About
About Repo Radar
I built this for myself. A dozen open-source projects, opened and checked by hand, every day. Then I decided to stop keeping it in a private file.
Who writes this, and why
I am a product person — a product designer and product manager. Over eight years I built several startups from nothing: deciding what the product should do, writing the specifications, and directing a team of five developers until the logic was right. I do that same work today on my own, with AI where the developers used to be.
That is why I check these projects myself: I am choosing what to build on, and one that looks alive but stopped moving two years ago is an expensive mistake to make at speed.
So every morning I open a dozen of them, read what they do, and write down what I found, because I need the notes myself. This site is that file, made public — nothing here is written for an audience I would not write for myself.
Since 21 Aug 2026. Every edition in English and in Russian.
These figures are recounted every time an edition is published — last on 9 Sep 2026. The words above them were written on 9 Sep 2026.
Most “awesome” lists are a pile of links. Nobody opened them. Nobody checked whether the project is still alive. You click one, and the last release was three years ago.
This is the opposite of that. Every project here was opened. The last release date was read. The licence was read. If a project has stopped moving, it says so.
It takes about an hour a day and produces one edition of a dozen projects. Here is what that hour goes on.
How one edition is made
- 1
Find candidates
Search for projects that fit the day’s theme, plus anything already waiting in the list from earlier days.
Roughly twenty candidates for twelve places
- 2
Open every one
Not the search result — the actual project page. Read what it does, who maintains it, and what it costs to run.
Every candidate, no exceptions
- 3
Check the facts
Three things read directly, never guessed: the date of the newest release, the licence, and the star count. A project with no release in eighteen months is treated as stopped.
Release date · Licence · Stars
- 4
Keep or set aside
Some projects fail the check. Some pass and still do not fit. Both get written down with the reason, and both get published.
69 written down so far
- 5
Write it up
Every project gets the same sections in the same order. Plain English first: what it is, why it matters, three good things, three honest drawbacks, and the exact commands to install it.
About 6,000 words a day
- 6
Write it again, in Russian
The whole edition, not a summary. Written from the meaning, not word by word.
Two full editions, every day
The part most lists leave out
Since the fourth edition, every edition ends with the projects that were opened and not published, each with the reason. There are 69 of them.
They fall into groups, and the groups are not the same thing:
Could not verify
No release history to read, a page that returns an error, or two sources that disagree. This says nothing about the project. It says the facts could not be proven.
Did not qualify
Checked, and it failed. Nothing new in over eighteen months, or a licence that is not really open source.
Waiting
Checked, passed, and queued for a future edition. Nothing wrong with these at all.
A list that only ever shows wins is not more trustworthy. It is less.
The rules the check follows
Read the newest release, not the list of releases
A project’s release list can be years out of date while the project ships every week. The single newest release is read instead. This rule exists because three good projects were nearly thrown out on one bad page.
For a Python tool, believe the package index
The published package is usually ahead of the project’s own release page. Once in 15 editions it was behind — and that got written down too.
A date with no year is not a date
Project pages often print “9 June” with no year. If the date is in the past, it belongs to this year. If it is in the future, it belongs to last year.
Eighteen months of silence means stopped
No new release in eighteen months and the project is treated as stopped, however popular it is.
Read the licence, do not assume it
15 projects here carry a licence that is not plain open source — source-available, or free only up to a certain company size, or open at the root with parts carved out. They look open. They are not, and you only find that by reading.
What 15 editions look like
When something is wrong, it stays wrong in public
3 figures published here turned out to be wrong. They were not quietly edited. The correction was printed in the next edition, and the original stays on the corrections page next to it, so you can see what changed and when.
Every entry also carries the date its facts were checked. A number here is true as of that day, and not a day later.
Who this is for
Honestly? Me first. I am the person who has to decide, this morning, whether a promising tool is worth building on — and who has been burned before by one that looked healthy and had been abandoned for two years.
So the projects here are chosen for that kind of work: things one person can run, alone, on a small budget, without a team and without a sales call. If that is your situation too, this was written for you by accident, and you are very welcome to it.
It is free. There is no tracking. The whole archive stays open whether you subscribe or not.
What I get out of it
15 mornings of this have changed how I work more than I expected. I have found tools I now use every day. I have found problems I did not know were solved. And I have caught myself twice about to build something that already existed and was better than my version.
That is the real reason it is daily. Curiosity does not work in batches.