How the radar works
From thousands of public signals to a handful of in-depth reports.
1. Collect
We read public feeds and APIs: developer and design communities, social platforms, code repositories, AI model and paper hubs, search trends, and a large list of technology blogs and newsletters in English and Korean. We respect each site’s robots rules and terms, and we do not bypass logins or rate limits.
2. Resolve to the original
Posts that talk about the same project are grouped under one topic and linked to its original source — a repository, a model card, a paper or the official site. Duplicate copies of the same post count once.
3. Score
What matters most is breadth: how many different channels and how many different people and sites are talking about a topic. Volume inside a single community is capped, repeat mentions by the same voice on the same day count once, and posts we found by searching for the topic itself are excluded. Growth versus the previous week and public metrics from the original source (new stars, likes, downloads) are added on top. A topic seen in only one channel cannot score above 30, and two channels cap at 55. “Hot” needs four or more channels and ten or more voices.
4. Research and write
For top-scoring topics our research system reads the original source and related public material from several angles, cross-checks claims, and drafts a report that our editorial rules then review: only the original source is linked, unverifiable numbers are removed, and limitations are stated.
Trends corner
Trends show how often each topic was featured across all sources on each day. They measure attention, not quality or adoption.
Corrections
If a report gets something wrong, email contact@muckle.kr. We fix it and update the date.