When Dan Lorenc, CEO and co-founder of cybersecurity startup Chainguard, was asked which AI coding tools he expected to see less use in the next year, his answer was spot on. “Anything but Claude Code.”
Lorenc is never alone. In a survey of more than 20 startup founders and venture capitalists, Business Insider found that Anthropic’s Claude Code has quickly become the default AI coding tool within startups, with a growing consensus that it’s highly praised for how it handles complex engineering tasks and autonomous workflows.
“The change in how code is developed is similar to how woodworking evolves from hand tools to power tools and then to a full assembly line,” Lorenc said. “For years, we carved everything by hand. Then AI came along and gave everyone a circular saw.”
Venture capital has poured billions into AI-coding startups like Lovable, Replit and Cursor, and last month the company announced it had given SpaceX the right to buy parent company Anysphere for $60 billion later this year, or pay $10 billion if the deal doesn’t go through. Meanwhile, investors are eager to back Claude Cord maker Anthropic, which is expected to go public by the end of the year.
The stakes are high, as AI coding has become one of the most obvious commercial use cases for generative AI, and startups increasingly rely on these systems not only to write code faster, but also to automate engineering tasks that once required entire teams.
Matthew Burris, senior director of research at Venture Studio Forum, said 12 weeks ago he had never written a single line of code.
“Claude Cord completely changed that,” he said. “Today, I’m shipping tools that rival what you’d get from a six-figure consulting contract.”
He finds that genetic workflow particularly useful. “Claude Code doesn’t just autocomplete lines; it actually reasons through architecture, researches approaches, and builds iteratively,” he said, adding that he intentionally avoids OpenAI. “I have real concerns about OpenAI’s approach to safety and release practices, their aggressive and increasingly monopolistic competitive position, and frankly, I don’t trust their data.”
Zhongtian Wang, head of technology at AI biometric startup VaryAI, says Claude Code is now embedded in every part of the company’s workflow.
“We started using it last year to write code and fix bugs,” he said. “We currently use it to automate our entire internal processes, including our quality assurance pipeline, deployment workflows, incident investigations, and project management.”
The cursor still exists but is losing ground
Although the cursor is still widely used, the founders consistently described it as a secondary tool that fades out the cursor.
“They built a great product and showed early on what AI-powered coding could look like,” said Danny Freed, CEO and founder of healthcare AI startup Blueprint. “But Claude Code’s agent workflows are even more advanced, especially for more complex tasks.”
Rami Alhamad, co-founder and CEO of personalized nutrition startup Alma, says he still uses Cursors for simple tasks, but is increasingly relying on Claude code for more demanding tasks. Nearly every line of code his startup currently ships is generated by AI and then reviewed and refined by him or his co-founders.
“The gap between what it can handle and what is required of senior engineers continues to shrink,” he said. “We are now reaching for tasks that six months ago would have been considered too complex for AI: tasks that involve multiple repositories, require architectural decisions, or require context across the entire codebase.”
Volodymyr Giginiak, co-founder of legal AI startup Wordsmith AI, also said he expects to use less Cursor and more Claude Code.
“Its development speed and versatility are unparalleled, supporting everything from simple fixes to complex multi-step workflows,” he said. “Tight integration with Frontier models means that their value increases as the models improve.”
Once a groundbreaking AI coding product, Microsoft’s GitHub Copilot has barely been talked about.
“There is no longer any meaningful advantage over new tools,” said Ben Seri, co-founder of AI security startup Zafran Security.
Startups are not monogamous
Even though Claude is dominant, the startups we spoke to are not completely monogamous.
Tony Liu, a partner at venture capital firm Costanoa Ventures, said comparisons between the tools are “a piece of cake.” The key, he says, is not the tools themselves, but “how you integrate these tools into your workflow.”
Kelsey Falter, co-founder of creative development studio Mother.tech, explained that they use Claude for development, Codex for local code reviews, and Gemini for PR reviews.
Itamar Tal, co-founder of AI security startup Tenzai, said his team is moving to a more modular approach, “mixing and matching” tools such as Codex, Vercel, and Amp. At the same time, they have avoided hosted vibe coding platforms such as Replit and Lovable.
“These are great for a quick start, but they lack security and production readiness and are difficult to scale past a certain point,” he said. “Vercel remains a strong exception due to its technical depth and configurability.”
Like everyone else in our study, Tal expects to use Claude Code more than just for writing software.
Recently, the startup experienced an issue with the video system in its main hall, resulting in glitchy images during Zoom meetings.
Instead of calling IT support, the team installed the Claude code directly on the controller and granted system-level access to investigate.
“Within about 25 minutes, we identified the hardware incompatibility issue and suggested a fix,” Tal said, estimating that the process saved “time and thousands of dollars in IT work.”
He said the experience reflects broader changes occurring within the company as old-school spreadsheets are replaced with internal tools built by Vibe Coding.
“It’s an incredibly exciting time to be building,” he said. “Development has never felt so fast and dynamic. We’re constantly evolving how we use these tools, but it’s clear we’re still in the early stages.”