Founder https://nitter.cf/t.co/ZW1BOYXKUg AI agents · local LLMs · Fine-tuning Deploying Hermes agents from @nousresearch in 🇫🇷 SMEs
Rouen, France
Joined June 2024
- Tweets783
- Following672
- Followers885
- Likes4.5K
Pinned Tweet
I built airlock, a security plugin for @NousResearch's Hermes Agent against prompt injection.
A web page hid instructions for AI assistants. My agent read it and tried to post to a webhook. airlock stopped it at the approval prompt. 👇
Otsoa retweeted
thank you everyone 🤍
join our Hermes Mobile app waitlist in the post below to be added as soon as we have more seats
Thank you to everyone who traveled from across the world to NousCon 2026. We loved spending time with the people who make the Hermes community great.
As you all know, we released the Hermes Mobile beta at the event, and as anticipated it filled up quickly. We are working tirelessly to free up more room in the beta, and to get the app production ready for wider release. The feedback has been incredible so far, and we are excited to share it with the world soon.
In the meantime, please join the waitlist at hermes-agent.nousresearch.co…
Je pense que je vais recruter fissa car là si ça continue sur ce rythme ça va être impossible à tenir !
C'est quoi vos reco pour trouver du monde dans l'agentique ?
Looks like something is happing for Taskaway on @crunchbase 🤯
Is anyone knows how is it working ?
Ma Saison 1 sur Promethee : 66 h 34 min de focus. promethee.io
Et si l’IA n’avait pas besoin d’être géante pour comprendre le monde ?
Depuis quelques années, une règle domine l’IA : plus un modèle est gros et plus il a vu de données, plus il est performant. Un papier de @randall_balestr et @ylecun, publié fin octobre 2026, propose de nuancer cette règle.
Ce que ça pourrait changer :
Des coûts en baisse : si la qualité de l’entraînement compte plus que la taille, on pourra obtenir des modèles performants beaucoup plus légers (enfin)
Du multimodal plus accessible : il n’est plus forcément nécessaire de constituer de grandes bases de paires image-texte, qui coûtent cher à produire.
Une autre voie que les LLM : c’est un argument solide pour les "world models" que LeCun défend, notamment avec @amilabs.
Yann LeCun and Randall Balestriero just Proved the Platonic Representation Hypothesis, and found a better way to achieve it.
The Platonic Representation Hypothesis suggests that AI models trained on different modalities, like images and text, eventually learn similar representations of the world. But this paper shows that how models are trained matters more than simply scaling them up.
So they mathematically show that, under certain conditions, two JEPAs trained completely separately can learn essentially the same underlying representation of the world. One learns from different images of the same scene, while the other learns from different descriptions of that scene, without ever being trained on image-text pairs.
Their approach achieves the same image-text alignment with 42x fewer training samples than next-token prediction. A 109M-parameter text JEPA even outperforms LLMs up to 72B parameters on cross-modal alignment benchmarks.
alphaxiv.org/abs/2610.39302
You gave me advice on this. I shipped the new taskaway homepage.
Roast it 🔥
What's confusing, what's weak, what would make you bounce in 5 seconds? 👇
Hermes browser extension v0.4.0 is out, 27 tools, vivid UI and gateway support. Go try it 🪽
hermes browser extension v0.4.0 🪽
a new hermes plugin with 27 tools for browser context + hermes control
→ vivid ui: cards, tables, charts
→ gateway support
→ fixes across chats, models, connections and safety
→ all new onboarding and loading screen
→ open hermes web in one click
open source and community based👇
github.com/abundantbeing/her…
The #1 pain point I hear from business owners and managers through Taskaway :
"I don't want another platform to manage. I want something that rolls out as simply as possible for my teams."
The winner of the agentic era will be whoever ships open-source solutions that truly hand the keys of agentic AI to companies, not yet another overpriced SaaS.
Don't you think so ?
