derCeoz @CeozX

Sunday means release day? :) MyBots pushed to GitHub. Install guides for humans and agents. Point your @NousResearch Hermes Agent at it. github.com/Fr3akaz0id/myb…

0 0 0 16 0

View Details

derCeoz @CeozX

2 weeks ago

@witcheer Want this on Android? You know, we all have unused phones :) Will open source soon, fixing little qol things before release. x.com/CeozX/status/2…

This is what i am cooking. @NousResearch Hermes agents that you can talk to. One of my bots teaches me languages. Uses the configured TTS and SST of your profiles. Thanks to @fayazara and @adolandev for the sdk and the robots :)

1 0 2 70 0

0 0 0 1 0

View Details

derCeoz @CeozX

2 weeks ago

@NousResearch @fayazara @adolandev STT lol. Oh did i mention that they react to "Hey Hermes" as well as "Hey Bot" when the wake is active?

0 0 1 32 0

View Details

derCeoz @CeozX

2 weeks ago

1 0 2 70 0

View Details

derCeoz @CeozX

2 weeks ago

@HermesWatcher Always open

0 0 0 13 0

View Details

derCeoz @CeozX

a month ago

@sudoingX Endless tokens and uncensored models

0 0 0 19 0

View Details

derCeoz @CeozX

a month ago

My arbiter fork now supports http for MCP usage. Had not enough vram on my Spark so it needs to live on another host. Give it a shot :) github.com/Fr3akaz0id/arb…

0 0 0 10 0

View Details

derCeoz @CeozX

a month ago

@0xBakeer created a Jev alternative that works very nicely. I forked it and added an OpenAI compatible endpoint, told my Hermes via auxiliary model setting to use it for approvals. Now only ESCALATE ones reach me (or if arbiter is down). github.com/Fr3akaz0id/arb…

0 0 1 163 0

View Details

derCeoz @CeozX

a month ago

This is exactly my experience. I play with whatever comes out, test it in real work with Hermes but always come back to Qwen.

Sudo su @sudoingX

a month ago

i've run models on everything from a 1660s to an h200. after all that, the most reliable lab i've found is alibaba's qwen team. i would say the most consistent results. qwen 3.8 27b dense and 3.8 flash next just do the thing and finish the task. i've run deepseek, glm, ling.

22 8 243 21K 57

0 0 0 11 0

View Details

derCeoz @CeozX

a month ago

Bonsai 2 is such a nice model

0 0 0 10 0

View Details

derCeoz @CeozX

a month ago

just look what @imbabybrooklyn cooked :) beautiful! x.com/imbabybrooklyn…

Brooklyn! @imbabybrooklyn

a month ago

One composer to rule them all; multiple tabs w/ multiple sessions. Just hover over the session and type. @NousResearch

26 6 249 35K 89

0 0 1 3 0

View Details

derCeoz @CeozX

2 months ago

@0xBakeer Qwen3.8 Flash Next on one spark. @MiaAI_lab vllm recipe as daily driver for @NousResearch Hermes agents. Few other Hermes powered by 3.8 27B running on 5090, 5060Ti and 3090. Can't complain.

0 0 1 123 0

View Details

derCeoz @CeozX

2 months ago

@DarioAmodei lol sure we are all doomed

0 0 0 6 0

View Details

Ahmad @TheAhmadOsman

2 months ago

Cry me a river you trained on humanity’s knowledge

Anthropic @AnthropicAI

2 months ago

We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them. We disrupted every operation in the report,

3K 12K 50K 43.9M 41K

29 64 1K 29K 29

View Details

derCeoz @CeozX

2 months ago

It happened! @seanphan finetuned a DFlash drafter for Qwen3.8-flash-next! First thing to try tomorrow :) x.com/seanphan/statu…

Sean Phan @seanphan

2 months ago

I finetuned a DFlash speculative-decoding draft head for nvidia/Qwen3.8-Flash-Next-NVFP4 - 78.3 tok/s on math (+36.1%) - 58.7 tok/s on code (+12.0%) - 37.8 tok/s on chat (-13.5%) Per workload, block 5, median tok/s against native MTP k=3 on the same prompts on 2x DGX Spark

7 12 111 9K 71

0 0 0 9 0

View Details

derCeoz @CeozX

2 months ago

herdr got an update, so helpful! have your herdr sessions from all machines on your main system. love it! x.com/herdrdev/statu…

herdr @herdrdev

2 months ago

0.9.0 is here, and it brings the most wanted herdr feature: all your machines running herdr, in a single client 🎉 control your agents and projects across local and remote machines without jumping between terminal tabs.