Why I keep a daily AI log
Where the AI radar came from, why every item links a primary source, and what to take from it.
By Ali Asif,
Every working day I read what the AI labs and the open-source world shipped. This site now publishes that as the AI radar: a short post with the biggest launch first, a line on why each item matters, and my own take at the end.
Where the idea came from
I got the format from the AI Search YouTube channel. Its roundups lead with the biggest release and keep moving. I wanted that in writing, with links, so I could check my own thinking later and so anyone else could check it too.
The one rule
Every item links the primary artifact: the model card, the repo, the lab's own post, the paper page. Never a tweet. Never a news site summarising a press release.
That sounds fussy until you try it. Half the "breakthroughs" in a feed turn out to be a benchmark run on a narrower task than the headline says, or a model that's gated behind a waitlist, or a license that quietly bans what you wanted to do. Opening the source takes two minutes and removes most of the noise.
How it runs
A script pulls lab blog feeds, trending models and daily papers from Hugging Face, GitHub releases, a few subreddits, Hacker News items above 80 points and a handful of YouTube channels. It drops anything I've already seen. Claude drafts a post only if at least three items clear a bar of real substance, and every link has to return a 200. A validator scans the copy. Then it publishes and pings me on Telegram. Some days that means three posts. Some days it means nothing, and that's correct.
I'll be straight about the archive. The weekly roundups running back to September 2025 are catch-up pieces. I wrote them this month, they're labelled as backfilled, and each one carries its real publish date. I don't want anyone reading them as a diary I kept at the time.
Why bother
I use the open-weight and tooling releases to make build-versus-buy calls for my companies. When a model that fits on one machine gets good enough at reading Arabic documents, that changes what I'd build for a client. I'd miss that if I only read summaries.
What to take from it
Two things, I hope.
First, I'm in this field every day, not once a quarter when a deal comes up, and I check claims against their source before I repeat them. That's the same habit I carry from audit. Second, I'm willing to put my judgment in public with a date on it, so you can see which of my calls aged well.
Read a few posts. Pick the one you'd disagree with and tell me why. That's the best way I know to start a conversation.