redshift's avatar
redshift
redshift@routstr.com
npub1ftt0...edsh
building Freedom AI Tech aka Routstr Chewing Glass Everyday Crunch Crunch Crunch Checkout https://routstr.com Stay humble stack sats
redshift's avatar
redshift 2 weeks ago
Kimi K2.6 seems lobotomized. Everyone is using 4 bit quantization and offering at prices comparable to 8 bit quantization. Users aren't sophisticated enough to choose the right providers yet. Routstrd chose @Private Provider Routstr AI as it has the best price, they proxy to @PayPerQ, which in-turn proxies to OpenRouter (correct me if i'm wrong) and OpenRouter just sorts by price/some custom algorithm. And we don't even know if the providers are lying or not. Kimi K2.6 is a good model for medium level tasks. It shouldn't be this bad. (Btw GPT 5.5 is great). Any ideas on how we can handle this better?
redshift's avatar
redshift 3 weeks ago
Nostr powers Routstr to make it the killer ecosystem built on social rails It is hard to keep up with all the new models released by AI Labs while also being productive with your work. In the last 48 hours, we've had 3 SOTA models released by MoonshotAI (Kimi 2.6), OpenAI (GPT 5.5) and Deepseek (V4 Pro) (2/3 already available on Routstr, GPT 5.5 soon). With Nostr and Routstrd, you don't have to. Your Routstrd daemon automatically fetches kind 38423 events published by users (only me for now) recommending must try models. These models show up on your OpenCode/Claude Code/Pi agent harnesses automatically as your routstrd daemon updates them every now and then. Here's the event I posted 10 mins ago: