🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

day 1 of asking for a raise

Sonnet 3.7 is now the state-of-the-art base model for agentic software engineering, according to our private evals Cognition.

For the month of March we’re running a Devin competition at Gumroad The winning engineers (and non-engineers) who use Devin the most (measured by merged PRs) will get $33,333, split amongst them!

Introducing Phantom 👻: a method to train robot policies without collecting any robot data — using only human video demonstrations. Phantom turns human videos into "robot" demonstrations, making it significantly easier to scale up and diversify robotics data. 🧵1/9

Devin is responsible for shipping 137 PRs in February, up 500% over Jan! March should be >200 PRs based on the first 10 days.

Introducing Devin 2.0: a new agent-native IDE experience. Generally available today starting at $20. 🧵👇

This is my favorite release since our original launch. Devin feels like a new product - much faster & more interactive, and now great for tasks that only 90% work, because the last 10% is so quick to finish yourself in the cloud IDE.

Free, continuously updated wikis & deep code Q&A for open source repos!

deepwiki.com/vllm-project/v… looking amazing! We found the quality of the documentation and diagram to be great and accurate. Will recommend :D

stellar-core's DeepWiki is phenomenal. We don't talk enough about the magic that goes inside of core but its clean code and extensive in-line documentation lend themselves well to deep machine understanding deepwiki.com/stellar/stella…

Most founders underappreciate having an AI software engineer in your Slack Example 👇

If AI is not the #1 committer to your company's code base, why not?

Our research interns present: Kevin-32B = K(ernel D)evin It's the first open model trained using RL for writing CUDA kernels. We implemented multi-turn RL using GRPO (based on QwQ-32B) on the KernelBench dataset. It outperforms top reasoning models (o3 & o4-mini)! 🧵

Excited to have Russell Kaplan on stage next! He's sharing how Cognition built Devin, their autonomous code agent 🤯! - "How many people have heard of Devin?" every hand in the room went up 😂! - Codebase understanding is hard, gap between advertised and "effective"

Devin ♥️ Open Source We're gifting $500 in agent compute units for OSS projects, in addition to free DeepWiki dot com documentation for every repo on GitHub

Many such cases

This past week has been the most explosive growth in Devin usage since our GA launch. Working hard to keep the servers up!

my experience as well - Devin just started working recently 🤷‍♂️ works best when you throw a ton of backlog tickets at it and it integrates with all the right things (slack, github pr comments, etc.)

The DeepWiki MCP server is live! How to use it + what’s inside 🧵👇

How DeepWiki works under the hood, in 2 minutes 📹 For more details on how we build Devin, check out Russell Kaplan's full talk at LangChain Interrupt 🔗👇