𓁚 (ra) 🧙 (mage) in a land of fire 🔥 and ice 🧊 YEG. head of platform @ https://nitter.gq/t.co/EoBj1Mgsiv & https://nitter.gq/t.co/sZBDpW6bBB
Edmonton, Alberta
Joined January 2011
- Tweets4.4K
- Following772
- Followers282
- Likes1.9K
Ryan Ramage retweeted
To act for you, an AI agent has to read your email, calendar, messages and files. It ends up knowing more about your life than most of your friends, so you should get to choose where it runs.
Today we're releasing QVAC BotPsy, a 0.8B model that sets alarms, moves meetings and answers messages by calling the right tool. It's small enough to run on your phone, which is where an agent that knows that much should live.
Models: huggingface.co/collections/q…
Keet team keeps rolling out features at a steady pace
Keet 4.25.0 is out on mobile & desktop 🐦
📖 Share contacts right in chat
✍️ Private notes on contacts
🌐 Group calls auto-adjust video to your connection
👑 Promote members to admin
🎙️ Smarter call transcripts & summaries
Still zero servers. Still knows nothing about you.
1
85
Now is the time to get all your python friends to do p2p.
I did a python example of p2p with no node or sidecar. Just a bare worklet in a thread. Unlocks all the holepunch p2p goodness...
github.com/ryanramage/bare-k…
Ryan Ramage retweeted
Day 9/30 of one article per day until my Premium subscription expires.
Today, I explored @keet_io and its vision for a different kind of internet communication.
What if your messaging app didn't depend on a central server?
Ryan Ramage retweeted
This Financial Planning Day, let's talk about the true cost of "free."
Most apps are free because your personal data is the product. Every message, contact, and location ping is harvested for ad revenue. 💸
Keet is free by design. We connect you directly to the person you're talking to, no middlemen in between, no data collection, and nothing to monetize.
Ryan Ramage retweeted
Bluetooth discoverability is officially live across the Holepunch stack. 🔥
With local-first discovery and background worker daemons, devices can find each other and keep apps running seamlessly—which technically means you can now chase @mafintosh with questions even when he's completely offline.
Dive into the latest features and explore the code on GitHub: github.com/holepunchto/bare-…
really excited about autobee. Multiwriter p2p just much simpler.
We've been cooking in the background. 👨🍳
Meet Autobee, our new open-source P2P collaboration engine.
After months of testing real-world peer-to-peer workloads, @mafintosh and the team rebuilt our collaboration engine from scratch.
⚡️ Faster indexing
⚡️ Lower memory & CPU overhead
⚡️ Fully open source
Find out more here > support.keet.io/release-note…
37
Ryan Ramage retweeted
The biggest upgrade to Keet is officially live 🚀
We’ve rolled out our next-generation chat engine, built from the ground up to deliver a faster, more resilient, and truly unstoppable peer-to-peer messaging experience.
You’ll notice the speed most where the wait used to be worst: joining a busy group chat, pairing a new device, or restoring from backup. Group chats also stop getting stuck waiting for enough members to confirm the order, so a quiet or thinly attended group chat will keep working.
Each of your devices converts its own group chats the first time you open them, so a chat may take a moment longer on that first open. We appreciate your patience while the distributed network catches up.
One ask: make sure everyone in your group chat updates too, or they can get locked out of a room once it moves to the new engine.
Update, hop in, and open a big group chat. No server holds your conversation. Now its order lives in the messages themselves
Ryan Ramage retweeted
P2P OTA updates have officially landed for React Native. 🚀
With pear-mobile and pear-runtime-react-native, you can ship JavaScript bundle updates directly peer-to-peer without relying on central servers.
Unify your deployment workflow across desktop, terminal, and mobile today. 🍐
Find out more 👉 pears.com/news/pear-mobile-d…
Ryan Ramage retweeted
QVAC now has a Python SDK.
Not a port and not a subset. The Python client is generated against the same contract the TypeScript SDK speaks, and it drives the same worker, so the same call returns the same result.
Python SDK install command:
pip install tetherto-qvac-sdk -f github.com/tetherto/qvac/rel…
Ryan Ramage retweeted
Hermes is an AI Harness from @NousResearch that is truly yours: you can shape its persona, it remembers you across sessions, and it gets better the longer you use it.
We ran it fully local on QVAC. The persona, the memory, the skills it learns, all of it stays on your machine.
Ryan Ramage retweeted
"This is not surveillance. This is love."
#ChatControl 🇪🇺
Ryan Ramage retweeted
You really should be using Keet: keet.io/ 👀
Ryan Ramage retweeted
If building P2P go to summit.pears.com/ and apply for the summit. We are always trying to expand our reach and meet new 🍐s. Final list still being compiled so still time to apply!
Ryan Ramage retweeted
QVAC SDK 0.14.0 is live.
This release makes the on-device stack faster on mobile, ships the developer-agent path, and takes local text-to-speech to 31 languages.
Main highlights:
- OpenCode and OpenClaw. The first official OpenCode plugin, plus a maintained OpenClaw compatibility path, both built on managed mode and qvac serve. Point a coding agent at a local model with far less setup and far fewer surprises.
- Brain-computer interface transcription, on the SDK. Take recorded neural signal data and decode it into text, fully on-device, no cloud. Stream it in chunks through a simple API. In 0.14 it runs GPU-accelerated on iOS.
- Text to Speech in 31 languages with our Supertonic3 upgrade.
VOICE AND SPEECH
- Supertonic3 multilingual TTS, 5 languages to 31.
- Chatterbox and Supertonic now run on the Android GPU, with lower memory use (especially on iOS), quantized s3gen Chatterbox support, and a fix for Chatterbox occasionally emitting random speech.
- Whisper transcription now runs on the iOS GPU. Parakeet runs on the Android GPU, with steadier real-time streaming.
VISION AND OCR
- VLM multi-tile batching: high-resolution Pan and Scan images are encoded in one pass instead of tile by tile, for faster vision throughput.
- OCR on ggml (EasyOCR and DocTR) reaches full speed parity with the onnx path, across Metal, OpenCL, and Vulkan.
PLATFORM AND RELIABILITY
- Dynamic compute backends on Linux: one build picks the right backend at runtime, and opens the door to ROCm and CUDA support without per-backend builds.
- Thinking tokens are kept out of the model context, so reasoning no longer fills the KV cache.
SDK 0.14.0 is now leaner and faster to start.
Let’s build.
Ryan Ramage retweeted
"We lost something in recent years. Peer-to-peer is how we take it back." 🍐
@mafintosh is bringing the P2P community to California for the second P2P Summit.
If you're building local-first, serverless, or sovereign software, this is where the future is being decided.
Apply at summit.pears.com/