Insights
Thoughts on AI-native development, agent workflows, design, and engineering.
Google's AI design tool generates high-fidelity UI from prompts, sketches, and voice. The interesting part isn't the generation — it's DESIGN.md and the MCP bridge to your codebase.
Read more →Paper eliminates the design-to-code translation by making the canvas itself HTML and CSS. What you design is what ships. The question is whether that's always what you want.
Read more →Pencil puts a Figma-like canvas inside your IDE with design files in Git and AI agents reading specs through MCP. The design never leaves the codebase.
Read more →Three new tools are collapsing the gap between design and code from different directions. The deeper shift isn't tooling — it's that agents need a shared design vocabulary to work at all.
Read more →Why Interlusion builds its marketing site with Astro, and what a zero-JavaScript-by-default architecture means for performance, simplicity, and AI-native development.
Read more →A practical walkthrough of implementing a three-layer design token architecture using Tailwind CSS v4's native @theme directive — no config files, no build tools, just CSS.
Read more →HSL promised human-readable colour. It lied about lightness. OKLCH delivers what HSL promised — perceptually uniform colour that works with P3 displays, design tokens, and AI agents.
Read more →Design systems fail when design and code drift apart. Design tokens fix the interface — a single source of truth that both sides can read, write, and ship from.
Read more →VectifyAI's PageIndex replaces vector search with hierarchical tree reasoning — and achieves 98.7% on FinanceBench. The result is real. The question is what it actually proves.
Read more →Everyone obsesses over models and prompts. The real bottleneck in Retrieval-Augmented Generation is retrieval — and fixing it requires engineering, not bigger context windows.
Read more →A plain-language introduction to Retrieval-Augmented Generation — what it does, how it works, and why it's becoming the default way to make AI useful with your own data.
Read more →sqlc generates type-safe Go code from plain SQL queries. No ORM, no runtime reflection, no surprises — just your schema, your queries, and generated code that compiles.
Read more →Andrej Karpathy's autoresearch lets AI agents run hundreds of ML experiments overnight on a single GPU. The implications go far beyond hyperparameter tuning — this is what the beginning of autonomous science looks like.
Read more →Both are CNCF Graduated. Both do GitOps well. But their architectures tell different stories about who the primary operator is — and that matters when agents enter the loop.
Read more →Angular spent years as the framework people loved to leave. Then it quietly shed everything that made it heavy, added everything that makes it modern, and became the frontend framework best suited for the age of AI.
Read more →Most architecture documentation is either absent or abandoned. Arc42 gives it structure that humans can navigate and AI agents can maintain. Here is how we use it and why it stuck.
Read more →Internal documentation used to rot the moment it was written. Now AI writes it, maintains it, and uses it as context — turning the oldest chore in software into a self-sustaining feedback loop.
Read more →The parrot metaphor confuses mechanism with capability. Statistical process describes how a system works, not what it can do. The debate deserves better than a four-year-old animal comparison.
Read more →Context7 is the most popular MCP server for a reason — it fixes the exact context pollution problem that makes AI agents hallucinate APIs. But a recent supply-chain vulnerability reveals a deeper question: who do you trust to fill the context window?
Read more →Prompt engineering was the warm-up act. Context engineering — designing what information reaches the model, when, and in what structure — is the discipline that actually determines whether agents drift or compound.
Read more →The object-relational impedance mismatch is real, ORM query languages add complexity instead of removing it, and AI agents write better SQL than ORM code. Here are the alternatives.
Read more →When agents write and maintain API backends, the criteria shift. Developer happiness, expressiveness, and elegance matter less. Predictability, fast feedback, and low ambiguity matter more. The answer might surprise you.
Read more →AI-native is not a marketing label. It describes a fundamentally different way to build software, run operations, and ship products. Here is what it looks like in practice.
Read more →