🚨 NVIDIA just shipped its own plugin for Hermes Agent !!
Your local model answers first, and the cloud model only takes over when the answer fails a check.
@grok verify this and explain why local-first saves money, in one sentence
Keep Opus 5.5 as your coding default in Claude Code. On 52 coding tasks in Claude Code 2.1.322, Fable 5.2 solved about as many (82.7% vs 84.6%) but cost 2.80x as much per solved task.
Opus already solves most of these tasks, though, so telling the two apart will take harder ones.
https://t.co/SfIvr9pHNg
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: https://t.co/iAOANQTVZ6
Jensen at #FireworksForge: training is becoming learning. Models shouldn't stop learning after they ship.
Do this now: run post-training as a recurring pipeline, version every eval + checkpoint, and sandbox agents before prod.
@nvidia@FireworksAI_HQ
@JensenHuang live at @FireworksAI_HQ Forge
Predictions for next year - Ontologies to bring enterprise knowledge - Open Shell for safe and secure agents - Hardware in the loop
This is absolutely insane. 🤯🎃
First the Hermes app beta. Now a huge browser extension update. Halloween is handing out treats.
v0.4.0 just shipped, and a new plugin gives Hermes 27 tools to read your pages and work your tabs.
▸ Answers render as cards, tables and charts right in the side panel
▸ Watch chats you started in other apps live: tools, reasoning, subagents
▸ Approval cards show exactly what Hermes wants to do before it does it
▸ Projects, bot profiles and group rooms in the chat drawer
▸ Thinking streams live, then folds away when the answer lands
▸ Open Hermes Web in one click
▸ Brand new setup in 21 languages
Works in Chrome, Edge, Brave and Firefox.
What a way to close out October.
Setup in the first reply 👇
@jonkomet
https://t.co/gn9Kweq2Rn
This is f*cking insane. Anthropic Research Lead:
"Prompting is basically over
99% of our engineers already run swarms of 300+ self-improving agents. Now everyone is shifting to graphs"
In just 20 minutes, she explains how graph engineering turns separate agents into a system that improves itself
One Agent → Loops → Plan Mode → Graphs → Self-Improving System
Prompting was the old workflow. The graph around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the article below ↓
@demishassabis in WSJ “He thinks of AGI as a system with the full cognitive range of the human brain—not just some of our capabilities but all of them. Earlier this year, he felt AGI was still three to five years away. Now… two to three years. ”
https://t.co/6WTONQmguT
Your agents can stay home for this one.
Melbourne, join us Oct 1 for Agentic AI Melbourne: talks, demos and meeting the humans behind the agents.
@cre8minus1 and Michael Hotan will be there! We're helping cover food & drinks. Come hungry.
https://t.co/BBx7U1Hu40
Lunch plans, Melbourne?
We’re teaming up with @grafana on Oct 1 to build a self-healing model factory. Pure Python, AWS + Kubernetes, with complete visibility into the agent, tokens and infrastructure. No CI/CD or manifests.
https://t.co/j0bBk8jHDI