Episode Details

Episode 74. Saturday, August 22, 2026. Two numbers landed on the same day and they tell the whole story: roughly ninety percent of executives say AI still hasn't improved productivity at their firms — and Nvidia told its largest customers the hardware underneath all of it is about to cost more than fifteen percent more. Jack King digs into why the productivity gap isn't an AI problem but a targeting problem, using Lowe's as the counterexample, then follows the cost thread through AT&T's routing playbook and into the quiet end of the one-giant-model-for-everything era. Plus: the Hollywood creatives now doing piecework training the models that displaced them — and why that lands on your marketing team as a disclosure obligation.

Today's stories:

• About 90% of executives say AI hasn't boosted productivity. Research by Mark Ma (University of Pittsburgh) across millions of Glassdoor reviews, thousands of corporate filings and hundreds of AI investment and layoff announcements, alongside an Atlanta Fed finding. Market reaction to AI-justified layoffs averaged near zero; more than half drew a neutral or negative response. Ma: using AI to justify cutting jobs is "a strategic miscalculation that cuts against the benefits of AI." — Fortune · Futurism · TechXplore
• Lowe's: shoppers who use Mylow convert three times more often. CEO Marvin Ellison on the Q2 call; online sales up 15.7% in a quarter where discretionary DIY demand was flat. Mylow Companion is now in store associates' hands. — Customer Experience Dive
• Nvidia warns customers of AI server price hikes above 15%. Bloomberg reports systems shipping early next year — including flagship Vera Rubin and Grace Blackwell configurations — will cost more than 15% more, driven by soaring memory costs. — Fortune · CNBC · TNW
• AT&T cut its AI bill 56% by routing to open-weight models. VP Mark Austin told The Information the company cut coding and advanced-task costs 56% for roughly a 2% quality drop, using LiteLLM routers to send easy work to Nemotron, Llama and Gemma. Open models now handle ~40% of employee queries; target is 60–70%. — PYMNTS
• A 27B-parameter agent beat the frontier at replicating research. Inherent, founded by Google DeepMind alumni, says its Faraday agent — running on Qwen 3.6 — outperformed Claude Opus 4.8 and GPT-5.5 at independently reproducing published paper findings. The lab left stealth in May with a $50M seed. — TechCrunch · Tech.eu
• The Hollywood creatives training the AI that replaced them. Writers, editors and VFX artists taking piecework labeling AI outputs via Mercor, Micro1 and Handshake. LA shoot days fell 48% between 2021 and 2025 (FilmLA); motion picture and sound recording emplo