Agents
This is as bad as saying there is no alpha in customizing your own models. I don't think people realize how much harnesses and models go han…
This is as bad as saying there is no alpha in customizing your own models. I don't think people realize how much harnesses and models go hand in hand. And thus the rest of the intelligence stack. Not
This is as bad as saying there is no alpha in customizing your own models. I don't think people realize how much harnesses and models go hand in hand. And thus the rest of the intelligence stack. Not attacking anyone here, but this kind of thinking is too shortsighted. This makes me question what exactly people are trying to build with AI or what they expect to build with AI. If you work on ambitious stuff (not all of this short-term AI slop), you quickly realize how significant it will be to own the intelligence stack. Harnesses, like evals and models, are things you want to own. Even if models can generate harnesses on the fly, which some can already, there are customizations that you simply cannot afford to offload. Not to mention that when RSI arrives, you don't want to sit and wait for others to build things for you. You must build it, and it's in your best interest to own it. The goal for everyone building with AI is to build like a frontier lab. Don't worry about costs; that will get sorted as intelligence becomes too cheap to meter. There is no alpha in building your own agent harness. The best techniques will get discovered, copied, and eventually incorporated into open-source harnesses like Pi. A few months after that happens, these techniques will be made obsolete entirely as labs release smarter models +…
Related
- Highly recommended. I've often claimed there's huge alpha in building agent harnesses. Turns out harnesses are compositional generalizers. T…
- i haven't seen a model that just works across agent harnesses. seems like it should exist. great opportunity for open-weight models. any tho…
- // Agentic Harness Engineering // Pay attention to this one, AI devs. (bookmark it) Most coding-agent harnesses are still tuned by hand or b…
- Impressive new paper from Microsoft and colleagues. Harness design is still hand-tuned almost everywhere. This work present an automated loo…
- // Scaling Laws for Agent Harnesses // If you build agent harnesses, this one is worth your time. (bookmark it) Most harness tuning treats e…
Source: DAIR.AI (X) | 2026-08-27