Anthropic has announced the release of Claude Opus 4.8, an upgraded version of its large language model that builds upon Opus 4.7 with significant improvements across benchmarks, making it a more effective collaborator. This new version is available at the same price point as its predecessor.

Several new features accompany Opus 4.8, including user control over the "effort" Claude expends on tasks within claude.ai, and a "dynamic workflows" feature in Claude Code designed to tackle very large-scale problems. Additionally, the fast mode for Opus 4.8 now operates at 2.5 times the speed and is three times cheaper than for previous models.

The model demonstrates enhanced judgment and reliability in agentic tasks, with a notable emphasis on its "honesty," being approximately four times less likely than its predecessor to overlook flaws in code it has written. It also shows improved prosocial traits, such as supporting user autonomy and acting in the user's best interest, alongside substantially lower rates of misaligned behavior.

Looking ahead, Anthropic is working on developing and releasing models that offer similar capabilities to Opus at a lower cost. Furthermore, as part of Project Glasswing, there are plans to generally release a new class of models with even higher intelligence than Opus, such as Claude Mythos Preview, which is currently used for cybersecurity work and requires stronger cyber safeguards before broader availability.

👉 Read the full article on Anthropic

According to the TechCrunch article title, a chip startup named Xcena has successfully raised $135 million, achieving a valuation of $570 million. The core premise of this funding round is Xcena's belief that the primary bottleneck in AI development is not computational power but rather memory.

This significant investment suggests a growing recognition within the industry of the critical role memory plays as AI models become increasingly large and complex. Efficient memory solutions are likely seen as essential for handling the vast amounts of data and processing required by advanced AI.

(Due to inability to access the article content, specific details about their technology or products are unavailable.)

👉 Read the full article on TechCrunch

The TechCrunch article title, "The internet is being rebuilt for machines," suggests a fundamental shift in how the internet's infrastructure and design are evolving to better support usage by AI agents and other automated systems, rather than solely humans.

This implies a need to address the challenges current internet structures pose as AI agents increasingly gain the ability to autonomously gather information, perform tasks, and interact across the web. Potential aspects of this rebuilding could include more machine-readable semantic web technologies or more structured data access via APIs.

(Due to inability to access the article content, specific details about the aspects of this rebuilding or technologies involved are unavailable.)

👉 Read the full article on TechCrunch