The automated research loop idea is interesting applied to SEO too — most site owners never systematically audit what's broken on their site, they just notice when traffic drops.
I built RankyPulse (rankypulse.com) partly around this: run a technical audit, get a prioritized fix list, re-run after changes. The "autoresearch" parallel is that you'd ideally close the loop automatically rather than waiting for a human to notice regressions. Still mostly manual today but the pieces are there.
The hard part on the SEO side is defining what "better" means in a way that's measurable quickly enough to be useful. Ranking changes take weeks to show up, so the feedback loop is much slower than inference benchmarks. Makes it harder to know if a hypothesis was actually right.
The automated research loop idea is interesting applied to SEO too — most site owners never systematically audit what's broken on their site, they just notice when traffic drops.
I built RankyPulse (rankypulse.com) partly around this: run a technical audit, get a prioritized fix list, re-run after changes. The "autoresearch" parallel is that you'd ideally close the loop automatically rather than waiting for a human to notice regressions. Still mostly manual today but the pieces are there.
The hard part on the SEO side is defining what "better" means in a way that's measurable quickly enough to be useful. Ranking changes take weeks to show up, so the feedback loop is much slower than inference benchmarks. Makes it harder to know if a hypothesis was actually right.
[dead]