X Web Viewer

As agentic automation helps developers get more done, the role is evolving toward higher-level decisions about what to build, how systems should work, and where taste and creativity matter most. Databricks Co-founder @pwendell, OpenClaw creator @steipete, and Codex engineering lead @thsottiaux discuss what that shift means for enterprises building, governing, and scaling fleets of AI agents in production. Watch the full conversation:

Banger paper on harness continual learning. (bookmark it) If you already are allowing your agents to rewrite their own prompts, skills, or memory files, this one is worth your time. (bookmark it) Continual learning has always tracked what changes in the weights. Modern agents accumulate experience in the harness instead, across prompts, memories, tools, skills, and routing rules. What this means is that if you update any harness component, previously reliable behavior can break with the model completely untouched. The paper names that harness-level forgetting and provides a way to measure it. Guarded harness evolution separates proposing an update from committing it. A Continual Optimizer drafts a candidate harness from post-execution feedback, and a Continual Evaluator commits only after checking current improvement, historical retention, and validity. Relative gains exceed 10% across textual reasoning, multimodal perception, and open-world interaction. Paper: Track more trending AI papers in our academy:

MR Tony Elumelu is Chairman and not CEO; he’s exempted from all their HR first name rules. He’s the board chairman. It’s just his accessibility that caused it. Never get familiar with greatness. Never!

NASDAQ COMPOSITE UNOFFICIALLY ENDS SESSION LOWER BY 272.67 POINTS, OR 1.04 PERCENT, FINISHING AT 26,058.42 S&P 500 INDEX UNOFFICIALLY ENDS SESSION LOWER BY 64.33 POINTS, OR 0.83 PERCENT, FINISHING AT 7,643.65 DOW JONES INDUSTRIAL AVERAGE UNOFFICIALLY ENDS SESSION LOWER BY 669.35 POINTS, OR 1.25 PERCENT, FINISHING AT 52,793.70

The most important agentic workflows are the ones where you collaborate with the agent, verify its results, and encode them as a skill or verifier for reuse. This works for things like writing, researching, coding, and other tasks. Domain expertise, in the form of human verification, is a crazy moat. Don't let anyone or any company tell you otherwise. And you can build incredibly valuable services and products with that. Protect it and don't give it away for free. Keep upskilling yourself and leverage the AI agents along the way. But don't forget how crucial it is to develop and hone taste, judge quality, and critical thinking. In simpler terms, don't offload understanding to your AI systems. Offload all the rest (boring and repetitive tasks). Careful automation goes a very long way. A lot of the narrative today is around eliminating the need/replacing domain expertise. But if you work on hard problems, which you should be doing with AI, you realize quickly how primitive AI models are in their "intelligence", capabilities, and adapting to extremely hard and important domains. There is a reason why math problems continue to be solved only by folks with deep math backgrounds. Learn from that. I am not saying the models won't get better. They will get better, but so will humans (it's important to be an optimist in human intelligence for this to be crystal clear), and so the agent-to-human relationship and interactions are the real moat and where all the value and discovery will come from.