If you have a problem with Autolith, please fill out a complaint form, and put it in an envelope addressed to the hospital you were born in.
Afterwards, join our Zulip and start a topic in the "Autolith Bugs" channel! https://t.co/zQe3fCalyT
@GlowingPsyop@LukasHozda pick your poison, I believe that qwen-27b is the bare minimum, heard people having good experiences with it, would be better if you could run something that's a lil bigger :)
@mont_py@LukasHozda Yes, likely, I released it today, and the build server (intentionally) lags behind the full test pipeline releasing. Also, what terminal emulator are you using? iTerm is by default pretty slow, but if you don't have issues with other agents doing fullscreen rendering, then hmm...
AL's new compaction works absurdly well. We produce structured checkpoints vi a cache-friendly forced-tool-call request:
- it contains decisions, constraints, completed work, next steps
- links to durable evidence
- stale facts detected on resume
- is portable across providers
@mont_py Very sorry that you are not having a great time!
You can also come to our zulip https://t.co/zQe3fCaTor, it is not standard to have so many issues, so you must have a very valuable setup for testing :)
@mont_py Hmm! Are you using opencode go, or zen? I only ever see Go users, usually with Deepseek 1017. This is the most popular setup when it comes to token count
Are you on Windows?
Would you mind showing me the errors/failures? :) we do really want to fix everything people run into
Theo claims LLMs were discovered when an autocomplete training run went way too long. Like someone left autocomplete on over the weekend and came back to ChatGPT. LLMs exist bc of the transformer and nobody found that by accident. What the fuck is he talking about