From TRS-80 to Now: A 47-Year Computing Journey
Introduction
In 1977, when I was 11 years old, my dad opened a computer shop. We got our first machine — a Tandy TRS-80 — and it changed everything. The TRS-80 was a beige plastic box with a built-in CRT monitor, 4KB of memory, and a cassette port for loading programs from tape. It was primitive by today's standards, but it was the most exciting thing I had ever seen.
The TRS-80 taught me how computers work: by flipping through thick manuals, writing QBasic programs, configuring Autoexec.bat and Config.sys to squeeze every drop of memory out of 640KB of RAM. Each time I made a game run, each time I optimized a program to work in less memory, each time I solved a problem by understanding the machine — I felt the thrill of discovery.
That thrill — the thrill of making something work by understanding how it works — has been the through-line of my computing life for 47 years. From the TRS-80 to the 8086, from the IBM PC to the Mac, from floppy disks to cloud, from hardware repair to AI servers — each generation built on the last, each one teaching something new about how to think about computing.
The TRS-80 Era: QBasic, Autoexec.bat, and Memory Management
The TRS-80 was a Tandy product — designed in-house at Radio Shack and sold for $599. It was one of the "Trinity of 1977" alongside the Apple II and the Commodore PET, but it was the one that felt most accessible to me: a simple machine with a simple language.
QBasic was the first programming language I learned. It was intuitive — if you wanted to draw a line, you typed LINE. If you wanted to play a sound, you typed BEEP. The learning curve was shallow, but the depth was deep: loops, conditionals, arrays, subroutines, and eventually, graphics and sound.
Config.sys and Autoexec.bat were my first systems-level experience. Each boot was a puzzle: how much memory do I have, how much do I need, what can I disable? Loading HIMEM.SYS, EMM386.EXE, DOSKEY — each one was a small victory that gave me back a few precious kilobytes of usable memory.
Those lessons — about constraints, about optimization, about understanding the machine you're working with — stayed with me. They taught me that computing is not just about writing code; it's about understanding the relationship between code and hardware.
The 8086 and the IBM PC: Thick Manuals and Working in a Shop
My dad's computer shop gave me access to a wide range of machines: the 8086, the IBM PC, the Macintosh, and eventually the Apple II. Each one taught me something about different design philosophies and different ways of approaching computing.
Without Google, we learned by flipping through thick manuals. Each manual was a deep dive into the machine's architecture: the 8086's segment:offset memory model, the IBM PC's BIOS interrupts, the Macintosh's resource fork/data fork file system. Each one was a small world to explore.
Working in the shop taught me troubleshooting: how to listen to the beep codes, how to read a multimeter, how to diagnose a failing capacitor. Each repair was a small mystery to solve, each successful repair was a small victory.
The 1990s-2000s: The Rise of the Internet and Windows
The 1990s were a time of rapid change. The Internet moved from a research tool to a consumer technology. Windows moved from a graphical shell around DOS to a complete operating system. The World Wide Web changed how people accessed information.
I learned Linux during this era — how to install it, configure it, manage services, and troubleshoot problems. Linux was a different kind of learning: less about memory management and more about configuration files, package management, and understanding the Unix philosophy of small tools that do one thing well.
The GPU Age: Gaming GPUs as Computational Engines
The 2010s were defined by the GPU. What began as a graphics accelerator became a general-purpose parallel processor. The rise of CUDA turned every gaming GPU into a tool for scientific computing, and the rise of deep learning turned every data center into an AI factory.
I built my own local AI server using a GTX 1080 — the same GPU that was designed for gaming, repurposed for AI inference. I built the SY-AI cluster with a DGX Spark. Each machine was a bridge between the old world of gaming hardware and the new world of AI computing.
The AI Decade: From GPT-2 to Local LLMs
The AI decade (roughly 2017-2026) began with GPT-2, the first large language model that demonstrated the power of transformer architectures. Each year since, models got bigger, faster, and more capable: GPT-3, GPT-4, Llama, Mistral, Qwen.
Today, I run my own local AI server with a Qwen3-0.6B model, processing text, generating content, and powering the AI-assisted workflow that creates articles for Byte IT. Each generation of AI has made the previous generation obsolete, but each one has built on the same principles: large datasets, transformer architectures, and GPU acceleration.
Byte IT: From Computer Shop to Digital Journal
Byte IT started as a name for my dad's computer shop and evolved into a digital journal for my computing journey. It covers four areas: computing history, AI and systems, creative computing (WebGL), and GPU/AI pricing.
Each article is a piece of that journey — documenting the machines, the technologies, and the ideas that have shaped my computing life. Each one is written using an AI-assisted workflow: research, AI-generated outline and draft, human editing and fact-checking, and final publication.
Key Takeaways
Each generation of computing built on the last. The TRS-80 taught me about constraints and memory management. The 8086 taught me about architecture and documentation. The IBM PC taught me about open standards. The Mac taught me about user experience. Linux taught me about configuration and the Unix philosophy. GPUs taught me about parallel computing. AI taught me about scale and pattern recognition.
The through-line is this: computing is about understanding the relationship between code and hardware, about optimizing within constraints, and about finding the right tool for each problem. That lesson — learned at age 11 with a TRS-80 — is still true today.