Pinned Tweet
We live in bubble on this app.
A lot of us are running agents, building workflows, creating skills, comparing harnesses, connecting tools, and etc..
Go ask one of your friends what Kimi or GLM is. They will probably look at you like a madman.
OpenAI’s consumer research found that 49% of ChatGPT messages are still people asking for advice or information.
Another 40% are requests to complete a task.
Meanwhile, 70.2% of sampled Codex users had assigned at least one task estimated to take a person more than an hour.
I’m sure your timeline makes it feel like everyone is running agents and orchestrating workflows.
They are not.
Do not confuse being surrounded by early adopters with being late.
You are early. Keep learning and keep building.
Hermes closing Jensen Huang’s GTC Berlin keynote 🔥🔥
A Microduck robot training demo with Hermes coordinating the models and simulation tools.
Huge moment for Nous and everyone building with Hermes. 🪽
Jensen closed his GTC Berlin 2026 keynote with this Hermes Agent demo as the finale.
He introduced it by saying the way to build an agent is to ask an agent to build it for you, so he installed and set up a customized Hermes in one prompt.
Then he used Hermes, running on @nvidia RTX Spark, to create a simulated training environment for the @huggingface Microduck. Hermes planned an obstacle course with Nemotron and had GPT-6 Astra build it using FreeCAD, Omniverse libraries, Blender, Newton, and Isaac Lab. Hermes trained thousands of simulated ducks on that course until the validated policy was ready to run on the real robot in the physical world.
At Nous, we believe open source is a practical tool as much as a philosophical ideal. It gives users the ability to build powerful, complex workflows chaining numerous local+cloud models, devices, and tools themselves instead of waiting for a company to enable them.
If you pay close attention to what's #̢̛ͥ̀̽̽ͨ̌̓͆̇̀̽ͦͨͫͯ̏̎̇̈̚͜҉̕͢͏͡҉̸̶̛͢͡҉̵̨̭̳̘̹̫͔͖̜̠͖̫̲̖̱̤̻̖͉͢͞͡ͅͅ͏̸̷̢͑̊̀̐ͣͭͭ̉͂ͨ̽͒̊̎͛ͪ̊̎ͩͧ̀͝͠͠͏̸̵̵̷̵ important, you'll see that Hybrid is always cooking at severe levels.
Congratulations fellas.
The more local models the better 🫡
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Sometimes people ask "What can you even do with hermes"
Just about anything truly.
Okay, this one's ridiculous.
Hermes can edit your videos in DaVinci Resolve now.
Hand it a folder of clips and a script. It builds the rough cut right on your timeline.
▸ Scans your footage and music
▸ Builds the cut from your script
▸ Adds markers for each beat
▸ Renders the final file
▸ Shows you the full edit plan before it touches your project
▸ On free Resolve, it hands you a timeline ready to import
One message. Rough cut done.
Mistral + NVIDIA are planning benchmarks for AI agents doing engineering work.
Can they use the tools, finish the workflow and get the result right? And how much time and compute does it take?
Would be good to see a full CAD to simulation task in there.
Today we’re announcing a new collaboration with @nvidia to bring AI to industrial engineering:
- faster simulation
- open benchmarks and supporting datasets for physics AI
- evaluations for agentic AI in CAE and EDA
More here: mistral.ai/news/mistral-nvid…
Perplexity Computer no longer needs the internet or paid credits to work.
Just a month ago I was asking if anyone was still using it them.
Well done @perplexity_ai
A fantastic collaboration!!
General Compute + Positron, targeting Q1 2027. Curious which models we’ll be able to run at launch.
Our heterogeneous cloud just got better. We're announcing a cluster scale deployment of @positron_ai as our next partner.
We believe demand for fast inference will outstrip supply for the next decade. Our bet on Positron is another step towards getting fast inference for opensource models into builders' hands.
Hardware goes live Q1 2027.
Thanks to @mitesh711, @trsohmers and the whole Positron team! We're excited for this partnership.
LFG!
Ornith is so underrated.
Their new 122B release has open weights under MIT, with some strong reported coding and agent benchmarks.
Anyone running this with Hermes yet?!
Aloha!🌺 Meet Ornith-1.5-122B: No Tradeoff Between Strong and Fast
Stronger. Faster. Leaner. Ornith-1.5-122B delivers frontier-level capabilities with exceptional inference efficiency. With superior instruction following, stronger task decomposition, and deeper understanding of real-world workflows, it achieves state-of-the-art performance among open-source models of comparable size, while running on a single NVIDIA DGX card!!
Open weights under the MIT license.
Coding
✅SWE-bench Verified: 0.845
✅TB2.1: 0.813
✅CyberGym: 0.800
✅NL2Repo: 0.512
Agentic
✅HLE: 0.446
✅BrowseComp: 0.785
✅MCP-Atlas: 0.746
✅Toolathlon: 0.708
🤗Hugging Face: huggingface.co/ornith-ai/Orn…
We can’t wait to see what you build with Ornith-1.5-122B.
Nebius announced $10M in support for European AI builders, compute grants, investor connections, and community programs with NVIDIA, Accel, and others.
The focus includes open models, coding agents, inference tools, and evals. The broader Builder Program is free to join, with credits, training, and technical guidance.
Any of my EU friends signed up yet?
The AI Builder Program is expanding across Europe with $10M in total support across compute grants, capital, and community programs.
Are you building open models, developer tools, or applications on top of them? The program helps builders go from experimentation to scale-up through:
- Compute grants for selected projects in areas such as open models, coding agents, post-training, and evals
- VC connections to help turn open source projects into companies
- Hackathons and demo days with communities including @AITinkerers, @techeurope_ , and @esra_robotics to build, connect, and share your work
We’re kicking off the European expansion this week at #NVIDIAGTC Berlin with @NVIDIAAI, @Accel, and a network of European VCs.
Building AI in Europe? Details below. 👇
You probably don’t need the most expensive model to change a button label.
Cognition just introduced Devin Coach, which checks your prompt before it runs. Vague instructions, unrelated tasks crammed together, or an expensive mode for a simple job? It suggests a fix.
They’re also adding an Efficiency Score based on estimated engineering work per unit of spend, including investigations and planning.
Enterprise teams can use that score to automatically give qualifying engineers more usage, within limits their admin sets.
What if your AI budget went to the engineers who use it best?
Blanket usage caps are a hiring freeze in disguise.
Introducing efficiency-based auto-approvals for your most productive engineers, plus Devin Coach to help everyone get there.
devin.ai/blog/efficiency-bas…
