Opinions, Not Press Releases
Browse a curated front page, then move through focused topic and cluster hubs without losing older work to chronology.
Search every article
Use site search to search the full archive, not only this page.
No posts match your search. Try another term or pick a topic.
Latest articles
Nine articles per page, ordered by publication date.
AI & AgentsPxpipe Review: Can Images Cut Claude Code Token Costs 60%?
Pxpipe turns bulky Claude Code context into dense PNGs. The measured token savings are real, but exact IDs and numbers can fail silently. See where it works and where text must stay text.
AI & AgentsCheaper Per Token. More Expensive Per Answer.
Sonnet 5 launched cheaper per token than Opus 4.8, then cost more per completed task on the full benchmark. Why cost per task, not price per token, is the number that lands on your invoice.
AI & AgentsHow to Use Claude Fable 5 in Claude Code
A practical Claude Code workflow for Fable 5: select the model, plan first, route implementation, verify with tools and handle Opus 4.8 safety fallbacks.
AI & AgentsDario Declared War on Open Source. The Real War Is Over Your AI Bill.
Anthropic accused Chinese labs of stealing its models and asked Washington to step in. Strip the geopolitics and it is a fight over the price of intelligence. Coinbase already cut its AI bill 50% with open weights and routing. Our read, and the EU hedge.
Delivery & QAAgile De-engineering
How a movement built to liberate engineers came to erode engineering culture in the enterprise. Alexandre Kotcherguine and Kevin Riedl trace the commercialisation, ritual capture, metric inversion, and craft erosion behind it.
AI & AgentsWhat an Internal AI Assistant Actually Costs in the DACH Region (2026)
Embeddings, vector DB, tokens, hosting, and the maintenance line everyone forgets. A directional per-seat cost breakdown for a DACH internal AI assistant.
Product & MVPThe MVP Is Dead. Build a Minimum Credible Product.
AI made prototypes cheap, not product judgment. The traditional MVP's roughness now damages the signal it was meant to collect. A practical case for building less, finishing one trustworthy path, and measuring evidence that deserves the next investment.
AI & AgentsLLM Gateways Compared 2026: LiteLLM vs OpenRouter vs Portkey vs RouteLLM
One endpoint over many providers, with fallback, caching, spend limits, and routing in one place. How LiteLLM, OpenRouter, Portkey, and RouteLLM differ, and how to choose on the constraint that binds you.
AI & AgentsSelf-Hosting LLMs in the EU: When Open Weights Actually Pay Off
The GPU is the cheap part. Here is the real cost of self-hosting open weights, the tokens/day break-even vs hosted APIs, when data residency forces your hand, and the vLLM production stack.
Follow the work that matters to you
Get a short email when we publish something new. Follow the whole blog or only the problems you care about.