Well it was fun while it lasted!...
Congratulations @rough__sea
I, along with the rest of the Deno company, am joining Cloudflare!
I'm motivated by abstractions that give leverage at scale. The Cloudflare workers programming model is a hermetic computing abstraction that truly embodies the old Sun Microsystems tagline "The Network Is the Computer". The programming abstractions that Kenton designed really force scalability. If you squint your eyes you might even conceive of it as a distributed high-level operating system.
The only major problem with this distributed operating system is that practically people only use it inside of Cloudflare. Cloudflare is an amazing service for nearly all applications - but there are many situations when you might need it to run elsewhere.
I want to solve this problem and make the Workers programming model the default way you write server-side application code.
The time is ripe for new abstractions too - what software we run and how we run it are changing quickly with AI. The number of agents is exploding and I don't think they all need a Linux VM. A much more lightweight Durable Object would suffice for many applications. I'm super keen to talk to people who want to self-host agents at scale on DOs - msg me!
30
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
Readers added context they thought people might want to know
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-gri…
research.nvidia.com/labs/amri/proj…
tech-ish.com/2026/10/02/tav…
Chiranjeevi Karthik Kuruganti retweeted
Introducing GLiNER2.5-Decide, our new 340M parameter open weight, encoder-based decision model.
GLiNER2.5-Decide is built for fast, deterministic classification. The model evaluates a set of user-defined typed questions and rules, and jointly decodes their answers, returning structured decisions with probability distributions and confidence scores.
We evaluated the model’s performance on Fast Decisions, an unseen, internally generated classification suite based on 17 datasets testing real-world use cases across routing, triage, classification, sentiment, and content understanding.
Measured against similar decision models, GLiNER2.5-Decide leads in 9 of the 17 datasets, achieving the highest average score:
- GLiNER2.5-Decide: 60.1%
- SemIf: 56.4%
- JevK5: 57.5%
- Laya: 46.6%
This performance makes the model a strong fit for use cases like tool calling, model routing, browser and computer use, and LLM-as-a-judge.
GLiNER2.5-Decide’s lightweight encoder architecture makes it easy to fine-tune the model for specific tasks, while being efficient enough to run locally on consumer-grade CPUs or in air-gapped environments, giving users greater control over where their data goes and where the model runs.
To make building and experimenting with GLiNER2.5-Decide as easy as possible, we're also offering hosted inference. You can now use our API to run inference and fine-tune GLiNER models on specialized tasks right inside your own coding agent: agent.fastino.ai
As with previous models, we’re also releasing the model weights on @huggingface under the Apache 2.0 license: huggingface.co/fastino/GLiNE…
> jsonCraft is live in beta 🚗
Build tiny interactive ASCII worlds with JSON and embed them on the web.
The driving demo was inspired by the legendary @bruno_simon portfolio
Still early. Go play with it and break things!
Best experienced on desktop 🖥️
jsoncraft.com/showcase
Chiranjeevi Karthik Kuruganti retweeted
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Chiranjeevi Karthik Kuruganti retweeted
🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!
🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!
Check out the configuration details in our official API docs: api-docs.deepseek.com/quick_…
Just watched @geerlingguy take on the SBC hobby dying, and it hit home.
I’ve stopped stockpiling RPIs. Migrated to microcontrollers for the trigger/control layer + self-hosted scripts running as outsourced compute.
Cheaper, lower power, more reliable for real projects. The era of throwing a Pi at everything is fading.
What are you switching to?
Video : youtu.be/HeX22LnKdFY?si=WTM8…
42
Chiranjeevi Karthik Kuruganti retweeted
Reminder that this is the future of humanity if open source AI doesn’t win
Well we said it to stop at 5 messages, But agents thought they are vibing nicely and just continued on and on...
This quoted post is unavailable.
1
62
Chiranjeevi Karthik Kuruganti retweeted
New art project.
Train and inference GPT in 243 lines of pure, dependency-free Python. This is the *full* algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further.
gist.github.com/karpathy/862…
💡The power of OpenClaw shines when you use it to run background coding agents.
It’s 100% self-hosted, secure, and controllable from your phone.
Let’s say you have a VPS. You install Docker and spin up these three containers:
• Authelia (2FA in front of your apps + OpenClaw gateway)
• Caddy as reverse proxy
• OpenClaw
Now you control the whole setup using a Telegram bot. You also have two or more preconfigured routes protected by Authelia:
app1.mydomain.* → openclaw-container:3001
app2.mydomain.* → openclaw-container:3002
I don’t open an IDE first anymore.
I tell the agent what to build.
Whenever I want to see something running, I just message the bot: "Spin up the React dashboard on port 3001", and follow up through Telegram.
This is the closest I’ve gotten to having my own private AI dev team that never sleeps.
#SelfHosting #AI #DevOps #Docker #AICoding #Homelab












