T3 code is honestly the best experience if you are using a ton of the subs from a lot of labs and you have multiple machines. Not for everyone, but hard to justify NOT using it with these recent updates.
The Voice retweeted
512MB RAM once forced programmers to think carefully
memory was limited
CPU was limited
storage was limited
so software had to be efficient
then hardware became abundant
now 32GB RAM is normal and some apps happily consume gigabytes just to display text and buttons
we solved the hardware problem and accidentally created a software efficiency problem
It's the year of the local LLMs seems like. Hype if it actually is usable unlike qwen 3.8 27b.
Aloha! 🌺Introducing Ornith-1.5, a family of open-source LLMs spanning 9B Dense, 35B MoE, and 397B MoE, trained with self-improving strategies.
It achieves state-of-the-art performance among open-source models of comparable size and delivers performance comparable to Claude Opus 4.8 across reasoning, agentic, and coding tasks:
✅Terminal-Bench 2.1 (86.1)
✅SWE-Bench (86 on verified, 65.1 on pro, 79.6 on Multilingual)
✅DeepSWE (56)
✅HLE (44.6)
✅ClawEval (81.4)
✅Tool Decathlon (71.2)
Ornith-1.5 takes a major step toward training foundation models through end-to-end self-improvement, extending the self-scaffolding strategies introduced in Ornith-1.0 into a more complete self-improvement loop: the model proposes new tasks, generates task-specific scaffolds, and produces solution rollouts for reinforcement learning, continuously creating new learning experiences from which it can improve.
All models, along with their quantized versions (FP8, GGUF, MLX, and NVFP4), have been released under the MIT License, enabling unrestricted commercial and research use.
📘Tech Blog: ornith.ai/ornith_1_5.html
🤗Huggingface: huggingface.co/collections/o…
The Voice retweeted
Aloha! 🌺Introducing Ornith-1.5, a family of open-source LLMs spanning 9B Dense, 35B MoE, and 397B MoE, trained with self-improving strategies.
It achieves state-of-the-art performance among open-source models of comparable size and delivers performance comparable to Claude Opus 4.8 across reasoning, agentic, and coding tasks:
✅Terminal-Bench 2.1 (86.1)
✅SWE-Bench (86 on verified, 65.1 on pro, 79.6 on Multilingual)
✅DeepSWE (56)
✅HLE (44.6)
✅ClawEval (81.4)
✅Tool Decathlon (71.2)
Ornith-1.5 takes a major step toward training foundation models through end-to-end self-improvement, extending the self-scaffolding strategies introduced in Ornith-1.0 into a more complete self-improvement loop: the model proposes new tasks, generates task-specific scaffolds, and produces solution rollouts for reinforcement learning, continuously creating new learning experiences from which it can improve.
All models, along with their quantized versions (FP8, GGUF, MLX, and NVFP4), have been released under the MIT License, enabling unrestricted commercial and research use.
📘Tech Blog: ornith.ai/ornith_1_5.html
🤗Huggingface: huggingface.co/collections/o…
Qwen 3.8 27b sucks to use and to run; it's so benchmaxxed it's useless.
I asked for a simple network scanner; it took over 2 hours in Qwen code to make anything and failed the task. Changed harnesses twice and retested with jspace, deekseek harness it still not a useful model.
I was hyped for Grok 4.6, used it to make a basic port of Monks and Mystics from CK2 - CK3 and it just couldn't do the job at a polished level. Fable just did the work and figured it out.
This aged poorly given prime agent just came out
"Long-horizon tasks are still a joke. They do not work, and I do not care what anybody says. Do not show me a stupid evaluation. Do not tell me about some dumb script you ran for 48 hours. Long-horizon tasks are not handled well. They simply do not work."
- Chamath at Stanford AI Club
"Second, complex problems also do not work. They are neither addressed nor handled well.
Why is this important? If AI develops like any other technology, we are going to experience an initial rise—the hype cycle. Then, we will see a natural contraction because, somehow and somewhere, something is going to fail. We are all going to see this, and then we will enter what is called the “trough of disillusionment.” I think the business and MBA folks will confirm whether that is true.
