OpenAI's coding tool; AWS tackles mainframes

Welcome to Runtime! Today on Product Saturday: OpenAI previews a coding agent, AWS launches a new service designed to migrate old workloads to the cloud, and the quote of the week.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Ship it

Surf's up?: Given that coding assistants have become one of the most promising uses for large-language models, it's been a little surprising that OpenAI hasn't been more aggressive in that space as developers flock to companies like Cursor and Windsurf. While OpenAI is reportedly in talks to acquire the company behind Windsurf, on Friday it unveiled Codex, a "research preview" based on a customized version of its o3 model.

"Codex can perform tasks for you such as writing features, answering questions about your codebase, fixing bugs, and proposing pull requests for review," OpenAI said in a blog post. ChatGPT Pro, Enterprise, and Team users will get first crack at the tool, which will be free to use during the preview period but the company said it will eventually roll out pricing plans based on usage.

Math is hard: As developers flock to AI coding assistants in hopes of making their day-to-day labor more productive, Google's DeepMind division rolled out a new coding agent this week designed to take on harder challenges. AlphaEvolve was built to "target fundamental and highly complex problems in mathematics and modern computing," such as optimizing algorithms, DeepMind said in a blog post.

AlphaEvolve uses Google's Gemini Flash and Gemini Pro AI models to "propose computer programs that implement algorithmic solutions as code," the company said. Google said it used AlphaEvolve to help the company's famous Borg computing infrastructure management system run more efficiently, and apparently it also made the process of training future versions of Gemini more efficient.

More than meets the eye: Over the last 15 years AWS has invested billions in order to make it easier to move workloads onto its cloud services from any number of different legacy platforms (moving out is a little more complicated). This week it announced that AWS Transform, which it previewed at re:Invent 2024, is now generally available for potential customers still running mainframe, VMware, or .NET applications.

The new service "uses specialized AI agents to remove the heavy lifting and automate complex migration and modernization tasks for your VMware, mainframe, and .NET workloads, delivering completed transformation projects up to 4x faster," AWS said in a blog post. However, at launch AWS told potential Transform customers that they would be required to keep any workloads migrated to AWS using the service on its cloud for at least 24 months, and after Duckbill Group's Corey Quinn pointed out how terrible that would be, the company removed that clause from its service terms.

Trust the process: Last year Celonis CEO Alex Rinke was skeptical about how quickly AI agents would roll out across enterprises, telling Runtime "I think it's a huge misconception to come in and think that you're just going to plug in the agents and it is just magically going to figure it out." He was right about that, but as attention has shifted toward technologies like Anthropic's MCP that could help companies build agents that actually work, Celonis is also working on the plumbing that could help fulfill the promises of agentic AI.

This week Celonis introduced new additions to its Process Intelligence API that make "it easier to securely share PI context, metrics, and recommended actions with AI platforms such as Microsoft Copilot Studio, Amazon Bedrock, or Salesforce Agentforce," it said in a press release. It also announced that it has acquired Orchestration Engine, a service that "connects and coordinates tasks in a given process," from Emporex and built it into the core Celonis platform.

AI boom: As we saw last week at ServiceNow's Knowledge 2025 event, enterprise software companies are locked in fierce competition to convince companies to put their flavor of agent-building and management software at the heart of their AI strategies. Boomi made a name for itself as an integration platform connecting applications across corporate networks, and now wants to extend those capabilities to agents.

Boomi AgentStudio is now generally available, and it "empowers organizations to design, govern, and orchestrate all AI agents at scale within a secure, no-code environment," the company said in a press release. It also introduced four new agents that help companies build and manage their APIs.

A MESSAGE FROM CIRCLECI

Enterprise leaders: What's the real cost of slow pipelines? Teams using CircleCI reclaimed $4.5M in productivity annually. Deploy on your cloud or behind your firewall. Learn more

Stat of the week

One clear divide between companies that have managed to deploy generative AI apps in production and those who have not is in most cases, the former already had a modern data strategy in place. New research from Fivetran bears that out: "42% of enterprises say more than half of their AI projects have been delayed, underperformed, or failed due to data readiness issues."

Quote of the week

"All code has security risks, and the more code you have, the more security risks there are. As this speeds up, we need the other side of tooling to speed up too, to catch those issues." — Chainguard CEO Dan Lorenc, on the rise of AI-assisted coding and the inherent security implications.

The Runtime roundup

CoreWeave's stock rose 22% Friday after it disclosed that Nvidia owns a 7% stake in the AI cloud provider.

Anthropic closed on a $2.5 billion credit facility, adding to a $4 billion line of credit it secured last year.

A MESSAGE FROM CIRCLECI

Stop losing millions to slow deployments. CircleCI's enterprise platform delivers 664% ROI while cutting development time by 50%. See how tech leaders are accelerating innovation. Learn more

Thanks for reading — see you Tuesday!