@Da7_Tech I use OpenCodex as a local proxy to route models from different providers into Codex, then I use LazyCodex to orchestrate what model does what when I enter my prompt into Codex.
Apologies for post length; I'll try to keep it terse.
Back in July I publicly stated on the TC Discord that I'd be slowly departing from the TC ecosystem, including relinquishing mod status.
Yesterday I sent Chad B. a DM letting him know of my departure and to remove my mod status, time permitting.
In the past couple weeks, I've provided the Discord mods with detailed insights into scammer detection techniques and referencing a couple tools. There are some "new" mods there which were very eager to learn, so I've passed on some of that knowledge and building blocks. They also know my DMs are always open.
Over the years I involved myself with some of TC's technical aspects (for example, I authored the PAUSELPDEPOSIT-chain Mimir code, which allows nodes to inhibit LP deposits on a per-chain basis), but soon no longer.
I'll continue to support https://t.co/wISNstogRE as a committer/maintainer (since familiarcow is a busy bovine these days), and I'll keep https://t.co/ob4i9cHsJN online too (since it only costs $33/year). So I'll still be in the TC Discord for those reasons, but those reasons alone.
All said and done, I still believe in -- and will probably always believe in -- THORChain's core ethos: providing a decentralised, permissionless, open-source native asset swap protocol.
@evildojo666 they might have the brains to develop large language models and harnesses for them, but not to double check what timezone they are actually currently using..
@morganlinton The benchmark showing Low, Med, High is confusing due to the LOW/HIGH/MAX mapping in DeepSeek docs.
And the screenshot of the table makes it look like you ran 1x LOW and 2x HIGH, which does not make sense considering the 87-91-87 scoring on your bench 😓🧙♂️