Today on Product Saturday: AWS overhauls its managed OpenSearch service with agents in mind, Clickhouse hits a revenue milestone and rolls out a tool for building agents, and the quote of the week.
Today: Snowflake's massive quarter shows once again that AI deployments hinge on data quality, Anthropic drops Opus 4.8 and picks up an enormous pile of cash, and the latest enterprise moves.
Today: Anthropic's Mythos Preview security model is finding a lot of bugs, but plugging those holes is still a challenge, an unlikely company joins the $1 trillion club, and the latest funding rounds in enterprise tech.
Today on Product Saturday: AWS overhauls its managed OpenSearch service with agents in mind, Clickhouse hits a revenue milestone and rolls out a tool for building agents, and the quote of the week.
Welcome to Runtime! Today on Product Saturday: AWS overhauls its managed OpenSearch service with agents in mind, Clickhouse hits a revenue milestone and rolls out a tool for building agents, and the quote of the week.
Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week.
Open and shut: AWS's OpenSearch service was designed to help developers embed search into their own applications (and to get around Elastic's relicensing of Elasticsearch, but that's a long story). This week it introduced a new version of the serverless managed service built to accommodate AI agents, which will be putting new stress on search engines as they roll out.
"The next generation of OpenSearch Serverless scales from zero to thousands of requests per second and back to zero when idle, offering up to 60% cost savings compared to the cost of OpenSearch Service clusters provisioned for peak capacity," the company said in a blog post. AWS engineers built a new storage layer underneath that service that helps manage resource consumption; "since we’re able to predict what you need and we’re able to deliver and scale back down in a very rapid fashion, you’re going to automatically save money," AWS's Tia White told The New Stack.
Just one ping: When (and if) AI agents start running more business applications, verifying that all those agents deserve access to sensitive corporate resources will be challenging for identity tools designed around people. Ping Identity expanded its core identity management service this week to address agent identity-verification, jumping on the headless software trend.
The new features "help enterprises support AI-driven operations without creating a parallel identity stack, preserving governance and control across human, non-human, and AI-agent access," the company said in a press release. The new features included a headless UI that allows customers to program agent identities through tools like MCP, as well as identity verification tools for the OpenClaw crazies.
House calls: ClickHouse has been a rising star in the database world for several years, and this week it announced that it crossed the $250 million annual run-rate revenue milestone at its second-ever user conference. The company, which has raised $1.2 billion in funding, also used the occasion to unveil ClickHouse Agents, a no-code tool for building agents on top of data stored in ClickHouse.
The new tool allows companies to build agents that can "connect natively to ClickHouse and to any MCP-compatible third-party system, drawing context from across an organization's stack, including a native integration with the AWS Agent Registry," the company said in a press release. It also launched a service called CostBench that allows customers to compare the costs of applying query engines like Databricks and Snowflake to ClickHouse data.
Around the loop: If CoreWeave is ever going to become a profitable AI cloud provider, it will need to convince customers to use its software services on top of the AI workloads they run on its servers. This week the neocloud introduced a new set of services that aim to improve the performance of those workloads by allowing customers to share data across the training and inference process.
"With reinforcement learning, production inference, agent observability, and autonomous improvement working as one closed loop, agents not only become more reliable, they compound in capability over time," the company said in a press release. It's another reminder that a set of best practices for running AI workloads is still a work in progress: "Assuming that you're going to get the perfect answer in your first deployment has been consistently proven incorrect in software development and cloud over the years, and it's no different in AI," CoreWeave's Corey Sanders told AI Business.
Context clues: Companies that have successfully deployed AI agents tend to have taken two steps before they got started; they organized their data using modern tools like data lakehouses, and they figured out how to provide context around that data to their agents. DataHub launched a new service this week that promises to help customers address that second step.
The new version of DataHub Cloud "serves as a context layer that sits between analytics agents, like Databricks Genie and Snowflake Intelligence, and enterprise data from data stores, like data warehouses and data lakes, to give agents trusted context to get analytics right and smarter with every query," the company said in a press release. That layer could become another sticky control point for agentic AI vendors, according to Constellation Research's Michael Ni: "Buyers should be clear on their context management requirements, as vector memory isn't business meaning, business meaning isn't governance, and governance isn't execution," he told VentureBeat.
According to new data from Cursor, an increasing number of developers are letting their AI agents take the wheel. At the beginning of the year developers only allowed 7.2% of AI-written code changes to reach the commit stage, but as of May, that number had risen to around 36%, "suggesting that developers are trusting agents to carry more work through the commit flow," Cursor said in the report.
"What's happened this year is initially the year started with all software companies going down. And what's happened in the last few weeks is we're seeing a separation between winners and losers." — D.A. Davidson's Gil Luria, commenting on earnings results from Snowflake and Salesforce this week.
SentinelOne laid off 8% of its employees and trotted out the usual "AI is making is more effective" justification while also reducing revenue guidance for the upcoming year, two data points that are totally unrelated.
Microsoft stirred up a hornet's nest this week — days before its annual Build developer conference — after threatening legal action against a security researcher who found several flaws in its Microsoft Defender security software, according to TechCrunch.
Thanks for reading — see you Tuesday!
Today: Snowflake's massive quarter shows once again that AI deployments hinge on data quality, Anthropic drops Opus 4.8 and picks up an enormous pile of cash, and the latest enterprise moves.
Today: Anthropic's Mythos Preview security model is finding a lot of bugs, but plugging those holes is still a challenge, an unlikely company joins the $1 trillion club, and the latest funding rounds in enterprise tech.
Today: GitHub and Grafana share more details on how hackers stole their code, Google Cloud booted Railway off its infrastructure for no clear reason, and the latest enterprise moves.
Today: Google unveils new AI models and developer tools as it seeks to get in on the agentic coding race, believe it or not, another supply chain attack targeting npm packages just dropped, and the latest funding rounds on enterprise tech.
