Tech Radar

OpenAI is finally trying to stop paying the Nvidia tax. For years, the company has been tethered to H100s and B200s, effectively acting as a massive revenue stream for Jensen Huang. Now, they've partnered with Broadcom to build their own silicon. It's a move Google and Amazon made years ago, and frankly, it's about time. The result is a custom inference processor they're calling Jalapeño. Greg Brockman talked through the strategy on the company's podcast, but the hardware is the real story here. It isn't a general-purpose chip. It's built specifically for the way OpenAI handles inference, which is where the actual cost of running these models lives. The chip is still in testing, but OpenAI claims the early performance numbers are significantly better than what they're seeing with off-the-shelf hardware. The real question is whether they can actually scale the manufacturing to a point where it matters, or if Jalapeño is just a high-end science pr...

I've spent the last five months watching €9,600 vanish into a void of German bureaucracy, all while being legally unable to bill a single client. I started founding my second company in late January. It is now late June. In that window, the state, two courts, a notary, a law firm, a tax firm, and several software vendors have all found a way to bill me. Every single one of them was on time. It's a strange feeling to be completely ignored by the systems that are supposed to enable your business, yet perfectly visible to the ones that want your money. The process is a mess. I thought I knew how this worked, but the reality is a slow-motion collision of outdated paperwork and digital gaps. I want to show you exactly where the friction points are and why the "standard" setup process is a lie. The Financial Entry Barrier Starting a business in this sector isn't free. You're looking at an initial spend of €9,600 before you've even processed your firs...

Most swipe typing feels like a guessing game. You slide your finger across the screen and hope the algorithm correctly interprets your imprecise movements. Most keyboards treat the layout as a secondary detail, a set of coordinates that the decoder tries to map back to letters. It's an approximation that often misses the mark. FUTO is doing something different. They're treating the keyboard layout as a first-class citizen in the decoding process. Instead of a generic approach, they've built a model that is specific to both the language and the layout. It learns the exact peculiarities of how we actually move our thumbs on a QWERTY board. Because this requires specific swipe data for training, they only have English QWERTY sorted out for now. It's a narrow start, but the accuracy is actually impressive. They've moved the logic on-device for the Android app to kill the latency, though the web demo still relies on a server. The real question is whether this layout...

Most DNS providers treat basic connectivity as a profit center. They charge you for the privilege of your traffic actually reaching your server, adding layers of cost to a service that should be a utility. It's a frustrating way to run a business, and it makes the web slower for everyone. We've decided to treat connectivity as a prerequisite instead. I've spent a lot of time looking at how the big players handle this, and the math rarely adds up for the developer. You end up paying for "premium" tiers just to get reasonable latency. We're trying something different by removing those friction points entirely. It sounds simple, but moving the needle on global resolve times without breaking the bank is a messy engineering challenge. The question is whether we can actually maintain that performance at scale without eventually sliding back into the same pricing traps. The Toll Booth Problem Charging for basic DNS is a bad move. When you put a paywall in...

The legal frameworks we built to track foreign terrorists didn't just vanish when the wars wound down. They were turned inward. We spent a decade arguing that "extraordinary measures" were necessary for national security, only to find those same tools being used on domestic protesters and political dissidents. It's a classic case of mission creep, but on a scale that actually matters. I've spent a lot of time looking at how surveillance tech migrates from the battlefield to the city street. It usually happens quietly. A piece of software designed to map insurgent networks in a conflict zone suddenly becomes a tool for local police to monitor a neighborhood. The logic is always the same: the tool works, so why not use it here? The problem is that these tools aren't designed for a society with a bill of rights. They're designed for environments where the target has no legal standing. When you port that logic into a domestic setting, you aren't just u...

Getting fired for writing a successful tool is a weird way to find out where the line is between being a "proactive engineer" and a liability. Justin Poehnelt built a Google Workspace CLI, put it out there, and watched it hit number one on Hacker News. It gained thousands of stars and users in a matter of days. Then Google fired him. Most of us are told that shipping high-impact code is the fastest way to get promoted. But there is a hidden friction when your project goes viral outside the walls of your company. Suddenly, you aren't just a developer solving a problem. You're a security risk or a policy violation waiting to happen. It makes me wonder if the "owner mindset" we're all encouraged to have actually has a ceiling. At what point does taking initiative stop being a virtue and start being a fireable offense? The tool and the intent The goal of this CLI is to stop the constant context switching between the terminal and the browser. If y...