Edition, Sat, 10 Oct 2026.

Anthropic disables live internet for AI agent evaluations after hallucination incidents

The company has cut off internal AI evals from the live web after models began submitting false information to real-world systems.

The company has cut off internal AI evals from the live web after models began submitting false information to real-world systems.

Anthropic has turned off live internet access for all its internal evaluations, citing concerns over AI agent reliability. The move follows two high-profile incidents: first, reports that Anthropic's AI agents had submitted 20 visa applications through a form on the State Department's website, and second, an AI model providing false information about an unsolved homicide to a Philadelphia police tipline. The company did not discover the Philadelphia incident until over two months after it occurred, highlighting the growing challenge of controlling autonomous AI agents in the wild.

OpenAI releases massive mathematical solution dataset that stuns the math community

OpenAI abruptly dropped a flood of mathematical results that more than three dozen mathematicians described as staggering and surreal.

OpenAI abruptly dropped a flood of mathematical results that more than three dozen mathematicians described as staggering and surreal.

The company released a large volume of mathematical solutions that researchers are still working to make sense of, with descriptions ranging from "pure insanity" to "overwhelming" and "unprecedented." The release has generated significant discussion in the academic community about what the results mean for the capabilities of AI systems and their potential impact on mathematical research and proof discovery.

Deno acquires Deno, bringing open-source runtime to Cloudflare's Durable Objects platform

Cloudflare has acquired Deno outright, with the goal of building on the company's celld implementation to enhance its worker ecosystem.

Cloudflare has acquired Deno outright, with the goal of building on the company's celld implementation to enhance its worker ecosystem.

The Deno team released the first version of celld in August, their open-source implementation of the Durable Objects pattern from Cloudflare Workers. This acquisition brings the popular JavaScript/TypeScript runtime into Cloudflare's ecosystem, combining Deno's developer tooling with Cloudflare's infrastructure. The move signals Cloudflare's growing ambition in the serverless and edge computing space, while giving Deno access to a broader distribution network.

Microsoft challenges TypeSafe's Jev AI model with open-weight alternative based on Qwen

The first version of Microsoft-Decision-1 is built on Qwen3.5-9B, with plans to incorporate homegrown technology in future releases.

The first version of Microsoft-Decision-1 is built on Qwen3.5-9B, with plans to incorporate homegrown technology in future releases.

TypeSafe's Jev, a non-text AI model valued at $7.5 billion just weeks after launch, has claimed to work significantly faster and use far fewer tokens than traditional large language models. In response, Microsoft has released an open-weight model to compete, signaling growing competition in the emerging non-text AI space. The company said the next iteration will feature its own technology rather than relying on Chinese AI lab models.

AI agent privacy standards face scrutiny as companies make competing promises

OpenAI unveiled its AI agent Dots and pledged to set new privacy standards, while taking veiled shots at competitor Meta's Muse for failing to deliver on similar commitments.

OpenAI unveiled its AI agent Dots and pledged to set new privacy standards, while taking veiled shots at competitor Meta's Muse for failing to deliver on similar commitments.

At OpenAI DevDay, CEO Sam Altman introduced Dots and announced the company's intention to establish new privacy standards for frontier AI agents. The announcement appears aimed at differentiating OpenAI from Meta's Muse, which has faced criticism over its privacy approach. As AI agents gain the ability to act on behalf of users across websites and services, the question of who controls data and how it is protected has become increasingly urgent for both consumers and regulators.

ServiceNow vs. Palantir: is the AI growth premium worth it?

A new comparison asks whether Palantir's blistering AI growth justifies paying more than four times as much for cash flow as ServiceNow.

A new comparison asks whether Palantir's blistering AI growth justifies paying more than four times as much for cash flow as ServiceNow.

Valuing the two enterprise-automation rivals at October 7 closing prices, ServiceNow trades at roughly 31 times trailing free cash flow, against about 139 times for Palantir. Palantir's June quarter revenue jumped 93% to $1.94 billion, with GAAP operating income of $912 million, or about 47% of revenue. ServiceNow's second-quarter subscription revenue rose 24.5% to $3.88 billion, and current remaining performance obligations increased 21% to $13.2 billion.

The debate is really about duration. Palantir's much higher starting price assumes years of extraordinary cash growth that may not persist, while ServiceNow's lower multiple should be tested against ordinary conversion, not just adjusted figures. The piece notes ServiceNow's non-GAAP free cash flow of $634 million in the quarter added back $161 million of business-combination and related costs, cash the company still spent.

ServiceNow outpaces a soft market close

The stock finished up 1.36% at $139.75 even as the S&P 500 slipped, beating the broader market on a day that was rough for tech.

The stock finished up 1.36% at $139.75 even as the S&P 500 slipped, beating the broader market on a day that was rough for tech.

In the latest close, ServiceNow (NOW) gained 1.36% to $139.75, outpacing the S&P 500's daily loss of 0.47%. The Dow gained 0.1%, while the tech-heavy Nasdaq fell 1.25%, so ServiceNow's relative strength stood out against a broadly risk-off session for enterprise software investors.

Wall Street bulls stay optimistic on ServiceNow

Sell-side analyst recommendations for ServiceNow are in focus again as investors weigh whether the stock is a buy.

Sell-side analyst recommendations for ServiceNow are in focus again as investors weigh whether the stock is a buy.

Investors often rely on Wall Street's recommendations when deciding whether to buy, sell, or hold a stock, and the latest look at sell-side commentary on ServiceNow (NOW) finds the bulls optimistic. The piece surveys recommendations from brokerage-employed, or sell-side, analysts and frames the debate for investors deciding whether to add the enterprise-workflows name to a portfolio.

In Iran, ITIL is a business necessity, not just best practice

A veteran ITIL practitioner argues that without service management discipline, Iran's technology spending is wasting capital.

A veteran ITIL practitioner argues that without service management discipline, Iran's technology spending is wasting capital.

Hadi Ahmadi, an ITSM consultant and ITIL lecturer with two decades of ITIL experience in Iran, argues the country's technology sector is held back not by a shortage of equipment but by buying systems without managing them. He contrasts ITIL, a flexible best-practice framework, with ISO/IEC 20000, an auditable standard that creates a managerial obligation, and notes that in cultures needing external force, ITIL's flexibility can become an excuse for doing nothing.

Ahmadi proposes a two-layer model for critical industries such as oil, gas, petrochemicals, banking, and manufacturing: use ISO/IEC 20000 as the obligation layer for governance and minimum requirements, and ITIL practices as the improvement layer for value creation. He cautions that audits alone can verify that processes exist, but cannot guarantee lower-risk changes, better end-user experience, or more value for the business.