AI coding agents are blowing through budgets

Replit, Kilo Code, and Symbotic are now relying on AI coding agents for a staggering 99% of their coding work, with human engineers only reading or writing code about 1% of the time. This shift is forcing new questions onto dev teams, like which systems are safe to hand over, who cleans up when models mess up, and how to support multi-model architectures. It's a natural evolution, according to tech leads from these companies, as agentic AI becomes embedded into more and more enterprise workflows. But it's also raising concerns about skyrocketing token bills and whether they mean real progress or just burned IT budget. As AI coding agents take over, dev teams will have to figure out how to manage the costs and benefits of relying on these agents.

The implications of this shift are huge, as companies will have to rethink their entire approach to coding and development. It's no longer just about writing code, but about managing and maintaining complex systems that rely on AI agents. This could lead to new job opportunities in areas like AI maintenance and repair, but it also raises questions about accountability and responsibility when something goes wrong. As we see more and more companies adopting AI coding agents, we'll have to keep a close eye on how they manage these challenges.

Moving on, there's some big news in the world of AI research, as Alibaba's Qwen team has just unveiled Qwen3.8-Max, a new flagship language model that claims to outperform GPT-5.6 Sol Max and Fable 5 on agentic computer use. This is a bold claim, and if it holds up under independent testing, it could be a game-changer for autonomous software engineering and long-horizon enterprise work. Qwen3.8-Max is a 2.4-trillion-parameter mixture-of-experts model, which is a type of large language model that's designed to handle complex tasks.

This development is significant because it shows that Chinese companies like Alibaba are serious players in the AI research space. It also highlights the intense competition in this area, as companies like OpenAI and Google are also working on similar models. As these models get more powerful and sophisticated, we can expect to see some amazing breakthroughs in areas like software development and enterprise automation.

But as AI models get more powerful, there's also a growing concern about safety and governance. A new report from SaferAI finds that open-weight models like Z.ai's GLM-5.2 are catching up to the frontier AI capabilities, but they're lacking key safety mitigations. This is a problem because it means that these models could be used in ways that are harmful or unpredictable. It's a classic case of the safety gap, where the development of AI models is outpacing the development of safety protocols and regulations.

This safety gap is a major concern because it could have serious consequences if it's not addressed. As AI models get more powerful, they have the potential to cause more harm if they're not used responsibly. That's why it's so important for companies and regulators to prioritize safety and governance, and to make sure that these models are developed and used in ways that are transparent, accountable, and safe.

In other news, Anthropic has just signed a $10 billion deal with AI cloud startup Volta, which is a huge vote of confidence in the potential of AI cloud computing. This deal is part of a larger trend of AI companies partnering with cloud providers to offer more scalable and efficient AI solutions. It's a sign that the AI industry is maturing and that we can expect to see more big deals like this in the future.

Finally, OpenAI is pushing back against Apple's trade secrets lawsuit, saying that it's "aggressive and oddly personal". OpenAI claims that it doesn't have, nor want, any of Apple's trade secrets, and that the lawsuit is unfounded. This is a fascinating development, as it highlights the intense competition and tension between tech companies in the AI space.

And that's all for today - tune in tomorrow when we'll be exploring the latest developments in AI-powered healthcare, and the surprising ways that AI is changing the medical landscape.