Latest news with #vibecoding
Forbes
23-05-2025
- Forbes
Unboxing The State Of AI With Andrew Ng
Alt title: Looking at Vibecoding and Human Skill Sets Description: We're in the middle of a revolution in tech, and part of it is this idea of vibe coding. For a lot of people, it started with Andrej Karpathy's now infamous text about just letting the computer do the work, and leaning back and chilling and vibing to it. Pretty soon everyone was talking about 'vibecoding,' the idea that you don't have to know how to write a program if you just ask AI to do it for you. On the one hand, this brings a fundamental democratization of tech to a larger audience. On the other, what does it mean for coding? I came across this example of a team that wanted to build a game from scratch using AI. So they tried asking Claude. The computer wrote the code, all right, but they were left with buttons that didn't work, and serious bugs that needed to be fixed. Eventually, the computer fixed the bugs, but only with extensive prompting from the human. Here's how the author described what happened after an initial failed attempt where the program didn't launch: 'The AI went back to work, and its second attempt actually launched. I also cheated a bit and checked the code, noticing another issue: … I continued this back-and-forth with Claude, refining through natural language rather than code edits. Fourteen iterations later, I had something satisfactory enough for me to share without being ashamed.' So the takeaway here is that you may not have to do the hard coding, but you'll still have to move the program development along by helping the computer make decisions or correcting its mistakes, however you describe that process. In the April event at Imagination in Action, I interviewed Andrew Ng about this and other parts of the tech world. He's got a long and impressive career, including academic work at MIT. One of the big points that Ng brought up when I asked him about vibecoding was that in past iterations of this revolution, when we make advances, we find that people still get value out of their coding skills. He mentioned everything from punchcards to COBOL: after COBOL was developed, he said, people were wondering if they still needed to code or not. Ng pointed out that today, he personally wouldn't hire people who don't know how to code. It's still a good skill to have, he insisted, even if the machine can do it, to some extent. 'Last year, there were some people advising others to not learn to code,' he said. 'I think we'll look back on that as some of the worst career advice ever given, because as AI helps with coding, coding gets easier, and that means more people should do it, not fewer.' Ng also mentioned excessive hype in the industry, and attacks on open source that often get thrown together with safety issues. 'I think there are a few lines of hype that have been amplified because of the fundraising of PR goals of a small number of companies, and that has really distorted perception,' he said. He talked about the use of sandboxing for making sure systems are safe, and promoted the idea of open source technologies to deliver value to the world at large. As for valuable skills, he said, people should know how to prompt LLMs, as well as having basic coding skills. 'I think at this moment in time, we are already seeing a very clear performance gap in many jobs. (Knowing) how to code, it's absolutely huge in software engineering, but it's already very (necessary) outside software engineering. As AI becomes better, as security becomes better, I fully expect this performance gap to continue to grow.' He talked about how 10X programmers tend to make more money than others, and how the technology saves people a lot of time. He also addressed the funding challenges to universities. At the end of the day, Ng suggested, we should still know how to code, even if we don't have to do it every time we open the terminal. This made a lot of sense to me, because we're also experimenting with collaborative platforms that blend together the terminal, the collaboration hub, and the hosting system, as in the Microsoft Azure AI Foundry agent system I described a few days ago. It's all part of reinventing how we build software and what it means to be a professional in the age of AI. Let's continue to think about what the workforce will look like in a few years. When Ng talks about 'more powerful workers,' I think part of what he means is that we'll be working inside of these tools that really make us all 10X, effectively.
Forbes
16-05-2025
- Business
- Forbes
Beyond No-Code: Windsurf's SWE-1 And More
Logo of this company Winsurf It's only been a few months since developers started using the new term 'vibecoding,' but new LLM capabilities put the no-code/low-code movement into hyperdrive. Now we have brand new announcements of a frontier model by Windsurf, SWE-1, that advances beyond no-code into an overall model approach to software engineering. Let's look back a bit at how we got here, and what kind of research are going on in the world of software development. First of all, no-code refers to the use of AI models to generate the code needed for an application or given resource without humans writing that code themselves. But as experts point out, this is not the entirety of what software engineers or software developers do. There's a context to the code that could also conceivably be automated. Check out this language in an academic paper from 2024, maintained at the ACM Digital Library: 'The relevance of low-code / no-code development has grown substantially in research and practice over the years to allow nontechnical users to create applications and, therefore, democratize software development. One problem in this domain still persists: many platforms remain low-code as the underlying modeling layer still requires professionals to write/design a model, often using Domain Specific Languages (DSLs). With the rise of generative AI and Large Language Models (LLMs) and their capabilities, new possibilities emerge on how Low Code Development Platforms (LCDPs) can be improved.' What we have here is the hint or suggestion that automation systems can look beyond just the generation of code, and into the life cycle of writing or designing something. There's also some interesting coverage of this idea in a resource from the Software Engineering Institute at Carnegie Mellon that introduces the term 'model-driven engineering' (MDE). Authors write: 'In this report, we use the term MDE to refer descriptively to a software development approach that treats models as the primary artifacts created and used by software lifecycle processes. The enabling tools and technologies include a broad spectrum of capabilities that may provide value for developers, acquirers, and end users.' That broad spectrum of capabilities is what innovators are looking at in exploring how to broaden no-code into democratizing the entire software development life cycle. Now, this week, we hear that Windsurf, a company known for its code automation approach, has a new family of AI models that are looking to do this exact thing. 'Writing code is only a fraction of what engineers do,' said Varun Mohan, CEO and co-founder of Windsurf, in a press statement. 'To truly accelerate software development by 99%, we had to move beyond 'coding-capable' models and build software engineering-native models. SWE-1 is our first step in that direction, building a foundation for the future state.' In the SWE-1 suite, there's SWE-1, SWE-1-Lite, and SWE-1-Mini. In describing the utility of the tools, Windsurf Co-founder Anshul Ramachandran uses the term 'flow awareness.' (for context, see this interview I did with Ramachandran at Davos). 'Flow awareness lets us see exactly where models succeed or fail, down to the individual decision point,' Ramachandran explains. 'That feedback loop is our competitive edge.' Maybe if you're interested in this process and what SWE-1 brings to the table, you want a little more detail… Some of the background of this type of pioneering involves what programmers typically do during a project. They write code, yes, but they use a set of three important resource environments – the IDE, the terminal, and the browser. The IDE is the environment where programmers often write the code and analyze it. The terminal is where they run the code. Programmers may use browsers to test the code, or to get information on best practices from sites like Stackoverflow. In fact, many programmers, when asked about how they use AI, suggest that they're using Stackoverflow much less, because of code automation. In any event, a model that can traverse all three of these environments is going to be immensely valuable as a broad-based engineering tool. So that's a good place to start in researching where we are at with the NCLC movement. It seems like the goal is to keep pushing the ball forward in terms of what people can do without technical knowledge – how easy it can be to spin up an application or codebase with just a few prompts to an LLM. This is a space that many of us will be watching for a great deal of potential disruption.
