About proomt
The engineering reading list, ranked.
There is more good engineering writing published than anyone can read, and it is scattered across company blogs, personal sites, research feeds and link aggregators. proomt reads all of it on a schedule, summarises each post, and orders the result so the first screen is worth your time.
Nothing is republished. Every post links to its original, which is the source of truth.
Where the posts come from
484 sources: company engineering blogs, individual writers, Hacker News, Lobsters, and Hugging Face’s daily papers. Each source is checked on its own schedule, between every half hour and once a day, depending on how often it publishes. A post normally appears here within a few hours of going up.
The crawler calls itself proomt-bot, honours robots.txt, and fetches any one site at most once a second. To be excluded, block that user agent or write to hello@proomt.com.
What the summary is
A language model reads the article and writes two sentences, a few takeaways, topic tags, and a quality score out of ten. Anything a machine wrote carries a small pixel mark, so you always know which words are not the author’s.
Summaries can be wrong. They exist to help you decide in five seconds whether something deserves twenty minutes, not to replace the reading.
How the ranking works
Every post gets a base score from four things: attention it has earned elsewhere, the track record of its source, the quality score, and heat borrowed from related posts.
Attention is counted in logarithms, so the fiftieth upvote moves a post far less than the fifth. Lobsters and Hugging Face points weigh more than Hacker News ones because those are smaller, more specialised rooms. Votes cast here weigh most of all. Outbound clicks count for very little, since nobody has to sign in to produce one.
Heat is the part that is ours. When a story breaks, the obvious post about it rises everywhere. proomt also lifts the quieter posts on the same subject, found by meaning rather than keywords, so you get the considered write-up next to the loud one.
Hot divides that score by age, with a half-life of about 28 hours, so the front page turns over daily. Top does not decay at all, which is how you find the best of a week, a month or all time. New is everything, newest first, with only the weakest filtered out.
No source may take a third slot on the front page until every other source has had two, so one prolific company blog cannot fill it.
Search
Search reads titles, summaries and takeaways, so a post can be found by what it is about and not only by its headline. A company name finds that company’s writing.
Your account
Voting and saving need an account; reading does not. Your votes are never shown to anyone, and your reading list is private unless you turn it on from your profile.
Feeds
Every list has an RSS feed, including the front page and each topic.