GPT-engineer, generate an entire codebase from chat

Curating the ocean of AI news keeping you informed on topics key to the craft of software development

🧰 AI tools that help you build software

WizardCoder model takes a substantial lead in open source code related tasks
Wizard Coder, a 15 billion parameter AI model, excels in code generation, take a substantial lead in performance compared to other open source models and even 3rd place when comparing to closed models such as ChatGPT.

Integrate ChatGPT in your IDE
Bito gives you a very nicely integrated Chat gpt panel in your IDE. There are nice add on features such as prompt saving in the form of Templates. It ships with a few and you can add your own. There is also integration to the context menu, so highlight code the select a Bito tool such as “simplify code”. Bito will make the change and show you the diff

AI-Powered Web Accessibility
UserWay's Fix My Code is an AI-powered tool that can optimize your code to meet WCAG 2.1 AA standards. It can create (or fix your pasted) login forms, popups, and nav menus.

Generate an entire codebase from chat
gpt-engineer enables you to specify what you want it to build, and the AI asks for clarification, and then builds it. The configuration options and clarification work stream sets this open source product apart from other tools making similar claims. Lots of folks are excited about this one.

🏗️ Utilizing AI in your software

Vercel Launches AI SDK for React and Svelte
Vercel has introduced an AI SDK for building AI-powered applications. It helps developers build conversational, streaming, and chat user interfaces in JavaScript and TypeScript. The SDK supports React/Next.js, Svelte/SvelteKit, with support for Nuxt/Vue coming soon.

Run LLM apps on iphone, android and more
MLC LLM is a universal solution that allows any language model to be deployed natively on diverse hardware backends and native applications. It offers a customizable workflow for developers to implement models and optimizations

Open Platform for Operating Large Language Models
OpenLLM is an open platform for operating large language models (LLMs) in production. It allows you to fine-tune, serve, deploy, and monitor any LLMs with ease.

📰 AI news for devs

92% of programmers are using AI tools
According to a new GitHub programmer survey, 92% of US-based developers are already using AI coding tools both in and outside of work. Developers say AI coding tools help them meet existing performance standards with improved code quality, faster outputs, and fewer production-level incidents.

Stack Overflow enters the AI arena
Stack Overflow which has to date been the source for many training models has announced how they will start to embrace AI to better their own product. This includes features such as Question Formatting Assistant, Chat Decipher—use AI to extract questions from chat transcripts such as Slack and group together ones that are similar or duplicates, and Title Suggestions—Using AI during the question writing process to help generate more descriptive, accurate titles.

An Open Reproduction of LLaMA lands on Hugging Face 🤗
Meta AI’s LLaMA is not licensed for commercial use, OpenLLaMA is a permissively licensed open source reproduction. They are releasing 3B, 7B, and 13B models trained on 1T tokens. Thus if you've been doing research work with LLaMA, you can switch to OpenLLaMA for utilization in commercial products.

The Double-Edged Sword of AI-Coded Software
AI is making coding more efficient but could lead to complications like increasing technical debt and code complexity. The surge in AI-written code raises concerns about manageability, but the efficiency gains and response to a talent shortage can't be overlooked. Still, adopting AI tools like GitHub's Copilot or Magic AI means we need to rethink our methodologies to ensure quality over quantity. What struck me is how AI is changing the very dynamics of software development.

GPT Can Use Tools Now—That’s a Big Deal
This is an entertaining discusses of the recent update to GPT called “function calling” which allows it to call out to external code, databases, or other APIs when it needs to. Each function in function calling represents a tool that a GPT model can use when necessary, and GPT gets to decide which ones it wants to use and when.

Can you trust ChatGPT’s package recommendations?
This article discusses the research done by Vulcan Cyber, which discovered that attackers can use ChatGPT to help them spread malicious packages into developers’ environments. The technique relies on the fact that ChatGPT sometimes answers questions with hallucinated sources, links, blogs, and statistics.

🏫 Read/Listen/Watch: Learn

Learn while you commute, do chores, exercise…

Chris Lattner discusses Mojo and the challenges of AI research [watch or listen: 3 1/2 hours]
Chris Lattner sits down with Lex Fridman to discuss the motivation behind the Mojo programming language and the aspirations of Modular AI.

Free Course on LangChain & Vector Databases in Production [course]
This is a free course on LangChain and Vector Databases in Production, brought to you by Activeloop, Towards AI, and Intel Disruptor Initiative. The course is designed to equip you with the knowledge and skills to train, fine-tune, and incorporate LLMs into AI products at your organization. In the first installment, you will learn how to leverage LangChain, a robust framework for building applications with LLMs, and explore Deep Lake, a groundbreaking vector database for all AI data.

The Mystery at the Heart of ChatGPT [listen: 38min]
In the What's your problem podcast Jacob Goldstein talks with Sam Bowman about the mysteries at the heart of large language models like ChatGPT. One of the most amazing things about these models is that nobody really understands precisely how they do what they can do, not even the people who build them.

Mark Zuckerberg future of AI at meta [Watch or listen: 2hr:42min]
Lex Fridman interviews Mark Zuckerberg they talk about Meta’s current investments in AI and in general where they see AI going in the near and long term. One point of interest was Mark describing how Meta has and will continue to leverage open sourcing their technology.

 Thanks for reading and we will see you next time

Follow me on twitter, DM me links you would like included in a future newsletters.