Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here.
Half of all LLM usage is for writing computer code
The tech industry insists that AI will “transform” how companies, both large and small, operate. Tech VCs and AI founders predict that major business functions will be reshaped, one by one, to be handled by AI agents. For a while, many speculated which function would be transformed first. It wasn’t customer service, legal, or marketing: it was software development. Generative AI’s first killer app is coding. Tools like Cursor and Windsurf can now complete software projects with minimal input or oversight from human engineers.
Businesses are rushing to capitalize on the efficiency gains offered by AI coding. Naveen Rao, chief AI officer at Databricks, estimates that coding accounts for half of all large language model usage today. A 2024 GitHub survey found that over 97% of developers have used AI coding tools at work, with 30% to 40% of organizations actively encouraging their adoption. (GitHub, owned by Microsoft, created one of the first such tools, Copilot.) Microsoft CEO Satya Nadella recently said AI now writes up to 30% of the company’s code. Google CEO Sundar Pichai echoed that sentiment, noting more than 30% of new code at Google is AI-generated.
The soaring valuations of AI coding startups underscore the momentum. Anysphere’s Cursor just raised $900 million at a $9 billion valuation—up from $2.5 billion earlier this year. Meanwhile, OpenAI acquired Windsurf (formerly Codeium) for $3 billion.
And the tools are improving fast. OpenAI’s chief product officer, Kevin Weil, explained in a recent interview that just five months ago, the company’s best model ranked around one-millionth on a well-known benchmark for competitive coders—not great, but still in the top two or three percentile. Today, OpenAI’s top model, o3, ranks as the 175th best competitive coder in the world on that same test. The rapid leap in performance suggests an AI coding assistant could soon claim the number-one spot. “Forever after that point computers will be better than humans at writing code,” he said.
One reason for the progress: AI coding tools are gaining stronger reasoning abilities and can process much more information at once. While models retain general knowledge from pretraining, they depend on specific project-related input—such as a software description—provided by a human when it’s time to build something. This information is stored in short-term memory, known as a context window. Currently, state-of-the-art tools can productively consider fewer than 100,000 tokens (units representing words and word parts) at once. But that number is bound to go up.
Google DeepMind research scientist Nikolay Savinov said in a recent interview that AI coding tools will soon support 10 million-token context windows—and eventually, 100 million. With that kind of memory, an AI tool could absorb vast amounts of human instruction and even analyze an entire company’s existing codebase for guidance on how to build and optimize new systems. “I imagine that we will very soon get to superhuman coding AI systems that will be totally unrivaled, the new tool for every coder in the world,” Savinov said.
Accenture research shows AI ‘reinvention’ of business still far away
A large percentage of that first wave of AI projects, numerous industry sources have told me, ran into unforeseen problems—such as messy or incomplete data, missing infrastructure, outdated IT systems, and a lack of in-house expertise—and never made it into production. Many of the projects that did go live failed to prove they were worth the time, money, or effort. One AI company founder told me that, based on his conversations with C-level executives, he believes the success rate of first-wave AI projects was less than 10%.
The global consulting firm Accenture recently published research on what separates the winners from the rest of the pack. The firm emphasizes the importance of “thinking big”—that is, scaling AI systems aggressively across users and business functions—as well as securing executive buy-in, reskilling employees, and making significant investments in AI and cloud infrastructure. Accenture refers to companies that meet these criteria and see tangible results as “front runners.”
Yet Accenture’s data shows that such companies are still in the minority. After surveying executives at nearly 2,000 companies with more than $1 billion in revenue, the firm found that only about one-third (34%) had made a long-term investment in a generative AI system focused on a core business function. “Accenture’s research revealed that a small minority of companies . . . are already achieving considerable success at reinventing their enterprises with gen AI,” the report states. It also found that among those surveyed, 15% are ready to “reinvent” themselves with AI, 43% are “progressing,” and another 43% are “merely experimenting.”
Some companies may have been better off ignoring the early AI hype and waiting for the models, tools, and infrastructure to mature. On the other hand, there’s something to be said for learning by doing—even if the first attempt falls short.
Google is putting AI models to work to protect against online and phone scams
Online and phone scams, some of them powered by generative AI tools, surged in 2024 and continue to rise. Now, Google is deploying some of its latest AI models to help protect users from these threats. One such model is Gemini Nano, a lightweight AI that can run directly on a user’s device.
Now, when a Chrome user enables Enhanced Protection mode in Safe Browsing—the browser’s highest security setting—the Nano model runs locally to scan web content for signs of fraud. It can recognize common scam tactics, such as bad actors posing as remote technical support staff, a tactic Google says is becoming increasingly common. The model is also capable of detecting novel scams it hasn’t encountered before.
Google says it plans to use the on-device AI scam protection in the browser on mobile Android devices in the future, and to expand the detection to more types of scams. Google already uses on-device AI to detect scams in other mobile apps. The company recently began warning Android users of possible scams within text messages and phone calls.
More AI coverage from Fast Company:
- How AI is reshaping student writing
- LinkedIn’s new AI tools help job seekers find smarter career fits
- AI scam calls are getting smarter. Here’s how telecoms are fighting back
- Apple eyes AI-powered search as Safari usage declines
Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.
No comments