Mon, Feb 20, 2023Replit x Weights & Biases Machine Learning Hackathon Winners
The Replit x Weights & Biases Machine Learning Hackathon was Replit’s very first machine learning-focused hackathon that took place on February 4-11, 2023 with participants selected from all over the world! After applying and being accepted into the hackathon, participants hacked for seven days on their projects, using both custom machine learning models and fine-tuning existing ones, all by combining the power of Weights & Biases and Replit. During the hackathon, builders collaborated and got help from mentors on our Discord (shout out to Morgan and Ayush!), worked through examples and demos from the opening ceremony and our list of resources, and even documented their journeys using W&B Reports. In the end, we gave away over 500,000 Cycles worth of prizes to the best projects created by combining the two technologies. Expert ML engineers from Weights & Biases judged each project to decide the ultimate winners: And the winners are...
Tue, Feb 14, 2023Announcing Ghostwriter Chat: The first conversational AI programmer
Today, we are announcing the beta release of Ghostwriter Chat. Now with chat, a proactive debugger, and awareness of your project’s context, Ghostwriter becomes the best AI pair programmer in the world. Building Ghostwriter Last fall, we introduced Ghostwriter, Replit’s coding AI, with Complete, Generate, Transform, and Explain Code - coding assistant tools that work with you as you code. Together they made coding faster by automating repetitive tasks and removing the drudgery of writing boilerplate code. That was just the first step towards our master plan to create a fully autonomous pair programmer – one that feels like working with another teammate.
- Tue, Jan 10, 2023
Apply for Replit’s first Machine Learning Hackathon
We’re extremely excited to announce Replit’s very first Machine Learning Hackathon, in partnership with Weights and Biases! If you’re interested in joining the waitlist, head over to the official site. Weights and Biases is a machine learning platform that tracks everything you need to make your models reproducible – from hyperparameters and code to model weights and dataset versions. If you’ve never worked with W&B, check out this example Repl for an introduction! The total prize pool is over 500,000 Cycles – there are multiple opportunities to win, with prizes for Best Weights & Biases Report, Best Repl, an Honorable Mention, and of course, the Grand Prize of 300,000 cycles.
Mon, Oct 31, 2022Meet Replit Ghostwriter, your partner in code
A Ghost ... that codes. Today we unleash Ghostwriter, Replit's Coding AI, to the public for 1,000 Cycles ($10) per month. Ghostwriter makes coding fast, easier, and more fun. Tasks that used to be annoying or slow, just aren't anymore. Your coding, but better. Code with your AI pair programmer, on any device, anywhere. Try Ghostwriter, and you will never go back. Celebrate Halloween by getting your spooky software creating companion today. Want instant access? You can buy it. Want to try it out? We're offering a limited trial on a first come, first serve basis. Apply here. For two months, we've been hard at work beta testing the AI features. Taking user feedback. Fine-tuning. Everything we can do to make Ghostwriter one of the most powerful AI programming tools on the planet.
Tue, Sep 20, 2022Productizing Large Language Models
Large Language Models (LLMs) are known for their near-magical ability to learn from very few examples -- as little as zero -- to create language wonders. LLMs can chat, write poetry, write code, and even do basic arithmetic. However, the same properties that make LLMs magical also make them challenging from an engineering perspective. At Replit we have deployed transformer-based language models of all sizes: ~100m parameter models for search and spam, 1-10B models for a code autocomplete product we call GhostWriter, and 100B+ models for features that require a higher reasoning ability. In this post we'll talk about what we've learned about building and hosting large language models. Nonsense Any sufficiently advanced bullshit is indistinguishable from intelligence, or so the LLM thought. LLMs are super suggestible -- in fact, the primary way to interact with LLMs is via "prompting." Basically, you give the LLM a string of text and it generates a response, mostly in text form although some models can also generate audio or even images. The problem is, you can prompt the LLM with nonsense and it will generate nonsense. Garbage in, garbage out. Also, LLMs tend to get stuck in loops, repeating the same thing over and over again, since they have a limited attention span when dealing with some novel scenarios that were not present during training.
Wed, Sep 7, 2022Ghostwriter AI & Complete Code Beta
Update: Ghostwriter is out now! In 2018 when we announced Multiplayer Mode, we said it's the most significant evolution of Replit to date. For the first time, you could share a URL with a friend, student, or coworker and get a shared text editor and runtime — no setup required. Replit Multiplayer is changing how an entire generation of programmers learn how to code and make software. Today, we're announcing Ghostwriter, which infuses state-of-the-art intelligence into nearly all IDE features. Ghostwriter sports an ML-powered pair programmer that completes your code in realtime, tools to generate, transform, and explain code, and an in-editor search utility that lets you find and import open-source code without leaving your editor (think Stackoverflow in your editor).