The AI Krakow reading pile for spring
A practical reading list for builders: model evaluation, interface patterns, local-first tools, and a few essays that make the current AI moment easier to navigate.
The best AI reading pile mixes technical depth with product judgment. Model capabilities change quickly, but the questions around usefulness, evaluation, and trust stay stubbornly relevant.
Start with evaluation. Learn how people test task success, compare outputs, and notice regressions. Without evaluation, every prompt change feels like weather.
Then look at interface patterns. Drafts, citations, structured outputs, tool permissions, and human review loops are where AI becomes legible to users.
Local-first and privacy-aware tools deserve attention too. Not every workflow should send every piece of context to a remote service, especially when prototypes turn into operations.
Finally, keep a few thoughtful essays nearby. The current AI moment is technical, but it is also cultural. Good writing helps separate durable ideas from the weekly fog.