Cursor, an emerging AI tool used for generating and editing lines of code, saw one of the most unusual bug reports ever posted by a user on its official forum on Saturday, March 15. The developer was using Cursor to generate code for a racing game when the AI programming assistant abruptly refused to continue doing its work and, instead, told the developer to “develop the logic yourself”. "Generating code for others can lead to dependency and reduced learning opportunities,” Cursor’s response read, as per the bug report. While AI assistants are known to have refused to complete their work in the past, Cursor’s response oddly ties into a growing debate that has gripped the software engineering community recently. With the rise of AI coding tools such as GitHub Copilot, Claude 3.7 Sonnet, Windsurf, Replit, Cursor and Devin AI, the tech industry appears to be divided over critical questions such as: Will AI revolutionise coding? If so, how much coding will it actually automate? What does vibe coding mean? And should people still invest time in learning to code? To understand what AI really means for coding skills, let’s take a look at the different dimensions of the unfolding debate. ‘AI will take over coding entirely’ While concerns about the impact of AI on software development have been brewing for some time now, the issue came to a head when Dario Amodei, the CEO of the AI startup Anthropic, made a proclamation that rattled more than a few. Speaking at a Council of Foreign Relations event last week, Amodei said that AI will be able to generate all the code required for developing software in a year’s time. "I think we will be there in three to six months, where AI is writing 90 per cent of the code. And then, in 12 months, we may be in a world where AI is writing essentially all of the code," he said. Amodei further opined that software developers will have a role to play in the short term. "But on the other hand, I think that eventually all those little islands will get picked off by AI systems. And then, we will eventually reach the point where the AIs can do everything that humans can. And I think that will happen in every industry," he added. The bold prediction made by Anthropic’s CEO was also echoed by Garry Tan, the president and CEO of startup incubator Y Combinator. "For 25 per cent of the Winter 2025 batch, 95 per cent of lines of code are LLM-generated. That's not a typo," Tan wrote in a post on X. For 25% of the Winter 2025 batch, 95% of lines of code are LLM generated. That’s not a typo. The age of vibe coding is here. — Garry Tan (@garrytan) March 5, 2025 Naveen Tewari, founder and CEO of InMobi, also stated that the mobile ads platform was on track to achieve 80 per cent automation in software coding by the end of 2025. “My CTO will deliver 80 per cent automation in software coding by the end of this year. We have already achieved 50 per cent. The codes created by the machine are faster and better, and they fix themselves,” Tewari said. On the future of software engineering jobs, Tewari said, “I think my software engineers will go away. They will not have jobs in two years.” In an appearance on NYT’s Hard Fork podcast, Anthropic’s Amodei said he envisions doing more with fewer developers over the next year as well. ‘AI is a tool, not a replacement’ Disagreeing with Amodei, IBM CEO Arvind Krishna said that only 20 to 30 per cent of code will be AI-generated as opposed to 90 per cent. “Are there some really simple use cases? Yes, but there’s an equally complicated number of ones where it’s going to be zero,” Krishna said during an interview at SXSW 2025, an annual tech, music, comedy, and film conference, held last week in Austin, Texas, US. He also opined that AI will make programmers more productive rather than eliminate their jobs. However, in 2023, IBM had reportedly planned to halt hiring for back-office roles as the tech giant expected to replace those functions with AI. Sridhar Vembu, the founder of SaaS major Zoho, said that while AI programming tools can excel at generating boilerplate code for many projects, it can only offer 10 to 20 per cent productivity gains depending on the nature of the project. What is ailing the software job market is not AI taking away jobs (not yet anyway). Here is my thesis, as a participant and observer of software for 30 years. Over those 30 years: 1. Massive over-capacity steadily developed in enterprise software due to a flood of VC, PE and… — Sridhar Vembu (@svembu) March 11, 2025 Offering a more nuanced take, Vembu said he was pessimistic about the software job market due to reasons other than AI. The former Zoho CEO, who stepped down a few months ago, believes that the enterprise software industry is plagued by massive inefficiencies caused by influx of venture capital, private equity, and IPO-driven funding. “It is those multiplied inefficiencies in IT, built up over decades, that are facing a reckoning now,” he said. What is vibe coding? Beyond the remarks on AI versus jobs, another dimension of the debate is focused on how large language models (LLMs) can turn any user with basic communication skills into a new type of programmer. LLMs have significantly lowered the barrier of entry to software development. This is where vibe coding comes in. ‘Vibe coding’ is a playful term that was recently coined by former OpenAI researcher Andrej Karpathy. It is the growing practice among users which involves asking an LLM like ChatGPT to create a software programme based on a specific description without knowing how the code actually works. “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. It's possible because the LLMs (e.g. Cursor Composer [with] Sonnet) are getting too good,” Karpathy said in a post on X. 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. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper… — Andrej Karpathy (@karpathy) February 2, 2025 “Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away,” he added. However, questions have been raised on whether the vibe coding approach can produce reliable code that is suitable for real-world applications. Experienced developers have also criticised LLMs and AI-assisted coding tools as becoming a crutch for young coders, arguing that they rely too heavily on AI and lack conventional programming knowledge. Learn to code or learn to prompt? Despite concerns about AI automating software engineering jobs, many believe that learning to code is still more valuable than ever. Andrew Ng, AI expert and co-founder of Coursera, said that discouraging others from learning programming on the grounds that AI will automate it, is “some of the worst career advice ever given.” “In the 1960s, when programming moved from punchcards (where a programmer had to laboriously make holes in physical cards to write code character by character) to keyboards with terminals, programming became easier. And that made it a better time than before to begin programming,” he said in a blog post. The AI industry leader also emphasised the importance of upskilling amid the rise of AI-assisted coding tools. Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the… — Andrew Ng (@AndrewYNg) March 13, 2025 “One question I’m asked most often is what someone should do who is worried about job displacement by AI. My answer is: Learn about AI and take control of it, because one of the most important skills in the future will be the ability to tell a computer exactly what you want, so it can do that for you,” he said.
