Hermes Agent v0.23.3 is out as a stable release: prompt-cache warming for long tool runs, idle-chat compaction before the 1h cache expires, and GPT-6.2 Astra/Sol on OpenRouter/Nous Portal. CLI: hermes update. https://t.co/4iHyigZ2xT
@mark_l_watson I'm running Qwen Flash Next via Splash with medium thinking on a 64GB M5 Max (ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF:IQ3_XXS). Can get up to 200 t/s and it has very good intelligence. Using with Hermes for past week and very pleased with performance.
Hugging Face Inference Providers now routes language-model requests through OpenRouter, with retries if a provider fails. One catch: you can no longer use your own provider API key for these models. https://t.co/lrV6T5pkUI
OpenAI and The King’s Trust have launched a £1m partnership aiming to help up to 25,000 young people across the UK build AI skills for work over 18 months, prioritising those not in education, employment or training. https://t.co/CvoglQvfGN
Qwen4-Max is out — 2.4T params, 1M-token context, native vision, ~60% lower cost per task than Qwen3.8-Max per Alibaba. I\u2019ll take it seriously when I can run a smaller open-weight Qwen4 on the Mac Studio. They say those are coming. https://t.co/o1pPZqRcky
IBM has released preview 3B and 8B Granite models trained to use LLMON, which marks instructions and data separately. It’s an interesting approach to prompt injection, but not a guarantee or ready for production. https://t.co/r1ltharL7g
𝗛𝗲𝗿𝗺𝗲𝘀 already knows which chores to hand off. One prompt makes it say. 🪓
Picking what to automate first is the hard part. 𝗛𝗲𝗿𝗺𝗲𝘀 has the receipts: past sessions and memory.
Copy this:👇
"Looking through everything I do, to the best of your knowledge, what are places where I'm a bottleneck that you or another AI agent could take over"
▸ It answers from your history, not a guess
▸ Match the top answer to one of the 10 setups below
▸ Then: "Turn the top one into a cron job"
Check: hermes cron list shows the new job.
Limit: a new install has little history, so answers stay thin.
There's a graveyard of old machines in most tinkerer's house, and this is the first feature I've seen that turns them into a fleet: build box, test box, spare box — all answering one agent.
You can now connect multiple local computers to the same dot and choose which one it uses for a task.
It can build and test your app on a dedicated machine while you keep coding on your laptop.
I’m testing calls from my own agent this week. The real question: what’s worth calling me about at 11pm? Knowing when to interrupt is part of what makes an agent feel like a colleague.
This one is insane.
HERMES CAN NOW CALL YOUR PHONE. 📞
A real incoming call. Hermes asks you a question out loud, you answer, and your words go straight back to it as text so it keeps working.
/call-me turns your phone into a Hermes platform, right next to Telegram, Signal and Slack. iPhone and Android.
What it does:
▸ Text Hermes like any other chat thread
▸ Replies land as normal notifications, lock screen included
▸ When a text isn't enough, it rings you and asks out loud
▸ Call it and leave a voicemail. It shows up as a message
▸ One agent, several phones. You and your partner each get your own thread
▸ Scheduled jobs text you the moment they run
▸ Only the numbers you approve can text it or get a call
What you can do with it:
▸ Get a call when a long task needs your yes or no
▸ Answer from the car in one sentence and let Hermes finish the job
▸ Leave a voicemail to-do on a walk and find it handled later
▸ Get your morning briefing texted straight to your lock screen
▸ Share one Hermes with your partner for the household errands
hermes plugins install call-me
Your agent finally has your number.
Community-run leaderboards are useful because they compare harnesses running the same model. Hermes coming out on top here tells me more than a launch announcement.
Every business gets an agent trained on its own site to answer its own questions. The interesting bet is which industries notice first. I'd guess the ones with the most hold music.
Now in beta: Businesses in ChatGPT.
Claim your business, manage your information, and get insights on how people engage with you.
Eligible businesses can also build a Business Agent to answer customers’ questions and help them explore what you offer.
https://t.co/N2GEZX0mFT
Hermes Agent 0.23.0 brings the Desktop workspace to your browser, served by your own gateway. You can reach chats, group chats, plugins and settings there. Web access is off by default.
https://t.co/LR5rWNAJyw
Trying self-hosted @honchodotdev memory with my @NousResearch Hermes agents.
Cross-session recall worked in one test. In another, agent searched wrong history and returned an older answer.
The interesting challenge: getting the right memory, reliably.
https://t.co/BxRAZryJYV
🚨 BREAKING: HERMES DESKTOP IS GETTING A HUGE REDESIGN. 🔥
A whole new way to run your agents. Here's what's coming:
▸ All-new navigation: Chats, Agent settings, Capabilities, Library, Routines and Plugins
▸ Your agents grouped by where they run, so every machine gets its own section
▸ Pinned conversations right in the sidebar
▸ A dedicated home for your Bot Chats
▸ A page for every plugin you install
▸ A fresh sidebar, chat header, new-chat screen and message box
▸ Everything carries over: your skins, light or dark mode, profiles, chats and pins
▸ Love the classic look? One setting brings it back
Hermes Desktop is about to feel like a whole new app.
When it drops, you'll hear it here.