The (mostly) technical blog of Lloyd Hilaiel.
Lloyd — Mar 18, 2026 Here’s the situation. An alien just landed on earth and he brought along this incredible new technology — a unified computational engine that blurs the boundaries between processing (general compute, graphics, kernels) and RAM. It’s this weird thing with RAM actually attached to these tiny compute units that play forward in a network. Weirdest thing you ever saw. And you just…
Lloyd — Mar 16, 2026 There’s a moment in every AI-assisted development session where you hit the wall. Not the wall of what the model can do — the wall of what the tooling can do. You’re building something, the AI is cranking, everything is flowing. Then you realize you need to change the tool the AI is talking to. The server binary. The thing underneath. In a normal workflow, this is where the…
What programming language would make an AI coding assistant most effective? It’s a question that sounds simple but reveals a fascinating tension at the heart of how large language models learn to code. The answer isn’t just “whatever has the most training data.”
Lloyd — Nov 23, 2025 The current paradigm of AI-assisted software development is often plagued by a subtle but pervasive friction: the necessity of constant, vigilant supervision. Rather than operating as an autonomous engineer, the AI frequently functions like a learner driver, requiring the human operator to sit in a state of perpetual anxiety, ready to slam on the brakes at the first sign of…
Here’s a hot take on the future of AI compute. When you hear about Google’s Project Suncatcher—their actual moonshot research into solar-powered satellite constellations with onboard TPUs—it paints a vivid picture of one possible future. In that world, the demand for computation is so vast that we literally have to go off-planet to power it. It leads to a provocative question: a decade from now,…
You might prefer to engage this post as a conversation: Your browser does not support the audio element. Over the past few weeks, I’ve been reflecting on a growing tension at the heart of the AI discourse. Mustafa Suleyman, now leading AI at Microsoft, published a piece that raised eyebrows: a caution against anthropomorphizing models, and by extension, against entertaining the idea of “model…
A couple days ago I finished a run and set the title in Strava—a moment later I was presented with a silly AI summary. Cute. Useless. And emblematic of a bigger trend: every product team is racing to bolt an AI assistant onto their surface. Meanwhile, the work I want done starts above any single app and crosses all of them.
When it comes to leveling up our human workers, we face a fundamental choice: do we give them tools to become more effective, or do we integrate those features directly into the systems they use? This isn’t just a theoretical question; it’s a practical dilemma with massive implications for the future of work, especially in critical fields like healthcare. The path we choose will determine the pace…
I’ve been tinkering with Linux since the mid-90s, starting with a Slackware DVD and the painstaking process of installing, and reinstalling, and reinstalling again. 💿 For decades, the promise of a truly custom computing environment has been a siren song for developers like me, offering total control at the cost of a steep, often frustrating, learning curve. The power to build a digital workspace…
🌴 The term “AI agent” is everywhere, but what does it really mean? In a landscape cluttered with buzzwords like “super agents” and “ambient computing,” the word has been stretched to the point of becoming almost meaningless. It’s time to cut through the noise and establish a clearer, more precise way to understand and categorize the semi-autonomous systems we’re building. Instead of getting lost…