[spilldahill]

this is a solid articulation of the tension between waiting for tech to be “ready” and building toward where it’s clearly headed. there’s a lot of leverage in being early with infrastructure that assumes capability will catch up, especially when it’s not tied to any single model or vendor.

what’s interesting is that agency, not just reasoning, might be the next unlock. right now we think of LLMs mostly as assistants, but products that let them perceive and act in the world could shift what “ready” even means. sometimes the model doesn’t need to be smarter, it just needs better tools.

[dacke]

Really late at responding here but I agree on the idea that agents lock up new frontiers for improving AI-powered products/systems. I think that the principle of going towards where tech is heading applies not only to the LLMs but also to the surrounding tools, infra and plumbing. You might not have skills to leverage LLMs yourself today but have some kind of unique user insight that will be easier to translate into a product soon.