@GarrettLord@levie This feels like the quality assurance function of the future. Assuring quality and security in non-deterministic systems is going to be constant challenge.
4/ When the quick is gone from quick and dirty, what are you left with? Most teams choose to address it after the next release. And then the next release. It doesn't have to be a big bang — but the conversation has to start.
1/ A scaling startup Founder recently said to me: "I know we cut corners to get releases out. I just don't know where — or what I should be worried about as we grow."
3/ The founder usually feels it first. The monolith gets heavy. Features that were cranked out in a weekend start taking weeks. Releases require a meeting to ship. Sprint planning gets longer. Defects show up in production.
@chamath Why not drop the per user pricing ($200/user/month) and make margin on the passthrough of the token (usage) costs? This would allow for wide adoption across companies of all sizes.
@hamudinaanaa Exactly. Sequoia nailed it: copilots compete on software margins. Autopilots win on services margins.
But in product dev consulting we hit the same wall: clients measure Copilot *usage*, not value.
No baselines for velocity, cycle time, defects, or sprint predictability — let alone revenue or CSAT impact.
Usage ≠ outcomes. This is the next big bottleneck.
Your Outcome Primitives paper is timely. How are dev teams actually building those baselines today?