08 — AI Hype, Real Productivity, and AI Spending

The viral tweet: where productivity gains actually go

Context: everyone talks as if teams were at peak efficiency, bottlenecked only by code production. Dax argues most software engineering environments aren't like that 40:57.
Most engineers just want to do their job and go home. Give them a button that does their work faster and the rational move is: same amount of work, cash in the extra time — not push the organization harder 41:40, 42:05.
The hidden cost: the few irrationally motivated people who guard quality are now "drowning in slop PRs," burning out, and leaving — OpenCode has hired several people fleeing exactly that situation 42:25, 42:41.
Motivation, emotions, and team dynamics remain a first-order factor in AI-era productivity 43:03.

Startups have it easier: exciting space, competitive hires, meaningful equity. Dax's long-held hiring take: instead of two okay salaries, "hire one person and combine the salary — you'll get someone that'll meaningfully change the direction of your whole life." OpenCode pays a lot and only needs ~20 good people, not a thousand 43:47, 44:16.
At big-company scale he has no good answer: past a certain size "there is just no good reason for someone to do much more than what their job strictly asks" 44:39.
The bottleneck was never typing: "Pre-AI I spent 95% of my energy thinking about what to do and 5% doing it. Now I spend 96% thinking and 4% doing. It's a 20% improvement, but day-to-day it feels as hard as ever" 45:23, 45:42.

Two narratives: (1) flexing — companies bragging about $10k/month per engineer as proof of being future-facing — "that's fake, that'll go away"; (2) reality — for companies with thousands of engineers, even an extra $1,000/month/engineer "breaks your whole budget" 45:51, 46:40.
The uncomfortable scenario: "there is a world where the net result of all these AI coding tools is the same amount of work gets done, but all the engineers are happier because their job is easier. That's not good enough for a lot of companies" — they'll say go back to typing 47:11.
Counter-pressure: restricting tooling is a retention risk at the top of the market — best engineers leave over bad tools (the Jira effect). But at most companies the reality is simply "use Copilot, your limit is your limit" 47:47, 48:12.
On the narrative that these restrictive companies will die from falling behind: "I just don't think it's that straightforward" 48:45.