Dax sent an internal memo naming three challenges — none new, but all "turbocharged by LLMs" 48:47:

  1. Shipping features not worth shipping.
  2. When iterating, the original design turns out off, forcing something hacky — and the LLM absorbs the hack invisibly.
  3. Needing to spend more time cleaning things up.

1. Features not worth shipping

  • It's too easy to respond to any problem by shipping a feature. The fix is restraint: explicitly defining what kind of thing they ship now versus what waits for more clarity 49:19.

2. The muted prickle — hacks without feedback

  • The classic dilemma: the system doesn't support the new feature, so you either redesign from first principles or absorb a temporary hack. That's a judgment call — and "that judgment is so distorted right now because the agent will just do the hacky thing for you" 49:40, 50:09.
  • Result: OpenCode shipped "way more hacks in places where we should have rethought the whole system" 50:34.
  • The mechanism: pre-AI, writing a hack came with a bad feeling — a "prickle" — that accumulated with each subsequent hack and kept judgment sharp. Agents mute it: someone else deals with the problem, the landmines still blow up later, but you don't feel it, so the feedback loop is gone 50:56, 51:37.
  • Analogy: the CEO who delegates everything and never sees the shop floor versus hands-on leaders (e.g., Stripe's CTO periodically doing support/coding rotations). "All of life is about having the proper feedback loops in place" 51:57, 52:24.

3. Cleaning things up

  • The daily pressure is real: thousands yelling for features, new competitors appearing constantly — "if we let ourselves get pulled by those forces, I don't think it results in anything good" 52:46.
  • The silver lining: cleanup is easier than ever with agents. Pre-AI, a new pattern only applied going forward; now you can ask the agent to implement it everywhere, kill dead patterns, and pay down tech debt cheaply 53:10, 53:40.
  • Why bother when it brings no revenue: "there's a million ways to be a successful company… I want to make sure we build a place that we're happy to work at 5 years from now" — not one that can't hire good people because the job is slogging through legacy 54:01.

The kicker

  • "The worst part about all of this is I don't think we're trading this off to move faster. I think we're moving at a normal pace" — it feels fast, but they're not outpacing competitors, and it's easy to trick yourself about productivity 54:39, 55:01.
  • His counter-philosophy: prepare a lot, set the foundations, "and then when you spring, you spring with so much more force" 55:22.