AI/TLDR fits my taste. Next, let it learn my interests.

What I tried

I need a product like this. Keeping up with AI is part of building with it, but it can easily become the thing that takes me away from building. I already know a Chinese product called AI Hot, so the need feels familiar. AI/TLDR has its own taste: black backgrounds, sharp yellow accents, monospace headings, and a dense feed that still gives each release a short explanation. It feels made for people like me. My browsing experience was smooth. There is substance behind the presentation, too. Searching for Jev brought up its launch and related coverage. The launch detail puts the announcement, documentation, quickstart and evaluations together. That is a useful path from noticing something to deciding whether to investigate it.

Where I got stuck

I did not hit a frustrating interaction in this visit. The next thing I want is a closer match to my interests. Categories such as Model, Tool and Paper help narrow the feed, and search works when I know the name. But I also want to find useful things before I know what to search for. For example, I would like to describe an interest such as coding agents, local workflows, or inexpensive classification models, and get a shorter reading list around it. That is the extra layer I would find valuable. I have not tested newsletter delivery or watched the service long enough to judge its coverage over time.

One thing I'd change

I would love a saved personal filter: describe what I care about once, then let new items be checked against it. Jev from TypeSafe seems worth trying for that step. Its typed decisions fit narrow questions such as whether an item concerns coding agents, supports local use, or includes something available to try. Those separate signals could help rank the feed for each reader. The FAQ says a scan runs every two hours. I wonder what that costs to maintain; I do not know the actual bill. Jev might reduce the classification work after collection, but it would not replace crawling or writing the summaries. I would try it on a small sample first, checking both cost and useful stories it misses, and keep uncertain items reachable rather than silently hiding them. The result I want is simple: less time sorting, without losing the unexpected find.

Would I pay

Yes. I would support this product as it is. The public site is free to read, and the Buy Me a Coffee link offers a way to support it. The useful curation, smooth browsing and distinctive design already work for me. Personal filtering would make it more valuable still.

■

Roger HE is building PersonWise — The presentation presents itself.

Respond to this review →