- The Rundown AI
- Posts
- Open-sourced AI coding agent
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
A new autonomous AI agent is going viral across the AI community today, and this time, it’s completely open-sourced.
If you’ve seen Devin but got access yet, you’ll like this one. Let’s get into it…
In today’s AI rundown:
Open-sourced AI software developer agent
Apple Vision Pro’s spatial Personas
Get text-to-speech audio trained on your voice
Anthropic uncovers 'many-shot jailbreaking'
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 3 minutes
LATEST DEVELOPMENTS
AI AGENTS
Image source: SWE-agent
The Rundown: Researchers from Princeton NLP have developed SWE-agent, an open-source system that turns GPT-4 into an AI software engineering agent that can autonomously solve issues in GitHub repositories.
The details:
SWE-agent achieves accuracy similar to that of Devin (a recently viral AI agent) on the SWE-bench benchmark, resolving 12.29% of issues autonomously.
The agent has an average task completion time of 93 seconds.
The system interacts with a specialized terminal, allowing it to open and search files, edit specific lines, and write and execute tests.
TOGETHER WITH NYLAS
The Rundown: Nylas surveyed 1,000 developers, product and engineering leaders, and executives to better understand how they’re harnessing generative AI today.
The in-depth guide details:
The future of work and AI's pivotal role
The importance of education and training for effective AI use
Navigating security and privacy challenges
Download the full report to stay ahead of the latest trends.
APPLE VISION PRO
Image source: Apple
The Rundown: Apple Vision Pro users can now enable spatial Personas in SharePlay-enabled apps, allowing for collaboration, gaming, and media consumption with other users in a virtual space.
The details:
Spatial Persona avatars are designed to replicate a user’s facial expressions and hand gestures.
Up to five people can watch movies and TV shows, play games, collaborate, and hang out together.
Users can control what they see and reposition objects without affecting other users' views.
The feature requires VisionOS 1.1 and is enabled through the FaceTime app.
Why it matters: Spatial Personas is Apple’s attempt to make virtual collaboration and social interaction feel more natural and immersive. If successful, it could revolutionize fields like education, remote work, and entertainment.
Watch the Spatial Personas demo here.
AI TRAINING
The Rundown: AudioBox is Meta's new foundation research model for audio generation. In this tutorial, we’ll show you how to leverage it to generate AI audio that sounds like your own voice for free.
Step-by-step:
Head to the AudioBox demo and scroll down.
Click “Record your voice“ to use your audio sample, or use text descriptions or pre-loaded voices instead.
AudioBox will then prompt you to read a short sentence to upload your vocals to the model.
Type the text you'd like to generate after recording (or using a sample recording).
That’s it! AudioBox will then generate two recordings in your vocal style 🎉
AI RESEARCH
Image source: Anthropic
The Rundown: Anthropic researchers have discovered a new "jailbreaking" technique called "many-shot jailbreaking" that can evade the safety guardrails of large language models (LLMs) by exploiting expanded context windows.
The details:
Many-shot jailbreaking involves inserting a series of simulated dialogues into the input to exploit LLMs' in-context learning abilities.
The likelihood of eliciting a harmful response increases with the number of dialogues (or "shots") included in the prompt.
The effectiveness of many-shot jailbreaking is related to the process of "in-context learning," where LLMs learn using the prompt context.
Anthropic has informed other AI researchers and companies about this vulnerability and is actively working on mitigation strategies.
Why it matters: The discovery of many-shot jailbreaking highlights the double-edged nature of the expanding context windows in LLMs. While longer inputs and in-context learning make models more useful, they also enable new classes of vulnerabilities.
NEW TOOLS & JOBS
📹 Magic Hour- AI video creation platform to streamline content production
🔎 SemanticPDF- Semantic search for your PDFs
💼 Canyon- All-in-one AI-powered job search tool
📄 Elicit- Analyze, search, and extract research papers
💬 Ellipsis- Reviews pull requests and turn GitHub comments into functional, tested code
🤖 Momentic- Automated software testing platform with AI
💻 C3 AI - Intern Software Engineer
✈️ Shield AI - Aviation Fleet Engineer
⚙️ Mistral AI - SRE Engineer
📋 OpenAI - Admin Business Partner
QUICK HITS
Replit launched Code Repair, a low-latency programming AI agent that can automatically fix your code in the background while developers work.
A group of 200 musicians, including Billie Eilish, Nicki Minaj, and Katy Perry, signed an open letter calling on tech companies to use AI music generation tools responsibly.
Microsoft is testing an AI-powered Xbox chatbot that can automate support tasks and process game refunds.
The UK and US governments announced a partnership to jointly safety test the most powerful AI models and develop a common approach.
Amazon Web Services expanded its free credits program for startups to cover the use of major AI models from Meta, Anthropic, and Cohere, in addition to its own Bedrock platform.
Researchers discovered that feeding AI image generators just 1,000 extra culturally diverse images can significantly reduce stereotypes and misrepresentations in outputs.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 500k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Join the conversation