Afterward, you typically see the slow and gradual adoption of the real, final solution. This happened with the internet, and it has happened in many other cases.
The problem is that we are spending hundreds of billions, potentially trillions, of dollars trying to figure out how to cross this chasm. So, what do we do?
If we do not figure this out, people will reach the trough of disillusionment and say that AI was a joke. I think we need to be able to bring AI into highly complicated environments and make it work.
What is my solution? At a very basic level, you need a symbolic space that guides the embedded space."
----
From "techniahqrobot" YouTube channel, (full video link in comment)
He used a local model and had an awful experience because local models that you can run on your 5090 are an awful experience.
No one is using local ai models outside of image gen and maybe writing for any serious work.
I'm sorry about a 27b model just isn't very good for coding work; at best it's ChatGPT 4o level, and anyone who tells you otherwise is coping.
I initially hated Grok Build as it was really dumb, but they did something from when I tried it last week, and now it seems to be able to get things done way better now; it's grown on me and might be my favorite tui atm.
Yesss
Fahrenheit is superior to Celsius and I’m tired of pretending this is a debate.
Fahrenheit feels like temperature was designed by an human being who walks outside, feels the air, and says, “Yeah, that makes sense.”
Zero is freezing. 100 is brutally hot.
70 is comfortable. 80 is warm.
The numbers feel like what they are describing.
Celsius feels like it was designed by a committee of people who hate joy and want every conversation about the weather to sound like a chemistry lab.
“It’s 22 degrees outside.”
What does that even mean to a normal person? Am I wearing shorts? Am I grabbing a jacket?
Nobody knows. You need a conversion chart and a government employee standing next to you explaining the vibes.
Fahrenheit gives you range. That’s the point. It gives you more numbers for the temperatures human beings actually live in. You can feel the difference between 68, 72, 76, and 80. Those numbers matter. Fahrenheit lets you describe the world with more precision without having to break into decimals like a lunatic.
And here’s the other thing...
Celsius is connected to the metric cult energy that came out of the French Revolution. And the French Revolution was evil. Both of the Revolution were to rip out tradition, faith, monarchy, hierarchy, the Church, the calendar, the clock, and anything that reminded man he was not God.
CELSIUS IS INHERENTLY EVIL
So yes, Celsius may be useful if you’re boiling water in a lab. Congratulations. But Fahrenheit is for living LIFE because Fahrenheit is for walking outside and instantly knowing what kind of day you’re about to have.
100 means hot. 70 means beautiful. Zero means don’t go outside unless you have a death wish.
That’s a real system.
I would 100% do something like this and my wife would have a similar reaction.
My wife has a strict rule: I am not allowed to be left unsupervised in public for more than twenty minutes. This weekend reminded me why.
I’m standing in a long line at a garden center with a massive 50lb bag of specialized bird seed.
This woman behind me taps me on the shoulder and asks, "Oh, do you have a lot of birds?"
Standard small talk, But I’d been in line for fifteen minutes and I was bored.
I looked at her, completely deadpan, and said: “Actually, no, This is for me, It’s a new holistic cleanse I’m trying, High protein, great for the hair.”
The guy behind her immediately stopped scrolling on his phone.
The woman’s eyes went wide. “You… you eat that? Is that even safe?”
I leaned in like I was sharing a government secret. “The results are incredible. I lost 30 pounds last summer. The only problem is the side effects. Last time I did it, I ended up in the emergency room for two weeks. Tubes, monitors, the whole deal.”
She gasped. “From the seeds? Did you get a parasite?”
“No,” I said. “I was perched on a telephone pole trying to sleep and a hawk knocked me off. Fell forty feet.”
The guy behind her made a sound like a tire blowing out. The woman just stood there, paralyzed, trying to process the mental image.
Right then, my wife comes back from the bathroom, sees the look on the woman's face, grabs my arm, and starts dragging me toward the exit.
“He’s kidding,” she tells the woman. Then she turns to me: “This is why we can’t have nice things.”
I don't care. The look on that lady's face was worth more than the bird seed. 😂
