Anthropic's Claude AI model, using sixteen agents working in tandem, successfully created a new C compiler from scratch, a project that took two weeks and cost approximately $20,000 in API fees, according to a blog post by researcher Nicholas Carlini. This experiment highlights the growing capabilities of AI agents in complex coding tasks, as the agents worked on a shared codebase with minimal supervision.
The project, which utilized Claude Opus 4.6, demonstrates Anthropic's advancements in AI coding experiments, amid a push toward multi-agent tools from both Anthropic and OpenAI. The AI agents reportedly produced a 10,000-line compiler.
In other news, defense attorneys for Roberto Carlos Muñoz-Guatemala, convicted of assaulting ICE officer Jonathan Ross, are seeking access to investigative files related to the killing of Renee Nicole Good. Attorneys are requesting training records and investigative files related to Ross, the ICE agent who shot and killed Good during a targeted operation in Minneapolis on January 7, as reported by Wired. Muñoz-Guatemala's attorneys filed a post-trial motion in the US District Court.
Meanwhile, a new attack chain is emerging that allows adversaries to gain access to cloud environments within minutes, as detailed by VentureBeat. The attack, known as the identity and access management (IAM) pivot, begins with a seemingly legitimate LinkedIn message from a recruiter. The developer is then tricked into installing a malicious package that exfiltrates cloud credentials, including GitHub personal access tokens and AWS API keys. CrowdStrike Intelligence research published on January 29 documents how adversaries are exploiting this vulnerability.
In the realm of AI optimization, researchers from Stanford, Nvidia, and Together AI have developed a new technique called Test-Time Training to Discover (TTT-Discover). This technique allows models to continue training during the inference process, updating their weights for the specific problem. According to VentureBeat, the researchers used this method to optimize a critical GPU kernel, achieving a performance that was twice as fast as the previous state-of-the-art written by human experts.
Finally, Valve announced that rising component costs, such as RAM and storage, are forcing them to revisit the pricing and shipping schedule of their upcoming Steam Machine. Analysts are divided on how much of these increased costs Valve will pass on to consumers, according to Ars Technica. The exact form of these changes is currently unknown.
Discussion
AI Experts & Community
Be the first to comment