Hey all - I made a thing with my clanker Kevin: Ototo (https://ototo.dev/) - yes this is clanker driven dev. The idea is to offload code exploration and other token'y heavy tasks to a smaller model to do the work and chuck the result back. This helps in a couple of ways (1) it saves tokens if you can offload to a cheaper or free local model (2) it can keep the main context clearer as it removes all the sherlocking thinking/exploration. It also saves more tokens than using anthropics own exploration agent.
I had originally played around with a graph based approach but found this to work better. Currently, I have this running with Qwen 3.8 27B on my framework (I had Kevin build me a custom inference engine to squeeze all the juice I could out of the AMD - https://gitlab.com/handmadedigital/projects/shocho).
It's been through a bunch of tests (results are available if people want) against a load of different repos, including comparing subsequent commits of large repos with claude, claude + exploration agent and claude + ototo and so far ototo seems to really help.
The site is very 'LLM'y at the moment - I haven't had time to edit it to make the text more readable for us humans. Sorry.
Hope it helps folks in their day-to-day.
--EOL
Comments URL: [https://news.ycombinator.com/item?id=49971028](https://news.ycombinator.com/item?id=49971028)
Points: 1