Signal

I trained a model. What is next?

✨ The gist A Kaggle Grandmaster shares practical advice for what to do after building a machine learning model, drawing from his experience competing in data science competitions. The post covers next steps beyond initial model training—essentially guidance on how to move from a working prototype to something more polished or deployable.

What are Agentic Workflows?

As organizations move beyond single-prompt AI interactions, agentic workflows are...

How to test agent experience changes without shipping them?

✨ The gist Most hypotheses about how to improve AI agent performance fail when shipped to production, so testing changes before deploying them saves wasted effort. Microsoft DevBlogs discusses methods for evaluating potential improvements—like rewriting documentation or changing command-line interface output—in a controlled testing environment rather than rolling them out live and hoping they work.

How to test agent skills without hitting real APIs?

✨ The gist When an AI agent uses an external API to do something, testing that integration properly is hard—running against real APIs is expensive, slow, and risky. Microsoft is explaining how to test agent skills by simulating API responses instead of hitting actual services, so developers can verify their integrations work correctly without those downsides.

What to consider when creating waterfall charts?

Don’t add an axis break indicator, zoom into differences, add connecting lines, and more do's and don'ts for designing waterfall charts.