One of the fastest-growing sectors in artificial intelligence is the development of chatbots that assist with writing computer code. These AI tools are changing how software engineers work, allowing them to focus on higher-level tasks while delegating routine coding to machines.
Cat Wu, project manager for Anthropic’s Claude Code, explained the shift: “The essence of it is you’re no longer in the nitty-gritty syntax. You’re not looking at every single line of code. You’re more trying to communicate this higher-level goal of what you want to accomplish.” Wu also clarified her stance on terminology: “We definitely want to make it very clear that the responsibility, at the end of the day, is in the hands of the engineers.”
Anthropic has recently released an updated version of its Claude chatbot, Sonnet 4.5, which it claims is highly effective for coding and complex tasks. The company reported that Sonnet 4.5 was able to autonomously code for over 30 hours on a project for London-based startup iGent before its public release.
According to Gartner analyst Philip Walsh, coding and software engineering are among the most common business uses for generative AI chatbots such as Claude, ChatGPT, and Google’s Gemini. Walsh stated, “That is often the first thing large organizations go after. I think there’s broad recognition among these AI model providers that coding is really where they’re getting the most traction.” He noted that Anthropic’s products are popular with developers but emphasized strong competition from other companies.
San Francisco and Silicon Valley remain central hubs for this technology race. Major players include OpenAI, Anthropic, Anysphere, Cognition, Harness, and Microsoft-owned GitHub. Jeff Wang, CEO of Windsurf—a startup acquired by Cognition after being pursued by several tech giants—said: “This is the most competitive space in the industry right now.”
Some AI assistants offer simple autocomplete functions similar to those found in email or messaging apps. More advanced versions act as agents capable of carrying out entire programming tasks independently.
Anthropic’s internal development process began when Boris Cherny created a prototype tool for personal use; it quickly spread throughout his team due to its utility. Wu described this adoption: “Over time, we realized that it was just virally spreading within Anthropic.”
The company reports that about 39% of users employ Claude primarily for coding purposes.
OpenAI reports different usage patterns for its ChatGPT product but has recently introduced GPT-5-Codex to improve performance on extended coding assignments.
Startups like Anysphere have become important clients for major AI model providers. Anysphere’s Cursor tool relies heavily on Anthropic’s Claude and has formed a partnership with OpenAI.
The term “vibe-coding” was coined earlier this year by researcher Andrej Karpathy when using Cursor’s Composer with Claude Sonnet: “There’s a new kind of coding I call ‘vibe coding’, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists,” he wrote online. Karpathy described issuing spoken instructions rather than typing lines manually: “It’s not really coding – I just see stuff, say stuff, run stuff, and copy-paste stuff, and it mostly works.”
Despite some platforms promoting no-code approaches accessible through chat interfaces—such as Sweden-based Lovable—most advanced tools target professional programmers.
Concerns have emerged about potential job losses among software professionals due to increased efficiency enabled by AI tools. While some executives claim productivity gains could reduce staffing needs, Walsh from Gartner offered a different perspective: “There’s so much software that isn’t created today because we can’t prioritize it,” he said. “So it’s going to drive demand for more software creation, and that’s going to drive demand for highly skilled software engineers who can do it.”
Recent research from Stanford University indicates early-career workers (ages 22-25) in fields exposed to AI have experienced significant declines in employment rates as automation becomes more capable; AI systems solved nearly 72% of standard coding problems in 2024 compared with just over 4% one year earlier.
Walsh cautioned against interpreting “vibe-coding” as something non-technical employees can easily perform: “That’s simply not happening. The quality is not there. The robustness is not there. The scalability and security of the code is not there,” he said. “These tools reward highly skilled technical professionals who already know what ‘good’ looks like.”
Wu advised her younger sister currently studying in college that pursuing a career in software engineering remains worthwhile despite rapid technological change.
