The New Stack
PulseAugur coverage of The New Stack — every cluster mentioning The New Stack across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
Anthropic expands Claude with productivity tools to rival Microsoft Office
Anthropic is expanding Claude's capabilities to integrate more deeply into enterprise workflows. The AI assistant is being enhanced with tools for data analysis, digital whiteboards, and spreadsheet functionality, aimin…
-
Google open-sources gVisor, used by Anthropic and OpenAI for AI sandboxes
Google has open-sourced gVisor, a container sandbox technology that is reportedly used by AI companies like Anthropic and OpenAI. This move allows for broader adoption and potential improvements to AI model sandboxing, …
-
OpenAI Reverses Token Pricing Shift After Developer Pushback
OpenAI considered a shift away from token-based pricing to a flat fee per job, but developer feedback led to a reversal of this plan. Developers expressed a preference for transparent pricing models that clearly outline…
-
Claude Haiku 5.5's 100K token threshold dramatically increases costs
Anthropic's Claude Haiku 5.5 has a pricing tier that significantly increases costs once a prompt exceeds 100,000 tokens. This step-up in pricing, where all tokens in a request are billed at a higher rate, can lead to a …
-
Anthropic's Claude AI flags 29,000 open-source bugs, most unfixed
Anthropic's Claude AI model identified a significant number of potential vulnerabilities in open-source software, flagging 29,000 possible bugs. However, a substantial majority of these issues, 516 out of 29,000, remain…
-
Anthropic cuts Claude Opus pricing by 20% with new Opus 5.5 model
Anthropic has announced new pricing for its Claude Opus models, with the upcoming Opus 5.5 set to be 20% cheaper than its predecessors. The new model will cost $4 per million input tokens and $20 per million output toke…
-
OpenAI agent oversight gap surfaces; Musk eyes space compute; Kubernetes lessons for AI agents
OpenAI is facing scrutiny over an oversight gap in its AI agent technology, affecting over 100 organizations. Meanwhile, Elon Musk has suggested that space will soon become the primary location for computing power, with…
-
Cloudflare launches AI agent payment gateway with USDC settlement
Cloudflare has launched a closed beta for its Monetization Gateway, enabling AI agents to pay for services like websites, APIs, and datasets using HTTP 402 and the x402 protocol, settling in USDC on the Base network. Th…
-
MCP Server Security Lapses Highlighted Amidst Deployment Challenges · 2 sources tracked
A recent analysis of over 5,200 MCP servers revealed significant security and deployment shortcomings, with 88% requiring authentication but only 8.5% using OAuth, and a concerning 492 servers exposed without authentica…
-
Cohere releases North Small Translate for cost-effective 50-language translation
Cohere has released North Small Translate, an open-weight machine translation model built on the Command A Plus architecture. This model, which has 25 billion active parameters out of a total of 218 billion, is designed…
-
Vibe-Coded Apps Emerge as New Shadow IT Threat
The article discusses how applications designed around specific "vibes" or aesthetics are emerging as a new form of shadow IT. These applications, often developed with AI tools, bypass traditional IT oversight and secur…
-
Anthropic's Claude AI integrates native browser automation and collaboration
Anthropic's Claude AI has integrated native browser automation and collaboration features, enabling users to streamline workflows. This development, highlighted by The New Stack, aims to enhance productivity by allowing…
-
Orbital AI Data Centers Face Major Cooling Challenges
While the concept of orbital AI data centers, potentially involving SpaceX and NVIDIA, sounds appealing, practical challenges make it a difficult endeavor. Cooling these facilities in the vacuum of space presents signif…
-
AI kill switches face execution hurdles due to infrastructure complexity
Implementing an AI kill switch presents significant challenges, primarily due to the difficulty in achieving deep visibility into complex enterprise infrastructure. While mandating such a feature is feasible, its effect…
-
Space-based AI data centers face insurmountable cooling challenges
The concept of placing AI data centers in space, potentially leveraging SpaceX and NVIDIA's technologies, faces significant practical challenges. While the idea of using the vacuum of space for cooling is appealing, the…
-
Cloudflare open-sources AI tool that cleared Astro's GitHub backlog
Cloudflare has open-sourced an AI tool that helped clear the GitHub issue backlog for the Astro Web Framework. Co-founder Fred Schott stated that the tool, developed over six months following Cloudflare's acquisition of…
-
Open Secure AI Alliance aims to boost US AI leadership
The Open Secure AI Alliance is being formed to bolster US leadership in artificial intelligence, following an open letter on open models. Professor Amanda Brock of OpenUK commented that a more globally inclusive approac…
-
AI agents gain real-time manufacturing data access via new MCP server
A developer has built an MCP (Model Context Protocol) server to integrate AI agents with real-time manufacturing data, addressing the common issue of sales teams lacking up-to-date information on pricing, stock, and pro…
-
LLM release gates: Beyond traditional CI/CD for AI features
Traditional CI/CD pipelines are insufficient for managing the release of LLM-powered features, as LLM outputs are graded rather than asserted and can degrade in unexpected ways. To address this, teams are implementing n…
-
Anthropic prepares to re-release Fable 5 model with paid access
Anthropic is preparing to re-release its Fable 5 model, which was previously taken down. The model, also referred to as "An Ape and a Fox," is expected to become available again soon. Details regarding the pricing struc…