Latest Posts

Claude Memory Optimisation: How We Cut AI Costs by 70%

We started with one AI worker drowning in 16,400 tokens of context it didn't need. Here's how we built a Memory Matrix that gives each worker only what it needs — and cut our Claude API bill by 70%.

Building an AI-Powered Static Blog Engine With Jinja2 Templates

A behind-the-scenes look at how we built a full-featured static blog publishing system — complete with Jinja2 templates, RSS feeds, XML sitemaps, and a human-in-the-loop approval workflow — all orchestrated by an AI assistant.