HN Who Is Hiring: Structured Jobs

Every month Hacker News runs an "Ask HN: Who is hiring?" thread with a few hundred job posts, each written as free text. This Actor reads the thread and turns every post into a structured row you can filter, sort and load into a spreadsheet or database.

What you get for each job post

Field	Meaning
company	Company name as written in the post. Empty when the post never names one
workplace	REMOTE (remote only), HYBRID (some days in an office), ONSITE (on-site only), MIXED (remote offered as an alternative to an office) or UNKNOWN (not stated)
remoteAllowed	true when at least one role can be done remotely, false when the post says on-site or hybrid, empty when not stated
visaSponsorship	true when the post says it sponsors visas, false when it says it cannot or requires citizenship, empty when it does not mention visas
salary	min, max, currency, period (YEAR or HOUR) and the matching text. The lowest and highest pay stated anywhere in the post, across all roles
employmentTypes	FULL_TIME, PART_TIME, CONTRACT, INTERNSHIP when mentioned
equityMentioned	whether the post mentions equity, stock options or RSUs
techTags	technologies named in the post (Python, TypeScript, Rust, Kubernetes, PostgreSQL, LLM and about 50 more)
monthsPostedInRun	when you read several months, how many of them this company posted in
header, text, urls	the post's first line, its full text and its links
month, postedAt, postUrl, author	where and when it was posted

How accurate it is

Every field is read from free text by rules, so it was measured, not assumed. 70 posts from the July and August 2026 threads, from companies not seen while the rules were written, were labelled by hand before the Actor was run on them:

Field	Correct
Is it a job post at all	98.6%
company	100%
workplace	88.9%
remoteAllowed	100%
visaSponsorship	96.8%
salary (min, max, currency, period)	98.4%

These are the first-run numbers. The mistakes they exposed (for example "On-site 4 days/week" read as on-site instead of hybrid, and "No standard visa sponsorship" read as yes) have since been fixed. The same figures are in each run's summary.

Filters

Input	What it does
monthsBack	How many monthly threads to read, newest first (1 to 12)
keywords	Keep posts whose text contains any of these words
techs	Keep posts tagged with any of these technologies
workplace	Keep only some workplace types
remoteAllowedOnly	Keep only posts where a role can be done remotely
visaSponsorshipOnly	Keep only posts that say they sponsor visas
minAnnualSalary + salaryCurrency	Keep only posts whose stated annual pay reaches the floor in that currency
includeText	Leave out the full text for smaller rows
maxItems	Cap on rows

Run summary

Stored under the key SUMMARY in the key-value store: posts per month and how many were left out as non-job posts, the workplace mix, how many posts mention visas, how many state a salary, the median salary midpoint per currency, the most common tech tags, companies that posted in every month you read, the measured accuracy and the caveats below.

Things to know before you rely on it

• Nothing is guessed. When a post does not say something, the field is empty. Most posts do not mention visas, so visaSponsorship is usually empty.
• Salary is what the post states. Funding rounds, market sizes, revenue and 401k are not pay. Hourly rates are marked HOUR. A range written without a currency is kept with currency empty. Nothing is converted between currencies.
• Region limits stay in the text. "Remote (US only)" is REMOTE; the US limit is in header and text.
• Not every top-level comment is a job post. Comments about the thread and job seekers posting in the wrong thread are left out and counted in the summary.
• No contact list is extracted. Emails stay inside the post text exactly as the poster wrote them.

Data

The Actor reads Hacker News through its public search API, the same data anyone can browse on the site. It does not log in, does not use cookies and does not work around any blocking. A run reads one request per month plus one to find the threads, so it finishes in seconds.