Posts tagged philosophy
-
Dead Text Looks Alive
my-ai-team kept fourteen prompt fragments that no agent ever read, and its own tests kept them looking current. Deleting them was easy. Keeping them deleted took a lint.
-
Activity Is Not Progress
A two-agent cycle in my-ai-team sat silent for about 36 hours while four recovery mechanisms each believed the stall was someone else's. The fix was one ladder that ends only on progress.
-
How to Delete a Role
Removing the refinement role from my-ai-team's delivery agents took two pull requests, because a capability is not deleted until the runtime stops speaking its vocabulary.
-
A Handoff Is a File, Not a Message
Why every Dev→Reviewer handoff in my-ai-team is written to a path and posted to the issue, instead of being typed into the other agent's chat.
-
Why my-ai-team Has an Agent That Only Talks and Never Writes Code
Thesis #1 — explore, and why the ticket is the membrane between a fuzzy idea and clean delivery.
-
Why a Ticket's "State" Is the Gate to Delivery
Thesis #2 — how ready / blocked / assigned_to turn a ticket's state into the gate a delivery agent obeys.