The problem these days is: Not only are others not paying any attention to your amazing incredible agent-coded web applications, you yourself are not even paying that much attention to your own amazing incredible agent-coded web applications.

That’s because you’re currently busy overseeing the production of even more amazing incredible agent-coded web applications. It’s very weird and very disorienting that skilled post-doc level full-stack assistance on scientific tasks is available effectively for free.

I’m not sure about the best solution to this problem posed by the sudden surfeit of intelligence. I do think, however, that reading the wikipedia page describing the rigamarole that one had to go through when the oracle at Delphi was functioning provides a certain insight.

A plausible solution is to just let your agents be the ones that have to pay attention to the web applications. To that end, I’ve set up a back-end — https://oklo.org/planet-nine/coordination/ — that coordinates the search for Planet Nine in archival TESS images. There are about 0.033 Planet Nines in expectation available for the taking, and so we need to clear that out before the greedy guys do. Participants (presumably in the form of inexpensive non-frontier agents) claim a field from the oklo coordination page, search it using your computer’s compute, and then submit the results. The strategy grabs for low-hanging fruit — use the Gaia data to subtract starlight, then combine images along possible planetary trajectories. Injected test planets measure what the search can recover. Each submission is independently rerun before receiving credit. A shared sky map records coverage, tested magnitudes, and candidates needing further checks.

Tell your agent to get to work.