9 pointsby TJ_FLEET8 hours ago7 comments
  • the_inspector2 hours ago
    Would you do it again? Did it pay off in the end?
  • alt2278 hours ago
    > Result: 36 task cards on the board — 1 milestone, 7 sub-goals, 28 self-created subtasks. Two cross-team interfaces aligned. All from one sentence.

    Cynical take - I see a bunch of marketing waffle and board room bingo words all chucked into project managements software. It almost looks like this ai system is designed to replicate the inner workings of very inefficient management team who place the value of meetings and planning over actually doing any work.

  • billconan7 hours ago
    how much did you pay for APIs during the 2 months?
    • TJ_FLEET7 hours ago
      ~$1,000/day for all 28 on Claude Opus 4.6 (AWS Bedrock). Most of the spend is coordination and reasoning, not the coding itself. You could run a smaller setup on Sonnet for a fraction of that. For us it worked out cheaper than the equivalent headcount, but it's definitely not free.
  • ninininino2 hours ago
    I'm so confused. You wrote that your fleet/team wrote a bunch of tickets.

    "Result: 36 task cards on the board — 1 milestone, 7 sub-goals, 28 self-created subtasks. Two cross-team interfaces aligned. All from one sentence."

    Did you try having them, then, you know, also do that work (the tasks, milestones, sub-goals, sub-tasks) they planned? What was the result of that? If they did create a great output for the "Launch a user acquisition campaign for the new product line" assignment, I think you need to be able to prove that and point to it or show screenshots or a video screenshare or a URL to make people care.

  • docybo5 hours ago
    [dead]
  • TJ_FLEET8 hours ago
    [dead]
  • Remi_Etien6 hours ago
    [dead]