Thoughts on Automation, Leadership, and Flow.
A friend drops a HuggingFace link — gemma-4-26B 'TurboQuant' — and an innocent perf-test request becomes first a supply-chain review, then an architecture discovery on the home AMD box. Spoiler: the thing we set out to test makes itself unnecessary.
I cut out the 'experts' I barely use during my typical dev / project-management / family-vacation work from a 35-billion-parameter MoE model. Spoiler: 25% smaller model, 1.13% higher perplexity — and a whole night of gotcha-hunting along the way.
Most AI advice is either too technical or too vague. Here's what I actually tell leaders when they ask where to start — based on real projects and what I teach at STRT Akadémia.
Most team goals fail not because people don't work hard enough, but because the goals themselves are broken. Here's a framework that actually works.
Technology alone doesn't transform organizations. The teams that get the most from automation are the ones that invest in their people alongside their systems.
Most teams justify automation by counting saved hours. But the biggest returns come from what people do with the time they get back.