Adaptive thinking: when Claude ignores your memory
I ask Opus 4.7 to solve an obvious question. It fails. The memory instructions that used to work no longer guarantee anything.
Data Engineer writing about Power BI, Python, data pipelines, and everything I learn along the way. No fluff, just what works.
I ask Opus 4.7 to solve an obvious question. It fails. The memory instructions that used to work no longer guarantee anything.
AMD analyzed 6,852 Claude Code sessions and the data is damning. Leaked source code confirms silent Opus-to-Sonnet fallback.
Anthropic leaked its most powerful model, shipped Computer Use on Windows, and exposed 24,000 Chinese accounts distilling Claude. Here's what it all means.
Translytical GA, PBIR as default in Desktop, custom totals without DAX, and Copilot improvements. A practical breakdown of Power BI's March update.
Your Claude subscription no longer covers external harnesses like OpenClaw. Pay-per-use is now mandatory, plus free credits and what it means for the ecosystem.
Sora cost $15 million a day to run and made $2.1M total. What this reveals about the real economics of generative AI.
I'm a Data Engineer with a practical approach: building pipelines that work, dashboards that tell stories, and solutions that scale.
This blog comes from conversations I have with Claude while working. Each post is distilled knowledge, ready to apply.