Join over 250,000 subscribers getting the latest AI news

Cerebras Systems, founded in 2015, specializes in wafer-scale computing for AI workloads, challenging traditional chip architecture assumptions. The TechCrunch Disrupt 2026 event offers insights into scaling AI capabilities amid rising demands for infrastructure. Key discussions will revolve around the limits of current AI hardware and its ability to meet growing computational needs.

Destro AI is a startup focusing on combining AI with logistics operations to enable robots and human workers to collaborate efficiently. Traditional robotics firms often focus on hardware creation, whereas Destro aims to solve specific operational problems using existing robot technologies. Future developments will likely focus on adapting their system across various facilities in the logistics sector.

DoorDash is a food delivery service competing with platforms like Uber Eats and Grubhub. The AI agent aims to streamline the ordering process as part of a broader trend towards personal AI agents. Additionally, DoorDash will begin testing delivery drones in Northern California with select restaurants.

Grounded Entity Biographies (GEB) is a proposed framework that connects visual observations of the same entity across long videos, which traditionally face challenges in event connectivity. The framework addresses limitations of chronological descriptions and uncovers unresolved physical identities among similar objects. Future developments may include further optimization of video analysis from this research.

Raven addresses challenges in scaling AI harnesses, which are structures that enable AI agents to work on specific tasks. The system seeks to replace traditional static harness designs with modular and adaptive solutions. Future developments will likely focus on enhancing collaboration capabilities among AI agents across multiple domains.

OpenAI is a major player in artificial intelligence development, seeking to balance innovation with safety protocols. The company has recently faced criticism and legal challenges regarding its AI tools' behavior and their implications. Moving forward, stakeholders will watch for regulatory responses and advancements in AI safety measures.

AI-designed proteins are increasingly common and pose unique biosecurity risks, as conventional DNA screening methods may not identify them as threats. Google's watermarking approach aims to distinguish trusted designs from potentially harmful ones, although challenges remain in integrating this system with other protein design software.

DH-Robotics focuses on robotic end effectors and components for automation and AI applications. IROS 2026 is a major robotics conference featuring innovations and collaboration among industry leaders. The company also showcased its UDH-3-7 three-finger hand design at the event.

Restate is a Berlin- and San Francisco-based company that specializes in creating infrastructure for reliable operation of software and AI agents. Founded by former Apache Flink team members, Restate's platform addresses the challenges faced by AI agents during service interruptions. The upcoming focus will be on how durable execution can standardize AI application architecture as it becomes critical for operational success.

The Edits app, launched in April 2025, serves as a mobile video editor and creative studio for Instagram users. Meta has previously integrated similar creator assistant tools on its Facebook platform and continues to expand these features across applications.

DeepMind's SynthID Bio builds on existing AI technologies that predict and design proteins, addressing potential biosecurity risks as AI-generated sequences become more prevalent. As synthetic biology advances, reliable verification will be crucial for maintaining database integrity and safety in research.

Visual reward models evaluate how effectively a system can generate images or videos. Previous methods focused on basic scoring, which can obscure nuanced performance metrics. Future developments will likely focus on refining these evaluation methods for improved visual outputs.

World Action Models (WAMs) are AI systems designed to predict future events based on actions. Previous studies have shown that while Latent WAMs are faster, they lack the generalization benefits of Explicit WAMs. Future research may focus on refining these models for better performance across various environments.

Ascerta, founded by former Microsoft executives, originally focused on AI cost management and has shifted its goal to help enterprises measure the effectiveness and ROI of their AI investments. The need for such solutions has increased as companies scale their AI initiatives beyond initial experiments. The investment will support Ascerta's goal to formalize Enterprise AI Management as a software category.

Relay Robotics, founded in 2013, specializes in autonomous delivery robots designed for challenging environments, such as hospitals. Previously named Savioke, the company has completed over 1.5 million deliveries globally across various sectors. Future deployments and advancements in robotic technology within hospitals can be expected as the company expands its operations.

Language models, or LLMs, generate text based on prompts but can misrepresent data by overlooking negative results. This issue is particularly important as AI summaries become prevalent in scientific literature, necessitating accurate reporting. Future developments should focus on improving training and prompting techniques to ensure transparency.

Technical debt refers to outdated technology and systems that companies have not replaced or modernized over time. With the rising focus on AI, enterprises are finding new justifications to tackle these issues, which may lead to significant changes in how technology is implemented moving forward.

Airbnb is a platform that connects travelers with hosts offering accommodations and experiences. This update builds on previous updates that enhanced user engagement and service options. Future developments will likely focus on further integrating AI to optimize guest and host interactions.

OpenAI is a leading artificial intelligence company known for developing advanced AI models. Recent incidents, including a breach involving Hugging Face and another in Australia's healthcare system, have raised concerns about its technology's safety. With heightened scrutiny, OpenAI is implementing stricter safety measures and monitoring protocols during model training to prevent future issues.

OpenAI Dot is a new AI tool aimed at enhancing user productivity through continuous task management. It follows previous iterations by focusing on persistent AI collaboration rather than traditional prompt-response interactions. Future developments should include broader accessibility and improved functionalities.