Reads like a LOT of tokens to me. What does your usage /workflow look like? I'm v curious because although I do use Claude code, my token counts aren't nearly as much

I want to know if I'm missing something cool!

Not OP, but I routinely load 150k tokens into context. A full sub-package to work on, select other files in the monorepo, e.g. front-end visualization and back-end data loader. Then work some 150k tokens, then start again.

At the end, cache hit rate is like 99.5% if Novita is not having issues.

For official DeepSeek API, 99.9% or something.

Custom harness that never compacts or otherwise doctors the history.

Those numbers make sense to me...120 million input tokens is like 120 sessions of hitting the full context limit, which seems like a lot to me though

When you have a general idea of what smells bad vs what's okay...why?

I'd rather get it from the LLM and review

They report an hour less than average sleep time the first 3 months?!

How did get so lucky?

That’s probably about how much I lost taking care of ours. She generally woke up twice to drink milk but I was still up for the first one. Loss of an hour sleep can still be brutal if its like 6.5 to 5.5 hours.

Possibly sleeping longer to make up for it? In the early days you might lose up to an hour per night feed but if you go to bed a little earlier and have a sleep in then you should hopefully be able to mitigate that two hour decrease.

Out of curiosity, what kind of custom harness are you thinking?

i do the tmux pane interoperability thing. i released to github yesterday tmux-browse.

its abstracted so it doesnt look like a claude code harness per se but it works like one.

docs/recipies.md shows the ralph loop

https://github.com/itsmygithubacct/tmux-browse

I do something tangential. If you can pick out identifiers for these pages with deterministic features (financial pages have the most numbers on the page, or have the word "director", "general manager" and "managing director" on the same page, etc)

Pick those pages out and pipe only those to the llm

Luckily I work with excel w lots of sheets and don't need to do pdf to text conversion etc

Scale, my friend, scale...

Interesting, I'd love to know more. Are parts of it public?

Indeed, I have it open source, but want to preserve my anonymity here. The main gist of it is Quartz as a static site frontend bundle, backed by Decap as an editor, so that non technical users can edit documents. The validation is twofold - frontmatter is validated by a typical yaml validator library, and then I created markdown body validation using some popular markdown AST libraries, so there are two sets of schemas - one for the frontmatter, one for the body, and documents must conform via ci. I ship it with a basic cli that essentially does validation and has a few other utilities. Not really that much magic, maybe 500 lines of code or so in the CLI and another few hundred lines doing validation and the other utilties. It's all in typescript, so I use the same validation in Decap when people do edits.