Datalayer

Follower and following counts, post totals, bio, every link in bio, verification and Threads handle for any public Instagram profile. The spend cap is checked before every request and stops the run - it does not trim results after your money is gone. Unreadable profiles are never charged for.

Apple publishes the app version on every review and nobody joins on it. This does: per-release rating distribution, so you can see exactly which update tanked your rating, plus conservative regression detection. Multi-country.

Reads company job boards at the source - Greenhouse, Lever, Ashby, SmartRecruiters - instead of scraping LinkedIn. No login, no cookies, no account to get blocked. One schema across all four, versioned so your pipeline does not break, plus a change feed of roles opened and closed.

Every weather actor returns a forecast. This answers the question buyers actually ask: is this normal for here, at this time of year? Thirty years of history per location, giving the anomaly in degrees, the rank among those years, and frost, heat and heavy-rain days against their own normals.

Trial records with the whyStopped free text classified into safety, futility, slow accrual, funding and the rest — unreadable at scale across 34,000+ terminated studies until now. Plus enrolment failure, overdue records and how many stopped trials never posted results.

The FDA gives the recall reason only as free text, so nobody can query it. This classifies it into root causes — undeclared allergen, sterility failure, nitrosamine impurity, software defect, component recall — and scores firms that keep failing the same way. 89.4% classified over 900 live recalls.

USAspending has no period-of-performance filter, and sorting by end date returns 1995 one way and 3017 the other. So you cannot ask which contracts expire soon. This computes it, flags the corrupt dates, and ranks expiring awards by value with incumbent, outlay rate and vendor concentration.

Google Trends data that actually comes back. Google requires a session cookie before serving its Trends endpoints and throttles back-to-back calls — this primes the session and paces requests, so runs succeed instead of returning empty. Interest over time, related queries, regional interest.

Grants.gov search returns a title, an agency and a date. Award size, whether you must match funds, and who may actually apply live on a second endpoint that nothing on the store joins up. This does, and its deadline radar correctly excludes forecasts, which have no closing date yet by design.

Institutional holdings with the part nobody sells: what changed. Positions opened, exited, added to and trimmed between quarters, classified on share count so a market move is never mistaken for a trade. Handles the 2023 units change and the split-row trap.

Search LinkedIn jobs logged out - no account, no cookies, nothing to get banned. Title filters and date ranges are enforced on every row, and the job/spend caps stop the run before the money is spent, not after. Full descriptions on request. Versioned schema that will not break your pipeline.

When did a competitor last raise prices, rename a plan or swap analytics? The Internet Archive knows but only reports byte changes — measured on one page, 62% of those changed nothing a person would read. This extracts prices, headings and third-party scripts and diffs those instead.

Scholar scrapers fail because Scholar blocks them. OpenAlex is open, keyless and unblockable. Three fields Scholar will not give you: the abstract rebuilt from the inverted index it ships instead of prose, field-weighted impact so citations compare across fields, and 134,038 retraction flags.

Everyone dumps /products.json. This assembles a brand profile: catalogue size, median-based price positioning, discount depth and genuine launch cadence measured from created_at — not published_at, which stores bulk-refresh and which fakes growth.

Steam reviews with the hours the reviewer had actually played. Some scrapers return that field; none compute with it. This weights sentiment by playtime and bands reviewers from drive-by to veteran, so a 12-minute opinion cannot outvote a 500-hour one.

Telegram puts a view count on every public channel post and a subscriber count in the header. Other scrapers return the messages and drop the counter, so a bought audience looks real. This returns both, plus median views as a share of subscribers, reach trend, cadence and links by domain.

Latest posts with full engagement counts from a list of X accounts, plus the part nobody else sells: what is new since your last run and which older posts are still gaining. Schedule it for a feed instead of a snapshot. The spend cap stops the run, it does not trim after.

GLEIF is the only free official source for who owns whom: direct and ultimate parents across 2.5M+ entities in 100+ jurisdictions. It also shows whether an entity still renews its LEI — 1,190,660 sit at LAPSED. A different question from whether the company is active, and the two often disagree.

Form 4 carries a transaction code saying what actually happened, and almost nothing uses it — so an RSU vest gets published as an insider buy and the automatic tax withholding as a sale. This decodes all 19 codes, isolates genuine open-market purchases, and detects cluster buying.