英文科技评测与行业媒体 feed · InfoWorld · 全站最新

Google is adding a new feature, Orchestrator, to Antigravity, its AI-powered development environment, to help enterprises improve software development efficiency by automatically coordinating multiple Gemini models, including the newly introduced Gemini 4 Hydrogen. The new feature, which allows developers to delegate planning, coding, and review tasks to different models within a single coding session, combines a planning model, such as Gemini 4 Carbon or Argon, with Hydrogen, which the hypersca

Flexera’s 2026 State of the Cloud Report agrees with many other cloud reports these days: Cloud is more strategic, cloud is more complex, AI makes everything harder, and better governance is the answer. This report, based on a survey of 753 cloud decision-makers and users, argues that enterprise cloud has entered a “value era” in which organizations are measuring success less by cost savings alone and more by business outcomes. The headline findings are familiar. Flexera says 64% of organization

AWS on Thursday said that it was upgrading the runtime layer of Amazon Bedrock AgentCore with greater compute capacity, more efficient memory utilization, and faster startup times to help enterprises deploy and scale AI agents more quickly and cost-effectively. The greater compute capacity includes support for up to four times more compute than the previous version, with each session now supporting up to eight virtual CPUs (vCPUs) and 32GB of memory, the hyperscaler said in a statement. In addit

Google has announced Gemini Decisions API, offering developers a way to handle routine decisions in AI applications without generating text responses, weeks after OpenAI launched a competing service. Powered by the company’s new Gemini Decision 1 model, the API allows developers to evaluate up to 1,000 questions in a single request. Unlike conventional generative AI models that produce responses token by token, the system assigns probabilities to specified options, allowing applications to make

As consumer-grade hardware makes hosting local AI practical for both developers and enterprises, many are discovering that popular open-weight models come with hardcoded safety filters and state-mandated biases. In this article, I will show how to use abliteration — a representation engineering technique — to surgically delete refusal vectors from the Qwen 3.8 27B model without expensive fine-tuning or full retraining. I stumbled into this barrier firsthand while running a simple experiment. As

When Microsoft started to deploy AI models in what is now its Foundry platform, it was building on its partnership with OpenAI. OpenAI was able to use Azure infrastructure to train its models, and in return, Microsoft became the first commercial partner to use them. Azure OpenAI has been a successful product for both businesses. But as more organizations got access to OpenAI models, Microsoft’s stable of models grew to include not only OpenAI’s, but also models from Anthropic and Hugging Face, a

AWS launched a new local emulator service for its cloud services on Wednesday to help enterprises accelerate application development and testing. The service, named AWS Emulator, will allow developers and AI coding agents to build, deploy, and test AWS-dependent applications on their machines without provisioning actual cloud infrastructure and paying for it, thereby potentially reducing development time and associated costs, the hyperscaler said in a statement. To achieve this, AWS Emulator rep

Microsoft researchers have found that rewriting fewer than 10 rarely used neurons in an open-weight AI model can turn any agent built on it into a sleeper backdoor that runs an attacker’s commands on cue, with no retraining required. Researchers from Microsoft Research and Microsoft Red Team described the technique called Lazarus in a research paper . It modifies dormant neurons in a model’s final feed-forward layer to recognize a trigger and inject attacker-controlled commands into the model’s

A configuration file meant to store local command permissions on Anthropic’s Claude Code is exposing thousands of API keys, database credentials and other secrets, according to findings by GitGuardian. The code security firm identified 31,088 distinct secrets in Claude Code’s “.claude/settings.local.json” files across 23,276 public GitHub repositories. Of those, 5343 were still valid when GitGuardian detected them. The exposed credentials include database connection strings, cloud provider token

Two protocols define how AI agents connect to internal resources and collaborate with other AI agents. The Model Context Protocol (MCP) is well known because it’s promoted as a general protocol for connecting AI agents to resources and other AI agents. But that’s not completely accurate. It’s really the Agent2Agent protocol (A2A) that’s enabling AI agent orchestration . The confusion stems from how quickly platforms have adopted these protocols and marketed them as features. Users building their

