|
Welcome back! Peter Steinberger, creator of OpenClaw, an AI agent that has taken the developer world by storm, recently made a stunning statement: he doesn’t even look at much of the AI-generated code he uses to make the agent. That could soon become the norm for a majority of people developing applications, said former GitHub CEO Thomas Dohmke. This idea is part of the impetus behind Entire, Dohmke’s new AI-code monitoring startup, which on Tuesday announced a $60 million seed funding round at a $300 million valuation led by Felicis. (GitHub owner Microsoft, Dohmke’s previous employer, also participated in the round.) Entire—named for its goal of reimaging all of software development—is developing tools for people to understand AI coding agents’ activities without having to necessarily wade into the code themselves, Dohmke said. While developers like Steinberger are increasingly letting AI handle the grunt work of coding without necessarily reviewing all of the work themselves, large companies can’t afford to take an entirely hands-off approach. “Creating apps without looking at the code is now an option. But from a security and compliance perspective, nobody can afford to ship code that nobody has looked at, you come into all kinds of legal indemnification issues,” Dohmke said. Entire’s first product, released on Tuesday, is an open source tool called Checkpoints, which keeps a log of what the AI agent is doing, including its step-by-step “thought” process and the actions it takes while writing code. To do that, it connects to developers’ command line interfaces, windows in which they give instructions (prompts) to the AI agents on what code to write. In case you’re wondering why the AI agent developers like Anthropic don’t do this, it’s because they’re focused on helping people use their proprietary agents. Checkpoints, in contrast, aims to monitor a variety of AI agents—from more than one maker—that developers like Steinberger use for different purposes. Humans can check the entire tool for information about how a part of their AI-generated application runs, or to fix potential problems. Entire currently works with Anthropic’s Claude Code and Google’s Gemini CLI—an interface that lets developers give coding instructions to the Gemini model—with plans to add support for other coding agents. Entire is part of a category of startups and cloud services known as AgentOps that observe or monitor how AI agents behave. A slew of software providers have recently released similar products that aim to make it easier for companies to keep track of how autonomous AI systems act within their applications, with firms including Microsoft and OpenAI promoting new agent-monitoring products. Entire is taking a tried and true business path for open-source firms: it’s planning to add subscription plans for cloud-hosted versions of the product in the coming months, Dohmke said. He said he came up with the idea for the company last summer while watching GitHub’s AI coding tools gain steam. After talking with Microsoft CEO Satya Nadella and other executives about his idea, Dohmke decided he wanted to focus on the product full-time, and left Microsoft in August to found Entire. “The world of software development and dev tools is about to drastically change,” Dohmke said. “That was too exciting of an opportunity for me to pass along.”
| 
