Codeium

🔔 July 2024 Changelist 🔔 🗨️ Better Chat: bigger models, 2x recall on reasoning, list of UX improvements 📊 Analytics API 🩺 Extension Diagnostics 🪄 Codeium Command Improvements 🚀 JetBrains Gets Better ... and more! Happy Friday! Read more: https://lnkd.in/gKdQxv6g

Are you really a tech company from the valley if you don’t have a billboard (or a few) around San Francisco? 🤔 Check out our new billboards throughout the Bay Area if you want to work at a place full of life. Now viewable down the 101 from San Mateo to Santa Clara. https://lnkd.in/gHsgmg8w #Codeium #SanFrancisco #TechStartup #Developer

Thank you Ed Elson for having Varun Mohan and Jeff Wang on the First Time Founders podcast! Lots of nuggets on how we think about what we are building and how we are building it. Watch the full podcast: https://lnkd.in/g3DNyg-A

Today, we are rolling out a better chat experience than ChatGPT for software developers. Chat has always been an important part of the Codeium experience, but now we believe it is the best chat experience for anyone writing code. What makes it the best? A combination of model, reasoning, UX, and accessibility. The same ingredients to any great AI product. As a company, we have full control of the entire stack, so we have innovated on all axes. ⚙ Model: Today, we are deploying a Llama 3.1-70B based Codeium chat model for all users, and an additional Llama 3.1-405B based Codeium chat model for all paying users. The Codeium versions of these models are tightly integrated with our reasoning stack, leading to better quality suggestions than GPT-4 for coding tasks, and are hosted on our industry leading infrastructure so that we can offer them for free (or very cheaply) to our users. 🧠 Reasoning: We are rolling out our next generation retrieval engine, which has a recall that is 2x better than state-of-the-art embedding-based RAG systems. Simply put, all of our users will now get more relevant, grounded Chat responses. 🖋 UX: We are rolling out a large suite of UX features and improvements: inline citations, chat history, slash commands, inline context pinning, deep insights, copy to clipboard, and a whole lot more. 🌐 Accessibility: Chat in more IDEs than any other tool - Visual Studio Code, Jetbrains, Visual Studio, Eclipse, XCode. If you need it self-hosted for your enterprise, we can do that too. Oh and also available on the web via Codeium Live. We don’t care how you work, we just want to help you work better. There really is no Chat experience quite like Codeium’s in the world of software development. It is likely one of the reasons why developers on Stack Overflow self-reported Codeium as driving more productivity and satisfaction than any other tool, including GitHub Copilot and ChatGPT. Read more about all of the improvements we have made to Chat on our blog, and use Codeium yourself, for free: https://lnkd.in/dhbqR_vK

Made it to the big stage at AI Engineer World's Fair last month! Took the opportunity to share some of Codeium's secret sauce. Here’s what I presented: Talk title: “Are embeddings actually limiting AI agents? How Codeium Breaks Through the Ceiling for Retrieval” Embeddings have been the gold standard for retrieval for years, but they’ve hit a ceiling. Even the largest models are within a few percentage points of each other on MTEB (massive text embedding benchmark). Enter Codeium’s vertical integration. This means we: 1. Train our own models 2. Build custom infrastructure down to the metal 3. Are product driven, not research driven The result is we can run massive, workflow-specific LLMs at a speed and efficiency unmatched by the competition. Think of this as: we spend $0.01 for every LLM call instead of $1.00. If you are a Codeium user, we’re willing to throw you 100x calls since that’s what will give you the highest quality code. Low cost —> more compute —> higher quality product experience We exploited this fundamental business advantage by building “M-Query,” a system that runs parallel LLM calls on every item in your codebase. Imagine asking “I want to do X, is file Y relevant?” on every single piece of code. We can get a ranked list of every relevant piece of code in your team’s repository within seconds. All this, so that our users & customers can generate the highest quality code conditioned on their existing codebase(s), documentation, organizational standards, best practices, and more. DMs open if you want to discuss or hear more about what we’re working on! It takes a village, our team is hiring 🙂 Thanks so much to Shawn swyx W and Benjamin Dunphy for inviting me to speak. And shoutout to the whole Codeium team for building a killer product. The full talk is linked in the comments below.

Kevin Hou, Head of Product Engineering at Codeium, shares how we cracked the ceiling of code retrieval. In the talk, learn about: • Why code context retrieval is hard • How we leverage compute to solve this for millions of users https://lnkd.in/gPnfpMVv

📈 Rising the Ranks on the Annual Stack Overflow Developer Survey Compared to 2023: - Almost 400% increase in usage - Fourth most admired AI tool out of any category, beating Gemini, Claude, etc - Clear Top 2 in admired code assistants (up from 5th in 2023) - Continued increase in desirability, unlike our competitors Read more: https://lnkd.in/eqBaQFA9

🎙️ Curious about what the future of coding looks like? Our CEO Varun Mohan shares how Codeium differentiates itself by training its own models, leveraging context from entire codebases, and focusing on improving the entire software development lifecycle beyond just code writing. From humble beginnings as a GPU virtualization startup to unveiling an $93M-backed AI toolkit that's simplifying software development, Tony Hoang's latest The AI Podcast episode highlights how Codeium is transforming coding efficiency and boosting developer creativity through self-training. Dive into it here 👉 https://bit.ly/4cQ0KRr

🎙️ Curious about what the future of coding looks like? Our CEO Varun Mohan shares how Codeium differentiates itself by training its own models, leveraging context from entire codebases, and focusing on improving the entire software development lifecycle beyond just code writing. From humble beginnings as a GPU virtualization startup to unveiling an $93M-backed AI toolkit that's simplifying software development, Tony Hoang's latest The AI Podcast episode highlights how Codeium is transforming coding efficiency and boosting developer creativity through self-training. Dive into it here 👉 https://bit.ly/4cQ0KRr

Inline refactoring better than GPT-4 and 4x faster? Read how we iterated on Command to make it state-of-the-art: https://lnkd.in/gnTs_itz