Devin is a fully autonomous AI software engineer created by Cognition Labs. It is designed to work alongside engineers or independently complete tasks for engineers to review. Devin can plan and execute complex engineering tasks requiring thousands of decisions, recalling context at every step, learn over time, and fix mistakes. It actively collaborates with users, reporting on its progress in real time and accepting feedback. Devin can use unfamiliar technologies, build and deploy apps end-to-end, autonomously find and fix bugs in codebases, train and fine-tune its own AI models, and more.

This article shares details on two versions of Meta's 24,576-GPU data center scale cluster. These clusters support the company's current and next-generation AI models as well as AI research and development across GenAI and other areas. Meta's long-term vision is to build artificial general intelligence that is open and built responsibly to be widely available so that everyone can benefit from it. Its infrastructure is designed to power this ambition.

The Limitless Pendant is a device that records everything its wearers hear and then uses AI to help them remember and make sense of it. Designed to be clipped onto a shirt or worn on a string around the neck, it uses a beam-forming technology that can more clearly record people around the user and not the rest of the room. The Pendant is part of a system that uses data from different apps to provide users with useful information. The company plans to eventually create AI agents that do things on users' behalf.

OpenAI's Batch API helps developers save costs and get higher rate limits on async tasks such as summarization, translation, and image classification. Developers just have to upload a file of bulk requests and they'll receive results within 24 hours, with 50% off API prices. A link to the Batch API reference docs is available.

Salesforce has introduced xGen-VideoSyn-1, a text-to-video (T2V) model that generates realistic scenes from textual descriptions. The model uses a video variational autoencoder (VidVAE) to compress video data, reducing computational demands, and a Diffusion Transformer (DiT) for improved temporal consistency and generalization.

Anthropic has added system prompts and updated dates for all models.

Powered by phi-3-mini, this space uses a rarity prompt to generate data about any topic. It isn't the most accurate, but it is fascinating and powerful.

Neural networks can represent and manipulate 3D objects in 2D scenes by conditioning on per object representations. This work may well be the Holy Grail of 3D object disentangling.

Researchers have introduced a new method called T3M for creating 3D animations guided by text inputs. Unlike previous techniques that relied only on speech, T3M allows for more accurate and customizable animations, making it a valuable tool for virtual reality, gaming, and film production.