The secret sauce? AutoPM. This feature basically turns Genie into a team lead. It spawns and coordinates multiple agents that: -Research new APIs
-Explore your codebase
-Write + refactor code
-Test and create PRs You literally scale your dev team with a toggle.
CODE
-
AutoPM: Scale your dev team with a toggle for AI agents.
By
–
-
Genie: Multi-Agent System That Works While Laptop Is Off
By
–
What makes Genie different? It’s not a glorified chat window. It’s a multi-agent system that can: -Understand your whole repo
-Plan complex tasks
-Work in parallel
-And even keep running while your laptop’s asleep Yes, your laptop can be off. -
Cosine Genie: first AI dev tool for real engineering work
By
–
Just used Cosine’s Genie and honestly… this is the first AI dev tool that feels like it was built for real engineering work.
— AI Breakfast (@AiBreakfast) 12 juin 2025
Not just code suggestions – I’m talking full tasks across large codebases, handled autonomously.
This beat Codex and Claude Code for a reason 👇 pic.twitter.com/NZrWGMrEWIJust used Cosine’s Genie and honestly… this is the first AI dev tool that feels like it was built for real engineering work. Not just code suggestions – I’m talking full tasks across large codebases, handled autonomously. This beat Codex and Claude Code for a reason
-
LangGraph+ Tensorlake Integration for Agent Document Understanding
By
–
LangGraph+ Tensorlake: Unlocking Document Understanding for Agents When creating agents that interact with data, the connections you have to that data matter Excited to announce an integration Tensorlake – a best in class document ingestion engine https://
tensorlake.ai/blog/announcin
g-langchain-tensorlake-integration
… -
Voyager SDK enables 777 FPS video inference deployment
By
–
No camera? No problem. The Voyager SDK lets you run high-speed inference on video files or full datasets. 777 FPS on video input 450 FPS on COCO dataset 1 command to deploy Metis makes #AI inference simple and seamless. Come and take a look! #EdgeAI #VoyagerSDK
-
OpenAI to Improve Custom GPTs with Model Selection
By
–
OpenAI is preparing an improvement for custom GPTs, where GPT builders will be able to set a "recommended model"! This addition will also let users select which model to use with custom GPTs. Codex will also get an option to set a number of attempts per task.
-
Anthropic Introduces Custom Integrations Using Remote MCP Protocol
By
–
Ah yes this seems to be new and described here: https://
support.anthropic.com/en/articles/11
175166-about-custom-integrations-using-remote-mcp
… -
LangGraph Opens Doors for Community Contributions and Development
By
–
Contributions welcome! We’ve just made it easier to get involved in LangGraph — check out the newly tagged issues and help shape the future of the project
-
Cursor integrates MCP settings to boost developer adoption
By
–
BTW we link to Cursor MCP settings prominently from inside our product @victormustar is helping MCP adoption inside Cursor hehe Could be cool to do some comms around this.
-
Claude Desktop now syncs MCP servers from web
By
–
Nice! Claude Desktop now also syncs your MCP servers (aka Integrations) from Claude Web So this means Claude is now compatible with remote MCPs not just local ones, very nice @AnthropicAI