Scanning @octocat
Reading public repositories on GitHub. This takes a few seconds.
Scanning @octocat
Reading public repositories on GitHub. This takes a few seconds.
@octocat on GitHub (opens in a new tab)· San Francisco· on GitHub since 2011
Not enough public code yet to read a clear skill profile.
Public repositories only · scanned 1 h ago
Six areas, each scored 0–100 from your repositories
Open an area to see the technologies and repositories behind it
UI frameworks, styling, mobile and desktop clients. Seen in 1 repository.
Share of code across analysed repositories
The last year on GitHub
Declared in a manifest, a config file or a real share of the code. The number is how many repositories use it.
No frameworks or tools detected yet.
Most recent push first, with the stack DevRadar found in each
DevRadar reads up to 100 of octocat’s own, non-fork repositories (public repositories only): dependency manifests (package.json, requirements.txt, pyproject.toml, go.mod, Cargo.toml, Gemfile, composer.json, Maven and Gradle builds), the files at the root and the size of each language.
A technology counts only when a repository declares it. Topics, names and READMEs are ignored. Recent work counts most: a repository’s weight halves for every year since its last push, and very small repositories count for less.
An area’s score is three quarters depth (how many repositories use it) and one quarter breadth (how many of its technologies you use). Scores are relative signals for conversations, not grades. Last read from GitHub on 17 Oct.
Create an account to score job posts against this radar, include private repositories, and keep it synced every night.
Servers, APIs, auth, payments and background work. Seen in 0 repositories.
Databases, ORMs, caches, streams and analysis. Seen in 0 repositories.
Containers, CI/CD, cloud and observability. Seen in 0 repositories.
Tests, end-to-end suites, linting and type checking. Seen in 0 repositories.
LLM APIs, agents, model training and vector search. Seen in 0 repositories.