About AI Daylee
AI Daylee is a daily newspaper about AI for people who don't work in AI: small business owners, people trying to get more done at work, and anyone curious what these tools can do for them. It is published by AscenHD LLC in Pittsburgh.
Who runs it
Brian Kolowitz, D.Sc., edits AI Daylee and wrote the rules below. He spent more than eight years on the faculty at Carnegie Mellon, holds a doctorate in Information Systems from Robert Morris University, and has spent 20+ years building software and applied AI, much of it in healthcare. Full CV ยท LinkedIn
How stories get picked
Every morning we search Brave News for each of the six sections. Before anything else, we throw out vendor marketing blogs, links that don't load, and anything we covered in the last 30 days (the same link, or a headline that is nearly the same).
An editor model reads what's left and keeps the stories that change something for regular people or small businesses. It rejects old launches presented as news and vendor advertorials. Each edition carries at most 12 stories, and a section can come up empty on a slow day.
What the AI writes
AI writes every story from its source, in the same four parts: what happened, why it matters, who's doing it, and three steps to try it yourself. Every story links its source. Read it before you act on anything.
Under each story, the Panel comments: five fictional characters, all written by AI and labeled that way. Their comments are opinion. They don't claim to have used anything.
Stories publish automatically. No one reads each one before it goes live, which is why every rule above is enforced in code and every story links where it came from.
How Try It prompts are tested
Each Try It prompt is for a task people actually search for, found through Google's search suggestions. We send the exact prompt to ChatGPT, Claude and Gemini through their APIs, with no extra instructions, and publish every answer as it came back. The page lists which model version ran.
A separate model that isn't one of the contestants scores the answers blind, three times in shuffled order, and we average the scores. Once a prompt is published, its text and answers are never edited. The ChatGPT, Claude and Gemini apps add their own instructions, so your results can differ.
Corrections
Spot something wrong? Tell us through the AscenHD contact page.