Tech Radar

Anthropic isn't just dropping another flagship model. They're actually running a live experiment with two parallel updates to see how Opus 5 handles user interaction. It's a weird way to roll out a release, but it tells us they're less worried about the raw benchmarks and more concerned with how the model actually behaves in the wild. Opus 5 is available today. It's designed to be proactive, and it gets surprisingly close to the intelligence of Fable 5 while costing half as much. I'm skeptical about the "half price" claim until I see the actual token drain on complex tasks, but on paper, the efficiency gain is there. The most interesting part is the effort setting. You can now manually toggle how much "thought" the model puts into a response. It's a blatant trade off. You either spend more tokens to get a smarter answer or you throttle it back to save money and get your results faster. The real question is where that line actual...

The U.S. government is treating open-weight AI models from China like digital contraband. To DC, these weights are a security risk that could leak into the wrong hands or give adversaries an edge. It's a classic geopolitical standoff, but it's happening inside our GPUs. American startup founders aren't buying it. They argue that blocking these models doesn't actually protect national security, it just handicaps the people trying to build the next generation of tools here. If you can't study the competition, you can't beat them. It's a frustrating position to be in. You're told to innovate faster while the government considers cutting off the very resources that help you understand the current state of the art. I've seen this tension before with hardware and chips, but software is different. You can't really "fence in" a model once the weights are public. We're left with a weird contradiction where the policy is to restrict ...

We've spent the last decade trying to remove every possible millisecond of friction from how we capture information. We want seamless syncing, instant search, and zero lag between a thought and a digital record. But I've started to suspect that this efficiency is actually a trap. I still use a fountain pen and paper for my most important thinking. Most people see that as a quirky habit or a waste of time, but there's a reason I haven't switched to a tablet. The physical resistance of a nib on paper isn't a bug. It's a feature. There is research suggesting that the act of handwriting recruits more of the brain than typing does. When you're forced to slow down because your hand can't keep up with your thoughts, you stop transcribing and start synthesizing. You're not just recording data. You're actually processing it. I'm curious if we've optimized the "capture" part of our workflow so much that we've accidentally br...

He was the tech industry's most relentless contrarian. For decades, he spent his time predicting the failure of tools and architectures that we now can't live without. It's a funny thing, watching someone be consistently wrong about the future while being technically right about the problems. Then there's the Giant Voice System. On paper, it sounds like a fever dream from a mid-century sci-fi novel. In practice, it's a reminder that the most efficient way to solve a problem usually involves a level of absurdity we're too scared to try. I've spent a lot of time looking at how these two things overlap. It turns out that the gap between a "failed" prediction and a working system is usually just a matter of timing and a bit of stubbornness. We should probably talk about why that happens. The Art of the Contrarian John C. Dvorak made a career out of being wrong. While other tech journalists tried to predict the future with cautious accuracy,...

Most people use LLMs as an oracle. They ask for a snippet of Python or a summary of a meeting, and they hope the model doesn't hallucinate the answer. But Timothy Gowers, a Fields Medalist, decided to treat ChatGPT as a high-speed sounding board instead. He didn't ask it for the answer to a century-old problem in algebraic geometry. He used it to probe the edges of the problem, throwing ideas at the wall to see what stuck. It's a subtle shift in workflow, but it's the only way these tools actually become useful for high-level intellectual work. If you treat an AI like a search engine, you're just waiting for it to fail. If you treat it like a collaborator that's read everything but understands nothing, you can move through a conceptual space much faster than you could alone. I've seen a lot of hype about AI "solving" science, which is mostly nonsense. AI isn't going to wake up and prove a theorem on its own. But Gowers' approach sh...

Most presentation tools are just fancy ways to lock your data in a proprietary cloud. You don't actually own your slides; you own a subscription to a viewer that might change its layout or pricing whenever the company feels like it. Bento does something different. It packs the editor, the viewer, and the live data into a single portable HTML file. It's an odd approach that feels like a throwback to the early web, but it works. I like that it's view-source honest. If you want to know how a transition is happening, you can just look at the code. The tool even checks a signed manifest at launch to handle updates, though you can toggle that off if you prefer your files to be truly static. The feature set is lean. You get the standard shapes, tables, and media embeds, but the hover states are where it gets interesting. You can set them to reveal content or focus a group, which makes it feel more like a prototype and less like a linear deck. The real question is whether...