Stellar Tech Labs

  When shopping for a new laptop or desktop, one of the first specifications people notice is the number of CPU cores. Six cores. Eight cores. Twelve cores. Sixteen cores. It's easy to assume that more cores automatically mean a faster computer. I used to think the same thing. But the more I learned about how modern software works, the more I realized that many everyday tasks don't come close to using every core available. In fact, it's surprisingly common to see one or two cores working hard while the rest of the processor sits mostly idle. That raised an interesting question. If modern processors have so many cores, why doesn't every application feel dramatically faster? The answer lies in how software is designed. The Observation Not every task can be divided into smaller pieces. Some jobs can be split across multiple processor cores with very little effort. Others have to be completed one step at a time because every new calculation depends on the previous one. Addi...

A while ago, I upgraded my internet connection expecting everything on the web to feel noticeably faster. Speed tests looked impressive, downloads finished quickly, and streaming became almost effortless. But something didn't add up. Some websites still felt sluggish. I'd click a link and wait several seconds before the page became responsive. Sometimes the layout appeared immediately, but images, buttons, and menus continued loading long afterward. Other times, the page looked complete, yet I couldn't interact with it because the browser was still busy doing something in the background. That made me wonder whether internet speed was really the problem. The more I learned about how modern websites work, the more I realized that downloading data is only one small part of the story. The Difference Between Bandwidth and Latency When people talk about "fast internet," they're usually talking about bandwidth. Bandwidth determines how much data your connection can t...

Over the last few weeks, I've spent a lot of time learning about local AI and what it takes to run large language models directly on consumer hardware. Like many developers, I was excited by the idea of keeping everything offline. Running models locally means better privacy, no recurring API costs, and complete control over your own data. It feels like a step toward a more independent way of building software. But while experimenting with local AI, I found myself paying less attention to the software and more attention to my laptop. The fans were running at full speed. The keyboard was getting noticeably warmer. The chassis became almost too hot to touch. That made me wonder whether the real challenge with local AI wasn't software at all it was hardware. The Observation Most of the applications we use every day don't keep our computers under constant pressure. Opening a browser, compiling a project, editing a document, or responding to messages usually creates short bursts ...

  2026.06.10 :: 7 min read {blog} :: #ai #physics #chemistry #local-llm #hardware #engineering Over the last year, I've noticed a growing shift in the developer community. More engineers are moving AI workloads off the centralized cloud and back onto local machines. Open-weight models are becoming easier to run, laptops are shipping with dedicated NPUs, and tools like Ollama have made local inference surprisingly accessible. At first glance, this feels like the definitive future: Better data privacy. Lower long-term cloud costs. Zero dependency on external third-party APIs. Complete digital sovereignty. But while reading about local AI and experimenting with autonomous workflows on my own hardware, I kept running into a glaring issue. If local AI becomes the default, where does all that sustained energy come from? The deeper I looked, the more I realized that the future of local AI may depend less on software engineering and far more on physical chemistry. The Compute Revolution Is...

At some point, almost every developer looks at a gaming laptop and thinks they've found the perfect machine. On paper, the specs seem unbeatable. You get a powerful processor, plenty of RAM, fast storage, and a dedicated graphics card all packed into a portable device. It sounds like the ultimate setup: one machine for coding during the day and gaming at night. But hardware specifications only tell part of the story. The reality is that what makes a great gaming laptop doesn't always make a great development laptop. 1. Raw Performance Isn't Everything Gaming laptops are built to deliver high performance, and many of them do an excellent job. The challenge is that gaming workloads and development workloads are often very different. A game might push the hardware hard for a limited period, while software development can involve hours of compiling code, running virtual machines, testing containers, and working with multiple applications at once. Under sustained workloads, heat...

Modern technology has improved at an incredible pace. Today's laptops are faster than ever. Storage drives can move data in seconds, processors can handle demanding workloads with ease, and even mid-range devices offer performance that would have seemed impossible a decade ago. Yet despite all this progress, one problem refuses to disappear: Battery life. No matter how powerful or expensive your laptop is, there is a good chance you'll still be looking for a charger before the day is over. So why has battery technology struggled to keep up with the rest of the industry? The answer has less to do with poor engineering and more to do with the limits of physics. 1. Battery Technology Doesn't Improve as Fast as Processors When people think about technology, they often assume everything improves at the same rate. That isn't how it works. Processors, memory, and storage benefit from decades of advances in manufacturing and design. Batteries, however, are limited by chemistry....