Generated by Google Gemini from user prompt. The Commoditization of Knowledge For decades, the software industry enjoyed a unique advantage: knowledge was scarce. Building successful software required an assembly of engineers, architects, security specialists, and domain experts. Organizations accumulated this expertise, forming a defensive moat. The companies that knew more could build more, move faster, and defend their positions. Much of the current discussion around AI focuses on productivity: faster coding, smaller teams, and cheaper products. However, the deeper economic transformation is that AI is continuously reducing the cost of acquiring expertise itself. Programming knowledge, security guidance, infrastructure design, and domain-specific insights are becoming available on demand. This does not mean expertise becomes irrelevant. Judgment, experience, and the ability to navigate ambiguity remain valuable. The difference is that access to expertise is becoming cheaper and mor...

For the past several years, the dominant narrative surrounding artificial intelligence has been one of relentless acceleration. The dramatic leap from GPT-3 to GPT-4 reinforced a widespread assumption: if we continue scaling data, compute, and model size, AI systems will continue becoming dramatically more capable. This assumption is rooted in what researchers call scaling laws—the observation that larger models trained on more data generally perform better. For a time, the results seemed almost miraculous. Yet there are increasing signs that generative AI may be entering a different phase of development. Rather than continuing along a path of exponential breakthroughs, the industry could be approaching a period characterized by diminishing returns, optimization, and commoditization. This is not necessarily the end of AI progress. It may simply mark the transition from a period of discovery to a period of maturity. The Data Challenge and the Plateau Hypothesis One reason some researche...

If you've spent any time on LinkedIn lately, you've probably seen people complaining about "AI slop." A lot of that criticism is deserved. The internet is filling up with content that feels mass-produced: generic posts, recycled ideas, articles that technically say something but don't really add anything. We've all seen it. What I've also noticed, though, is that the definition of AI slop seems to be expanding. These days, some people are willing to dismiss a piece of writing simply because they suspect AI was involved. Maybe it has a certain phrase. Maybe it's unusually polished. Maybe the structure feels familiar. Once someone spots what they think is an AI fingerprint, the content itself almost stops mattering. I think that's a mistake. We're starting to treat AI involvement and low quality as if they're the same thing, and they aren't. What Real AI Slop Looks Like Real AI slop absolutely exists. It's content generated with litt...

One of the most common demands we place on artificial intelligence is objectivity. We want AI systems that rise above politics, religion, culture, and ideology. We want them to act as neutral referees capable of seeing reality more clearly than we can. But what if the real value of AI is not its ability to eliminate bias, but its ability to make bias visible? When we ask modern AI systems complex questions, they typically respond by generating a balanced summary of opposing viewpoints. We often view this as a successful, neutral outcome. Yet something important is happening beneath the surface: the AI may be neutral, but the user is not. Human beings do not experience reality directly; we experience it through assumptions, values, and mental models. Every one of us operates inside a worldview. The problem is not that these frameworks exist, but that they often become invisible. Once a worldview becomes part of our identity, it stops feeling like a perspective and starts feeling like re...

Walk onto many elementary school playgrounds, recreation leagues, or youth sports fields across the United States, and you'll find soccer among the most popular entry points into organized athletics. Its appeal is easy to understand: the sport is relatively inexpensive, widely available, and generally viewed by many parents as a lower-risk alternative to collision sports. For years, however, youth soccer in the United States faced a major challenge. As players progressed, many encountered the high costs associated with travel teams, private coaching, and elite development programs. Those barriers often limited participation and created opportunities for athletes to migrate toward school-sponsored sports with lower direct costs to families. Today, that landscape is beginning to evolve. Organizations throughout the soccer ecosystem are investing in programs designed to broaden access and reduce financial obstacles. Major League Soccer's MLS GO Play Fund supports recreational socc...

For the past two decades, "learn to code" was the ultimate career advice. Tech executives and university advisors still beat this drum, insisting that a Computer Science (CS) degree remains the safest bet for the future. They are wrong. While a tiny fraction of elite engineers will still need formal CS training, the vast majority of students should look elsewhere. We are entering an era where studying pure computer science makes about as much sense as majoring in Latin to understand modern literature. To understand why, we have to look backward. The Desktop Revolution: History Repeating Itself When personal computers first entered offices decades ago, the early adopters were not software engineers. They were accountants, civil engineers, and business professionals. They learned BASIC and other rudimentary programming languages to solve immediate, practical problems. Their applications were ugly and unpolished, but they worked. Eventually, systems grew too large and complex. C...

Lessons From Doing It Myself I’ve been an inventor on several patents during my time at Microsoft and Clearbrief, but in those cases I had the luxury of patent attorneys handling the heavy lifting. I would describe the idea, write a technical explanation, and then the legal team would transform it into a polished application with claims, formatting, examiner correspondence — the whole package. This time was different. I had an idea I believed was patent‑worthy, but I was completely on my own. No attorneys. No corporate infrastructure. Just me, a blank document, and two AI assistants — Copilot and Gemini — to help me figure out what the USPTO actually expects. What follows is what I wish someone had told me before I started. 1. Validating the Idea (or: Asking AI if I’m Crazy) I began with a short summary and a bulleted list of the core concepts. I asked both Copilot and Gemini whether the idea seemed original and whether it had potential value. Both said yes. That was enough encouragem...