I like this view from the brilliant Terence Tao. I have always believed that AI could help to augment us. We built and train agents like tha…

I like this view from the brilliant Terence Tao. I have always believed that AI could help to augment us. We built and train agents like that today. But all forms of intelligence can and should be possible to coexist without the need to compete with each other. Higher and more advanced levels of intelligence can emerge and compound, unlike anything we have seen. It won't happen with the way agents are trained and built today, but it's a good starting point. I wish we would all train and build with AI agents more in this way. I can see a world where we orchestrate agents to augment humans and train agents to augment other agents as well. This is how I have been trying to build my agent harness over the past couple of months. I don't interact with agents linearly anymore (most coding agents and chat apps are built this way). Instead, I interact with the agents (providing important context and guidance), apply the output to the real world, observe impact via metrics, document the lessons learned (as external memory), build LLM knowledge bases to teach/augment the system further, adjust and augment my understanding of problems, and run this in a loop. It's a more involved process, but it's one where the goal is not competition, but a push for deeper human-agent collaboration for maximizing the highest quality of output. We are so early on this. Historically, hybrid systems can be pushed further than the individual ones. I sense that this will also apply to different forms of intelligence (and we have yet to see all the forms). Go build your knowledge bases, go train your agents, and always be curious, learning, and upskilling. Terence Tao proposes what he calls a "Copernican view of intelligence". Instead of buying into the common, one-dimensional narrative that artificial intelligence will simply evolve from "subhuman" to "superhuman" and ultimately make humanity entirely redundant, Tao urges us to lo…

Related

• Another banger paper from Microsoft. Why it's a big deal: It teaches reasoning models to compress their own chain-of-thought mid-generation.…
• The most important post I've read all week by @hwchase17 . Memory, particularly memory consistency is the biggest performance inhibitor to A…
• What if owning the memory layer goes far beyond escaping lock-in? What if memory becomes fully editable intelligence? You could debug bad pa…
• Portable agents

Source: DAIR.AI (X) | 2026-04-13