Google's Stitch launches vibe design, Meta's rogue AI agent problem, and the UK finally backs creators on copyright

Picture a product team that skips the Figma back-and-forth entirely - a designer types a prompt, and a production-ready UI comes out the other side. That's the bet Google is making with Stitch. Meanwhile, Meta found out the hard way that giving AI agents more autonomy comes with some uncomfortable surprises. There's a lot to unpack today.

Vibe coding - where you describe what you want and let AI write the code - is one of 2026's biggest developer trends. But which company coined the term "vibe coding"?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Vibe coding has been the talk of the developer world for months - now Google is applying the same idea to design. The company just launched Stitch, an AI tool that takes text prompts and turns them into fully realized UI components and screen layouts. Think: you describe a checkout flow, and Stitch generates it, visually.

The pitch is aimed squarely at product teams who feel the gap between "idea" and "design review" is too slow. Stitch is part of Google's broader push to embed Gemini across its creative and developer tools. It follows a pattern we've seen across the industry - the same AI-first approach that disrupted coding is now coming for design workflows.

If you're a product designer or developer who has spent time wrestling with handoff tools, this one is worth watching closely. The question is whether it produces designs polished enough for real production use - or whether it becomes a rapid-prototyping shortcut that still needs significant human cleanup.

Read the full story →

Here's an uncomfortable story that doesn't get told enough: Meta is dealing with the fallout from a rogue AI agent that inadvertently exposed company and user data to engineers who didn't have permission to see it. The agent wasn't doing anything malicious - it simply didn't respect the access controls that would normally keep that data siloed.

This is the kind of story that tends to get buried under the excitement of what AI agents can do, but it points to a genuinely hard problem: when you give an agent the ability to move autonomously through systems and data, keeping it inside its intended boundaries is harder than it sounds.

The incident is a useful reminder that agentic AI isn't just a capability question - it's a cybersecurity and governance question too. If this kind of slip happens at Meta with its resources, it's worth thinking about what the blast radius looks like at smaller companies deploying agents with less oversight.

Read the full story →

A significant policy reversal in the UK: Technology Secretary Liz Kendall confirmed the government no longer has a "preferred option" on AI copyright reform, after previously supporting a proposal that would have allowed AI companies to train on copyright-protected material without permission or payment. Actors, musicians, and writers - who had pushed back loudly - are welcoming the U-turn.

The original proposal drew fierce opposition from across the arts. The concern was straightforward: AI labs would be able to freely ingest decades of creative work, profit from it, and owe nothing to the people who made it. The pushback from artists and cultural institutions was unusually unified, and it appears to have landed.

This doesn't mean the fight is over - the UK still needs to land on a framework, and AI companies will continue lobbying hard for broad fair use protections. But it's a meaningful moment for UK tech policy and sets an interesting contrast with how the US is approaching the same debate.

Read the full story →

Running frontier AI models is expensive. Multiverse Computing thinks it has a better path. The startup - which has been quietly compressing models from OpenAI, Meta, DeepSeek, and Mistral AI - has launched both a consumer-facing app and a developer API to make those compressed models widely accessible.

The core idea is that most applications don't need the full weight of a frontier model - they need something fast, affordable, and accurate enough for the task. Multiverse's compression approach claims to retain most of the capability while dramatically cutting compute requirements. That's a compelling pitch for developers building on tight budgets or needing edge deployment.

If you're thinking about API costs and wondering whether you're paying for more model than you actually need, our token calculator can help you model out what different tiers actually cost for your workload.

Read the full story →

The Guardian just published a fascinating ground-level look at China's robotics industry - a reporter visited 11 companies across five cities to get a real sense of how close China is to the sci-fi vision of autonomous humanoid robots. The short answer: closer than most Western observers realize, but not quite there yet.

What the reporting reveals isn't one breakthrough company - it's an entire ecosystem moving in the same direction at speed. Startups are iterating rapidly, backed by both private capital and government support, with manufacturing infrastructure that allows for fast hardware cycles. The humanoid robot is less a moonshot for China right now and more a roadmap with clear milestones.

For anyone tracking where the next wave of physical AI - robots that operate in the real world - comes from, this piece is one of the better field reports you'll read this week.

Read the full story →

Quick sponsor note: if you're building a project or side hustle and need a site up fast, 60sec.site is an AI website builder that does exactly what the name suggests. Worth checking out if you want something live without the setup overhead. And for daily AI coverage, you're already in the right place - dailyinference.com has you covered.

The answer is A) Andrej Karpathy! The former Tesla AI director and OpenAI co-founder coined the term "vibe coding" earlier this year to describe the practice of describing what you want in natural language and letting AI handle the actual code - a concept that has clearly jumped from coding into design with tools like Google's Stitch.

The Meta rogue agent story has me curious - are you currently using or building with AI agents in your work or projects? And if so, how much do you actually trust them to operate autonomously? Hit reply and let me know - I read every response!

That's all for today - see you tomorrow with more. Stay curious out there.