Blog
Notes from the build.
Flutter vs React Native: choosing a cross-platform framework
Flutter and React Native are both sound choices in 2026. How they differ in architecture, performance, UI, native access, web, hiring and on-device AI.Conversation design for LLM assistants: beyond the prompt
The design work left once the model is fluent: persona rules, openings, clarifying questions, repair, escalation wording, memory, voice and transcript review.Next.js vs Astro: choosing a web framework for your product
Next.js and Astro are built for different jobs. A practical comparison of rendering, performance, SEO, AI crawlers, hosting and team fit, with scenarios.RAG on Cloudflare: Vectorize, D1, R2 and AI Search
How to build retrieval-augmented generation on Cloudflare with R2, Queues, Vectorize, D1 and AI Search, the limits to plan around, and when to use pgvector.FinOps for AI: keeping GPU and inference bills under control
How AI spend differs from normal cloud spend, when self-hosted GPUs beat API inference, how to tag cost to features and customers, and a monthly review list.RAG vs fine-tuning vs long context: choosing the right one
RAG, fine-tuning and long context solve different problems. What each one changes, what it costs to build and run, and how to choose for your case.The EU AI Act: what product teams need to do and when
The EU AI Act for teams shipping AI features: risk tiers, provider and deployer roles, the dates that apply after the 2026 changes, and a working checklist.Fine-tuning LLMs: when it pays off and how to do it
When fine-tuning an LLM is worth it and when it is not: the methods, the data you need, the workflow, hosted and open-weight options, and the upkeep.AIoT explained: what changes when connected devices get AI
What AIoT means in practice: where intelligence sits across device, gateway and cloud, the use cases that hold up, what is hard and a realistic first project.Guardrails for AI agents: prompt injection and permissions
How to defend an AI agent in layers: prompt injection explained, least-privilege tools, approval gates, sandboxing, output checks, audit logs and red-teaming.