One-person data studio · Japan
App Store charts and company job boards only show what is true today. Since August 8, 2026, I have been recording them on a daily schedule and turning the results into data products for developers, analysts and sales teams. The pipeline is built and maintained with Claude.
Track record as of October 8, 2026
Products
Each product reads an official public endpoint and adds what the source does not keep: history, a common schema across vendors, or the change since the last run.
Job postings from Greenhouse, Lever, Ashby, SmartRecruiters and Recruitee in one schema. Delta mode returns only the roles opened or closed since the last run, with first-seen dates, plus a per-company rollup of who started hiring and for what.
Top Free and Top Paid, top 100, for 10 storefronts: US, JP, GB, DE, FR, KR, CA, AU, BR and IN. The Apify actor is built and tested but not public yet. The daily history collected so far is available on request.
The same history and derived signals over HTTP, for AI agents that pay per request with x402 instead of signing up for an account and an API key.
How it works
A job scheduled for 06:00 JST reads each source's official public endpoint. Today that is 20 App Store chart feeds (10 storefronts, Top Free and Top Paid) and the Hacker News front page.
Rows are appended as NDJSON, one file per source per day, and committed to git so that every snapshot has a verifiable timestamp. A day that was not recorded cannot be recovered later.
Apify actors serve current data on demand. Derived signals, such as rank momentum and hiring deltas, are computed from the same files.
{ "ts": "2026-10-06T21:00:02.048Z", "country": "us", "chart": "top-free",
"rank": 7, "appId": "6754134922", "name": "PineDrama - Short Dramas",
"artist": "TikTok Ltd.", "releaseDate": "2025-12-23",
"url": "https://apps.apple.com/us/app/pinedrama-short-dramas/id6754134922" }Built with Claude
I run the studio alone. Three Claude Code agents handle the work that would otherwise need more people: finding sources, building actors, and repairing them when a source changes.
Searches for long-tail public data, opens every candidate URL to confirm it is reachable, checks robots.txt and terms, and looks at how many similar actors already exist on the Apify Store.
Turns a target spec into a deployable Apify actor: typed input schema, pay-per-event billing, timeouts and retries, Dockerfile and README. Both actors were built with Claude Code.
Reads run logs, separates transient errors from structural breaks such as a changed response schema, and proposes the exact parser fix, starting with the sources that matter most.
Data rules
No names, contact details or anything else that identifies a person.
Nothing behind a login, a paid license, or terms that prohibit collection.
Public JSON APIs and open data before HTML pages, with robots.txt respected.
Every snapshot is timestamped and kept append-only.
About
Sofue Data is a sole proprietorship in Japan, run by Takuma Sofue, a software engineer. I started it in August 2026 to find out how far one person can take a data business when AI agents do most of the building and upkeep.
Roadmap
Questions, dataset requests and partnerships are welcome.