Show notes
Everybody quotes the failure stat. Almost nobody reads it right.
Alan Espinoza of Reconstructive AI returns for the full Texas Drive Time conversation, and he opens by reframing the number the industry hides behind. It isn't that seven out of ten implementations failed. It's that seven out of ten people struggled to use what you bought them. Different problem. No software fix.
Then the harder question: what are we actually training these models on? An architect, a GC, and an owner will each get a different answer out of the same tool, because each one is prompting it to defend their own…
Website: podcasters.spotify.com