I can’t decide whether Agile development is completely dead or about to make a huge comeback. Or maybe Waterfall will. Who knows? Coding agents are changing everything — and I don’t think we’ve quite figured out how they are going to update our methodology, or how we are going to organize ourselves to take advantage of them. We have barely scratched the surface. Folks who have been around a while will remember the Waterfall method for developing software. Because coding was the hardest part of a

PuppyGraph on Tuesday said it was adding an AI-assisted graph schema builder and MCP integrations to PuppyGraph 1.0, the new version of its distributed graph query engine for enterprise data. The update is designed to help developers reduce the complexity, latency, and costs associated with letting AI agents query interconnected data across thousands of database tables and multiple platforms. The AI-assisted schema builder automatically maps existing tables into graph entities, relationships, an

Microsoft researchers have uncovered a memory-safety bug in Rust’s standard library that had gone undetected for nearly 11 years, alongside other flaws that can trigger undefined behavior even in applications written entirely in safe Rust. Two of the findings affect stable Rust, including a separate issue that could cause applications to access freed memory. The nearly 11-year-old bug involves Rust’s process::exit function. Microsoft’s FORGE Lab researchers found that calling the function from a

Anthropic has introduced Claude Fable 5.2 and Mythos 5.2, claiming significant improvements in handling complex software engineering tasks while cutting API prices by 20% compared with the previous versions. Fable 5.2 and Mythos 5.2 share the same underlying model but have different safeguards. Fable is broadly available, while direct access to Mythos is restricted to vetted cybersecurity and life sciences organizations to limit potential misuse. The models are both priced at $8 per million inpu

I read the recent piece about AWS cloud capacity being “largely sold out” through 2027 , and my first reaction was fairly simple: Well, that’s that. I guess I’ll stop recommending Amazon Web Services to any client looking to host AI workloads. Clearly, the cloud is full. Please take a number. Your GPU will be available sometime after the next presidential election, assuming the data center hamsters are still running. Of course, I suspect I can still log in to the AWS console today and provision

Running a gold mine can be profitable. Finding and digging the next one is expensive. As analyst Benedict Evans argues , Anthropic’s business has the same challenge, with more than 80% of its third-quarter revenue coming from models released in the preceding six months. A lab can make good money serving its best model while spending heavily on the one it will need next, but it could make amazing money if it could go 18 months between major training runs instead of six. Reality, however, bites. T

My most-cited BYTE cover story was 1994’s Componentware , which made the case for a two-tier architecture best exemplified then by Visual Basic and its component ecosystem. The prevailing strategy for software reuse, object-oriented programming , hadn’t gained the traction we all expected. Visual Basic showed us why it hadn’t. Objects were written in the same languages, and by the same specialized coders, as the programs that reused those objects. VB flipped the script. Code-savvy component buil

Agentic AI has become a powerful enterprise business tool with increasing capabilities, and it has never been easier to address business pain points by applying a large language model (LLM) to develop a plausible automated solution. But is that solution as reliable as you need it to be, especially when the insights gleaned are used to inform critical business decisions? LLMs are incredible resources that play a significant role in a business’s agentic AI infrastructure. But when they have to wor

The UK’s digital watchdogs have warned enterprises that they can’t evade their legal responsibilities by blaming errors on AI agents that they have deployed. “By taking action on behalf of users, such as making purchases and communicating with businesses, these agents could transform how people use digital services, engage with markets and unlock wider innovation,” said the Digital Regulation Cooperation Forum (DRCF), a body comprising the UK’s Competition and Markets Authority (CMA), Financial

OpenAI is buying Pierre Computer Company in order to incorporate the San Francisco-based software developer’s Code.Storage infrastructure layer for Git into its Codex AI-powered programming tool. Pierre Computer, whose employees number just barely in the double digits, has got plenty of clout for its small size. It’s well-known for Code.Storage offering, as well as for its open-source software libraries, Trees and Diffs, which OpenAI already uses in Codex. OpenAI has experienced some of the limi

A memory-safety flaw in CPython can allow malicious AI model templates to escape code sandboxes and execute arbitrary code inside inference servers, according to security researchers at RunSybil. “By exploiting this vulnerability, attackers could gain a foothold inside servers responsible for processing millions of user requests,” Darius Houle, a security researcher at RunSybil, wrote in a blog post. “They also could strike directly at the underlying cloud infrastructure hosting those workloads.