Over the past two years, software developers have watched AI coding tools evolve from advanced autocomplete to, in some cases, tools that can build entire applications from text prompts. Tools like Anthropic’s Claude Code and OpenAI’s Codex now allow people to work on software projects for hours at a time, writing code, running tests, and fixing bugs under human supervision. OpenAI says it is currently using Codex to build itself, and the company recently published technical details about how the tool works under the hood. This made many people wonder. Is this just AI industry hype, or are things actually different this time?
To find out, Ars reached out to some of Bluesky’s professional developers and asked them how they feel about using these tools in the wild. The responses showed that employees generally agree that the technology works, but are divided on whether that’s entirely good news. Although this is a small sample self-selected by potential participants, their opinions are still valuable as professionals working in this field.
David Hagerty, the POS system’s developer, told Ars Technica beforehand that he was skeptical of the marketing. “All the AI companies are touting their capabilities,” he says. “Don’t get me wrong, LLMs are revolutionary and will have a huge impact, but don’t expect them to write the next great American novel or anything like that. That’s not how they work.”
Roland Dreier, a software engineer who has made significant contributions to the Linux kernel in the past, told Ars Technica that while he acknowledges the hype, he has been keeping a close eye on advances in the AI field. “It sounds like incredible hype, but cutting-edge agents are amazingly good at the moment,” he said. Dreyer said there have been “incremental changes” over the past six months, especially after Anthropic released Claude Opus 4.5. We used to use AI for autocomplete and occasional questions, but now we’re telling the agent, “This test is failing. Please debug and fix it,” and expecting it to work. He estimated that complex tasks like building Rust backend services using Terraform deployment configurations and Svelte front ends would be 10x faster.
