Europe is completely missing out on the AI agent coliseum. We’ve got @bot being chaotic, @NousResearch and @openclaw throwing hands, and @Muse is just sitting at the border waiting for a visa approval, likely related to EU AI Act and GDPR.

Human jobs will rise from execution to arquitects/reviewers in the next months.

One of the favorite solutions among client projects have been linked to finding expensive mistakes instead of automations.

If slowing the AI labs that take safety most seriously doesn’t stop the race and instead helps the least responsible labs win, why do it?

If the goal is genuinely to minimise existential AI risk, under what assumptions does weakening or slowing the most safety capable frontier labs make humanity safer, rather than simply transferring technological leadership to actors with weaker safety standards and constraints?

Ox Alpha is a great model, close to frontier. Across EQ challenges, agentic workflows, and business doc analysis, it gets slightly beaten by Opus 5 and Fable, and half the time, even lighter plays like Kimi K3 and DeepSeek Flash get close to it. What are your thoughts?

Leveraging the latest llm's in real projects and learning on the side to complement the gaps, truly is one of my favorite ways to learn nowadays

Most people prompt AI like they’re writing legacy code, specifying the method, not the intent. German military doctrine ("Auftragstaktik") solved this 100 years ago: define the goal, let subordinates adapt. AI isn’t just engineering. It’s applied psychology.