Posts

2025 was an interesting year. I’m writing this post in March 2026, not December 2025 or January 2026 as is customary, because guess what? 2026 hasn’t been a picnic either. As with 2024, I’m still extremely grateful for the support I’ve received since starting PiForge. While the company faced its first real challenges with contracts and late payments, we managed to navigate through them and come out stronger. The new office has been a game-changer, and having a split between work and home has been great. ...

The Problem Hiring has been coming up a lot in some of my social feeds again. It’s a problem I’ve experienced, and seen often from both sides of the table. And most companies don’t get this right. Do I have all the answers? No. But I do have some thoughts. The Trigger The conversation trigger for this post was: “During the developer hiring process, has anyone found an alternative to take-home code tests and live code tests? Companies have stopped using take-home tests as they believe that it’s too easy to cheat using LLMs. About 70% of technical hiring managers that I’ve discussed this with admit that live code (and architecture) tests aren’t working, as too many skilled developers produce poor results due to the artificial time limit.” ...

Trigger warning: This post contains opinions. Specifically mine. It also contains a confession or two, which I’m not thrilled about, but here we are. I don’t have many people in tech I actively look up to. The industry is drowning in loud, obnoxious voices who are very good at having opinions and very bad at having done anything. But Carmack is one of the rare exceptions. He’s always been the quiet, thoughtful one. The hacker in the original sense; someone who understands systems deeply enough to bend them, who ships because building is the point, not the clout. I’ve written before about my own struggle with that identity vs. the need to actually make money, that tension between doing the craft and doing the business. Carmack, for most of his career, has never had to compromise on that. He gets to just build. ...

Around a year ago I started one of my blog posts with this musing on “The Promise vs. Reality of AI in Tech”: I’m concerned about the industry’s direction with these [improvements in AI]. Instead of creating better developers, we’re often replacing them. Rather than building amazing accessibility tools, we’re creating deep fakes and virtual companions. Instead of developing better MVPs for real problems faster, we’re seeing low-quality products marketed as revolutionary simply because they were built without coding experience, for no purpose other than a quick cash-grab. ...

🚨 The guy who built Clawbot/Moltbot/OpenClaw just got hired by OpenAI. Let that sink in for a moment. This project is one of the biggest security nightmares of recent memory. Every day there are more stories of problems, exploits, vulnerabilities. Yet the project has gone viral in ways that few open source projects do, and people are eating it up. Meanwhile, I know developers who started similar projects and abandoned them specifically because of the security risks. They made the responsible choice. They understood the implications. They walked away. ...

We spend more effort learning how to prompt LLMs than we do communicating with other humans. People complain that AI is bad at understanding humans, but in truth, humans are rediscovering how bad we are at explaining ourselves. Prompting works because you’re forced to do interface discovery consciously. You test phrasing, observe outputs, refine assumptions. With humans, we think discovery is rude or inefficient, so we skip it. Then we’re shocked when the call fails. ...

⚠️ Generative AI isn’t lowering the barrier to entry. It’s lowering our standards. I hate that a talentless non-creative can use an AI music generator which, to most people sounds fine, but to others is clearly terrible, and make more money than my hard-working and super talented musician friends. I hate that someone with no experience can pick up an AI code generator and rip off someone else’s hard work solving a hard problem they struggled with over many years of trial and error, and then be successful because people don’t care it’s an obvious knockoff. ...

This came up in my feeds today. I’ve said it before and I’ll keep saying it until I can’t anymore: never before has there been a time where we needed to be more intentional about training and mentoring juniors than now. As much as I love and use it, my biggest worry with current AI tooling has always been that it takes the thinking and understanding away from you. This is different from previous coding tools. Many just obfuscated or abstracted concepts, others assisted with tedious tasks, but this time the tool does it all. And it’s making us lazy, and it’s making us less capable. Studies are starting to show this. ...

The words we choose matter. People have been calling me pedantic for a long time, but I’ve never seen the problem with that. Words have particular meanings. Even synonyms are just “similar”, not usually the same. Some time back I wrote a blog post about “Why You’re NOT an Engineer (also Why It [Probably] Doesn’t Matter)” The key point I made: there IS a difference between a coder, a programmer, a developer, and an engineer (although some people disagreed). ...

I’m really happy to see my LinkedIn feed finally catching up with what I’ve been saying for the last nine months: vibe coding sucks, and fundamentals are important. Also, before the haters go and hate, “all vibe coding is coding with AI, but not all coding with AI is vibe coding”. My dislike is and always has been with “the vibe”. I’ve been reading books like Masters of Doom and Hackers: Heroes of the Computer Revolution lately, and looking back at those pioneers, they didn’t just have an intense love for what they did, they had a crazy deep fundamental knowledge of the things they worked with. And if they didn’t understand something? They took the time to figure it out. ...