Qwen 3.6 open source ranks in the top 10% on LiveBench. Can we finally code fully local?
Tested Qwen3.6-27B on a MacBook M4 Pro, 100k context:
- Via LM Studio: 11 tok/s, totally usable for small tasks. Really solid quality in both English and French for a local model!
- Via Claude Code: unusable. Claude Code sends a 75k token prompt upfront: 13 minutes to process the first message 😅
Great progress on open source, but let's be honest: full local dev isn't there yet
Can't wait to see what's next!
Introducing the Google Workspace CLI: https://t.co/8yWtbxiVPp - built for humans and agents.
Google Drive, Gmail, Calendar, and every Workspace API. 40+ agent skills included.
Tired of App Store rejections for your iOS submissions? You can use “Greenlight” to pre-check the main issues when submitting iOS apps!
greenlight preflight /app/MyIosApp/
The repo is in the comments.
Claude Code finished 10 minutes ago but you didn't notice? Or it asked you a question and you missed it?
Install an audio hook that plays a sound whenever Claude needs your attention or completes a task!
Details below 👇
ElevenLabs + Revolut.
We're deploying ElevenLabs Agents across Revolut's customer support - a company that's reimagined what financial services can be and scaled to 70M+ customers globally with relentless focus on product and execution.
Both our companies obsess over metrics - and the numbers look great:
- 8x faster resolution
- 99.7% success rate
- 30+ languages
- 24/7 support
@Revolut tested every provider and chose us for voice quality, latency, security, and partnership. Operating across nearly 40 markets, they needed complexity and speed without sacrificing quality. Their team is world-class and moves fast - so it was even more for our Forward Deployed Engineers to be embedded with them for months, building and iterating together to get this right.
I see a lot of hype around Speech To Text tools: Super Whisper, Parakeet, Whisper... But on MacOS, it's built in natively! System Settings > Keyboard > Dictation. Turn it on and add a shortcut.
To be honest, it holds its own pretty well against Whisper, Parakeet and the rest.
TTS/STT models (Text To Speech / Speech To Text) are having a moment right now. If you use Telegram, I built a small tool a while back that transcribes all voice messages. Free and Open Source using open source Whisper models from @OpenAI 👇👇
https://t.co/F89UhWNp5p
Want to run Claude Code with a LOCAL model? Whether it's for privacy, company policies, or just curiosity, here's how with Ollama:
- Download Ollama
- Pull a model: ollama pull ministral-3:3b
- Run Claude Code: ollama launch claude --model=ministral-3:3b