CBC
11-05-2025
- Entertainment
- CBC
'Vibe coding' makes designing apps easier than before — but it comes with risks
Cyanide ice cream. Cholera-inspired chocolate cake. A recipe featuring so-called ingredients the CBC's Language Guide discourages from repeating verbatim. These can all be found on a website created by U.S.-based tech entrepreneur Tom Blomfield that uses artificial intelligence to generate recipes after a user suggests a list of ingredients. "Some more mischievous users started to push the envelope on what kind of recipes his AI-powered site would generate. And they found that it would generate things that they thought were funny but are potentially dangerous," Emanuel Maiberg, a reporter for the tech website 404, told The Current host Matt Galloway. Blomfield built with a new method recently dubbed "vibe coding," where people use AI tools to build a program, app or game with prompts or suggestions much like how one would use ChatGPT to generate a written text answer. But the example calls attention to how building apps solely by "vibe" may lead to problematic, and even potentially dangerous results. Since Maiberg's story was published on April 2, the cyanide ice cream article has been removed from archive, but others including the cholera cake remain. CBC reached out to Blomfield for comment, but did not receive a reply. Vibe coding was coined by Andrej Karpathy, a Canadian computer scientist and co-founder of artificial intelligence giant OpenAI, in a 2023 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," Karpathy wrote, as though describing something more akin to a meditation session than developing a computer program. Vibe coding has become possible recently, experts say, because AI tools have become sophisticated enough to build functional — or mostly functional — apps with little more than general prompts or suggestions. "The difference between last year and now is that large language models — LLMs — have gotten good enough that they can actually produce, you know, medium-scale games or apps, things like that. It actually works," said Michael Guerzhoy, an assistant professor teaching programming and machine intelligence at the University of Toronto. The process has opened doors for budding app makers like Chioma Janelle Efejedia, a psychotherapist and social worker based in Kitchener, Ont. Not knowing how to code, she might have had to pay a programmer thousands of dollars to make a mental health app. Instead, she vibe-coded her own app called OMA Life, which offers guided mindfulness in various languages including Ibo, Yoruba, and Urdu, culturally relatable relaxation sounds, and access to a directory of therapists. "I just think, you know, where tech is right now gives a great opportunity to say, OK, I can meet this need," Efejedia told CBC Radio's Manjula Selvarajah. 'Build something really cool' Tobin South, a researcher in AI security at the Massachusetts Institute of Technology, says he's excited by vibe coding's "ability to unlock your everyday person to build something really cool." "I went to a party the other day and someone wanted a really cool app to organize the party. And so I made a little app with bingo cards inside of it, and a party agenda. And I was able to just bring this into existence with the English language rather than typing any code," he said. But he also cautions that it can create "massive security risks" when they aren't written and checked by experts in the field. "If you're starting to build personal finance tools or other tools to augment your life, these things can get really tricky. You do not want your bank details leaked all over the internet because you vibe coded," he said as an example to Galloway. He likened traditional app building as something made out of Lego bricks by trained experts, brick by brick, drawing on the work and experience of previous versions and notes from their creators. Vibe coding, meanwhile, is sort of like dumping your hand blindly into a box of bricks and making something out of whatever you've clawed out. "Sometimes ... this leads to a Lego construction, a Lego house that might fall down, that's missing some essential bricks that hold it all together," he said. 'Unexpected, dangerous results' Someone vibe coding on their own won't benefit from the institutional knowledge of working in a tech corporation, either. "If you work at Google, there's already someone breathing down your neck about security and making sure everything's done the right way," South said. In other words, if you make a recipe app without vibe coding, there's almost certainly someone on your team making sure that if someone asks for a recipe with cyanide, it won't actually go ahead and make one. In late 2024, Google CEO Sundar Pichai said that 25 per cent of all new code for the company's products were made with AI, though under the supervision of human employees. Maiberg cautions that as more programmers — not just amateurs or hobbyists like Efejedia — use AI to build their code and programs, more lines of code that have never been checked by a human may creep into our collective technological backdrop. That could mean errors might never be found until the consequences rear their ugly head -- from a rude ice cream recipe, to easily hackable personal banking info to something worse we may have yet to predict. "I think my concern, and the concern of other people, is that we can get unexpected, dangerous results from having so much code written by AI in a way that we don't fully understand," Maiberg said.
