contribute to solving this as well. Having all that data in a context graph is a real & valuable win! But, making it useful for a specific task is a specific, separate challenge (introducing relevant Slack threads, JIRA tickets, Notion pages, and code references).

Here are some articles on Stripe's and Ramp's approaches: