Value-First AI Daily - Jul 24, 2026
July 24, 2026
The show's first Friday at its new 1:00 PM Central time opened with a live Top 3 - Google ATLAS, FLUX 3, and HubSpot's Agent Hub - then an honest reckoning when Chris and Nico caught the list missing Claude Opus 5 and ranking HubSpot first, a bias Chris traced on air to the brief he wrote. The through-line: what an AI-native operation actually costs, and why the meter is yours to build.
Moments from this episode
2 short cuts from this conversation.
Key takeaways
The largest grounded study of real AI use (Google ATLAS, 15 million interactions) puts AI at about 21% of tasks in a typical job, with fewer than one in ten interactions ending in full automation - size an agent program to collaboration, not replacement.
A show grading its own tools: the sealed Top 3 missed Claude Opus 5, the day's biggest release, and ranked HubSpot first; Chris traced the bias, live, to two lines in the brief he wrote for the curating agent.
An AI operation's cost is only a black box if you let it be - the token numbers are already tracked on the backend, and the meter is yours to build; the honest work is naming which numbers are measured, estimated, or missing.
Dharmesh Shah's public correction from token-maxing to value-maxing (return on token) lands as a value-first idea: spend tokens freely when the return is real, never as a vanity metric.
Model movement: Opus 5's front-end coding benchmark jumped sharply and runs faster at low effort, but the bigger signal is what a Haiku-5-class small model would unlock at that cost.
Keep learning
You just watched the work. Now build it.
The next step is doing it yourself — live, with people on the same path.
Get the next conversation in your inbox
One signal-dense note when a new episode lands. No noise.