# Runtime: Cloudflare gets serious about data > Source: > Published: 2026-10-03 15:00:38+00:00 *Welcome to Runtime! Today on Product Saturday: Cloudflare's new data query engine builds on open formats and serverless thinking, MongoDB levels up during a tough week, and more.* *Please forward this email to a friend or colleague! If it was forwarded to you,* __sign up here__ *to get Runtime for free every week, or* __level up here__*.* ### Ship it **Fill 'er up**: Four years into the generative AI boom, the companies that have [__managed to implement AI and get results__](https://www.thestack.technology/ron-van-kemanade-group-coo-lloyds-on-pivoting-to-agents-eying-200-cobol-applications/) started with clean, well-organized data to feed into AI models and harnesses. Such goals are, of course, easier said than done at enterprises with dozens of data silos constructed by different divisions, acquisitions, and legacy technology providers. Databricks and Snowflake have probably been the biggest beneficiaries from the push to modernize older data pipelines and storage strategies, but [__the rise of open data formats__](https://www.thestack.technology/how-open-file-formats-shook-up-the-snowflake-databricks-rivalry-and-created-a-new-battleground/) has leveled the playing field and allowed companies to store data wherever they like and bring the best data query engine for their needs to that data, rather than the other way around. After announcing the Cloudflare Data Platform a year ago this week, the company made it official this week with the launch of Cloudflare Basin. Basin, which is now generally available, "brings an end-to-end analytics platform to the Developer Platform, enabling you to collect data from a variety of sources, such as apps, infrastructure, devices, and other Cloudflare services, then query it to answer analytical questions," the company [__said in a blog post__](https://blog.cloudflare.com/cloudflare-data-platform/?ref=thestack.technology). Based around Apache Iceberg, the platform includes data pipelines, a data catalog, and the R2 SQL engine. Cloudflare is pushing the speed and affordability of the platform as a selling point against more well-established data query engines, but Basin customers will still be able to read and write data using query engines compatible with Iceberg, which is getting to be most of them at this point. "That kind of data portability is only possible with free egress, which allows developers to access their data in Cloudflare from the wide variety of tools in the ecosystem, regardless of region or cloud," it said in the post. ### Delivery, continued **Atlas shrugged**: It was a very strange week for MongoDB, which saw the return of former CEO Dev Ittycheria to the big chair after [__the abrupt departure of Chirantan “CJ” Desai to Meta__](https://www.thestack.technology/meta-makes-its-enterprise-ambitions-official-poaches-mongodb-ceo-desai/), the ringing of the opening bell at the Nasdaq, and a point release for its flagship database. MongoDB 9.0 doubles throughput performance on large instances and improves throughput by 20% on transactional workloads, according to the company. MongoDB also introduced Atlas Infinite, a new edition of the cloud-based managed version of MongoDB that "separates compute from storage so each can scale independently, giving any workload, agentic or not, the elasticity, responsiveness, and retrieval accuracy they demand on real-time operational data," it [__said in a press release__](https://www.mongodb.com/company/newsroom/press-releases/mongodb-launches-mongodb-9-0-the-best-version-ever-built-and-atlas-infinite-for-ai-scale-demand?ref=thestack.technology). There are now two versions of Atlas for customers that want to consume MongoDB through the cloud: Infinite, and Atlas Core, which is the same cloud version customers have been using to date with a new name. **Go with the Flow**: ServiceNow is the big kid on the block in the IT service management space, but [__Freshworks has been nipping at its heels__](https://www.thestack.technology/freshworks-ceo-dennis-woodside-we-can-be-the-servicenow-alternative/) over the last several years by focusing on the small and medium-size business market. This week ServiceNow introduced Flow, a new version of its flagship service that was designed for "speed and simplicity," the company [__said in a press release__](https://newsroom.servicenow.com/press-releases/details/2026/Introducing-Flow-by-ServiceNow-a-new-AI-service-desk-that-deploys-instantly/default.aspx?ref=thestack.technology). "Flow is a new standalone product with faster deployment and the same reliable results expected from ServiceNow's comprehensive service management portfolio," the company said. It's a take on [__the headless software concept__](https://www.thestack.technology/enterprise-software-brings-out-the-guillotine/), which involves designing enterprise software tools to work with agents or collaboration tools like Slack rather than forcing customers to log into a distinct user interface. **The Decision**: TypeSafe lit a fire underneath enterprise tech two weeks ago with [__the release of Jev__](https://www.thestack.technology/runtime-jev-is-an-llm-without-the-ll/), a so-called "decision model" that generates structured output designed to be consumed by software agents. AWS jumped on the bandwagon this week with the release of Strands Decider 2B, which was released under the Apache 2.0 license. The new model is "a 2 billion parameter model, suitable for running on a local CPU or GPU, which can return answers to meaningful questions in tens of milliseconds," AWS said in a blog post. It's based on a fine-tuned version of Alibaba's Qwen, which is quite a twist in the ongoing discussion over open models and distillation. ### The rest of The Stack **Tale as old as:** The UK government's trial for an the country's largest AI-enabled pension is not going to plan. Strategic outsourcer Capita, who won the £239 million ($316 million) contract to deliver the pension scheme, is being investigated by the National Audit Office after missing "key service levels" repeatedly. **Know your enemy**: Microsoft's annual Digital Defense Report dropped with Redmond reminding IT leaders that identity management is going to be one of the most prominent playgrounds for AI-powered attacks. The report emphasised the importance of short-lived credentials and strict least privilege access. **Chipping in**: Synopsys won a vote of confidence from OpenAI this week, partnering with the AI model company to train a model that can be used in the process of designing new chips. GPT-Synopsys promises to help those experience designers come up with new ideas and let novice designers get up to speed more quickly. **Alphabet soup**: There's nothing more that enterprise tech likes more than an acronym and a standards fight. Mary Branscombe takes a deep look at the competition and cooperation between some of the leading protocols that help govern AI agent activity, including MCP, A2A, and ARD. ### Quote of the week "I've talked with some folks who have built harnesses at security vendors that said I should title a report I was working [on] as 'Agent Harnesses, What Are They?'" — *Forrester's Jeff Pollard,* __explaining to The Stack__ *how the harness became the enterprise tech product of 2026, even if people are still struggling to understand how they work and how to use them safely.* ### We're also reading: [**__Inside Larry Ellison's Wide Web of Power, Money and Politics__**](https://www.vanityfair.com/story/larry-ellison-profile?ref=thestack.technology): Vanity Fair goes deep on the mercurial Oracle founder, who seems to spend more time outside enterprise tech these days than in it. *Thanks for reading — see you Tuesday!*