Hacker News Who is Hiring Scraper & HN Threads to Markdown

Turn the monthly "Ask HN: Who is hiring?" thread into structured job records (company, role, location, remote region, seniority, salary converted to yearly USD, visa, links) and export any Hacker News story with its full comment thread as clean Markdown. Everything comes from the official Firebase HN API and the Algolia HN Search API: no HTML scraping, no proxy, no blocking. Turn on watch mode (stateKey) and a daily schedule returns only the job posts added since the last run, and you pay only for those.

Input	A mode (who-is-hiring, front, new, best, ask, show, jobs, search, story), optional keyword / date window, story IDs or URLs. Job filters: keywords, location, remote only, posted within N days, seniority, employment type, minimum yearly pay
Output	Jobs: company, role, location, remote / onsite / hybrid, remote region, seniority, employment types, years of experience, salary (raw, yearly, USD, one per role in multi-role posts), visa, YC batch, URLs, full post as Markdown
	Stories: title, URL, points, author, comment count, comments with depth / parent_id, and the whole thread as one Markdown document
Price	$0.004 per job post, $0.003 per story, $0.01 per run. Seen, filtered and non-job comments are free
Speed	A full monthly hiring thread (about 250 posts) in about 2 seconds: one API request for the whole thread. 20 front-page stories with comments in about 4 seconds

{
    "type": "job",
    "id": 49925063,
    "hn_url": "https://news.ycombinator.com/item?id=49925063",
    "thread_month": "2026-10",
    "headline": "Discord | Senior Software Engineer, Application Security | REMOTE (US West Coast) or San Francisco, CA | $196K-$245K base + equity",
    "company": "Discord",
    "role": "Senior Software Engineer, Application Security",
    "location": "REMOTE (US West Coast) or San Francisco, CA",
    "remote": true,
    "remote_detail": "US West Coast",
    "remote_region": "US West Coast",
    "remote_timezone": null,
    "seniority": "senior",
    "salary": {
        "min": 196000,
        "max": 245000,
        "currency": "USD",
        "interval": "year",
        "annual_usd_min": 196000,
        "annual_usd_max": 245000
    },
    "salary_text": "$196K-$245K",
    "employment_type": null,
    "employment_types": [],
    "visa": null,
    "urls": ["https://discord.com/jobs/8656854002"],
    "markdown": "Discord | Senior Software Engineer, Application Security | ...\n\nDiscord is where communities and friend groups spend time together...",
    "created_at": "2026-10-01T18:09:56.000Z"
}

Jump to the task you have:

1. Get every job from this month's HN "Who is hiring?" thread as a table
2. Get a daily feed of new HN job posts (job alerts)
3. Find senior, remote or well-paid HN jobs only
4. Export Hacker News threads as Markdown for an LLM, RAG or notes
5. Monitor what Hacker News says about a product or topic
6. What this Actor cannot do

1. Get every job from HN "Who is hiring?" as a table

{ "mode": "who-is-hiring", "maxItems": 500 }

The Actor finds the newest "Ask HN: Who is hiring?" thread (posted by the whoishiring account on the first weekday of the month), reads the whole comment tree in one request and turns each top-level post into one record. Download the dataset as CSV or Excel and you have the month's HN job market in one sheet.

• Older months: sinceDays: 365 processes every monthly thread of the last year (maxItems caps the total). A specific month: paste its thread URL into storyIds.
• What gets parsed: HN posts follow the loose convention Company | Role | Location | REMOTE | Full-time | $salary | url. The parser also reads em-dash headlines (Acme (Remote US) — Senior Engineer — TypeScript / Postgres), finds the role wherever it sits ("Multiple roles", "Forward Deployed Engineers"), keeps tech-stack segments and flags such as ONSITE in tags, and never loses text: headline and markdown are always there.
• Not job posts are skipped, not billed: complaints about the thread, job seekers posting in the wrong thread ("Location: … Willing to relocate: …") and replies to a role. In the October 2026 thread that was 6 of 247 top-level comments.

2. Get a daily feed of new HN job posts

{ "mode": "who-is-hiring", "maxItems": 500, "stateKey": "hn-hiring-watch" }

Schedule this once a day (Apify Schedules, or call it from Make, Zapier, n8n or the API). The first run exports the current thread. Every later run returns only posts that no earlier run with the same stateKey exported. When a new month's thread appears, the run also re-checks the previous thread the watch was already reading, so late posts are not lost; a thread the watch never read is not back-filled (a new watch does not buy last month's posts).

What to expect: posting is front-loaded. In October 2026, 100 posts arrived on day 1 and 77 on day 2 (73% in two days), then 3 to 25 a day. So a daily schedule costs about $0.4 and $0.3 on the first two days of the month at the new price, and a few cents a day after that. Send the dataset to Slack, email or a Google Sheet with Apify Integrations.

3. Find senior, remote or well-paid HN jobs only

{
    "mode": "who-is-hiring",
    "maxItems": 500,
    "seniority": ["senior", "staff", "principal", "lead"],
    "minAnnualSalary": 150000,
    "stateKey": "hn-hiring-senior-150k"
}

Contract or part-time work only (freelancers, side gigs):

{
    "mode": "who-is-hiring",
    "employmentTypes": ["contract", "part-time"],
    "remoteOnly": true
}

Remote Rust or TypeScript jobs posted this week (all filters are AND-ed; within a list, any word matches):

{
    "mode": "who-is-hiring",
    "keywords": ["rust", "typescript"],
    "remoteOnly": true,
    "locationContains": ["europe", "worldwide"],
    "postedWithinDays": 7
}

• keywords searches the company, the role and the whole post (case-insensitive substring: rust also matches "trust").

• locationContains searches the parsed location, the remote region and the first line of the post, because HN posts often list several cities there.

• remoteOnly keeps posts that say REMOTE, including Remote (US) and Remote / Hybrid.

• postedWithinDays uses the time the post was added. In the first N days of a month it also reads last month's thread, so a weekly run never misses the posts from the end of the previous month.

• seniority is one of intern, junior, mid, senior, staff, principal, lead, manager, director, vp, executive, read from the role only (rank words such as Director / VP / Chief win over level words; "Associate Director" is a director). none matches posts without a level word.

• employmentTypes is any of full-time, part-time, contract, temporary (fixed-term), internship, read from the headline ("| Full-time |", "| FT |", "| Full Time or Contract |"), else the role ("Data Engineer (contract)"), else an explicit sentence ("This is a part-time position"). A post offering several types (employment_types: ["full-time", "contract"]) matches when any is wanted. Add unknown to keep posts that state no type: in the October 2026 thread 180 of 259 posts state one (full-time 168, contract 14, part-time 4, internship 1; 6 posts offer two or three).

• minAnnualSalary compares the top of the published range, converted to a year (hourly x 2080, monthly x 12) and, with salaryCurrency: "USD", to US dollars at the day's ECB rate. A post that lists several roles matches when any of its roles reaches the floor (see salaries). Posts without a published salary never match.

• remote_region says where a remote job can be worked from, read from the remote qualifier only (REMOTE (US) or NYC -> United States, Remote, USA only -> United States, EU -> Europe, Berlin / EU-remote -> Europe, (UNITED STATES ONLY) -> United States, a bare Remote -> Worldwide). Since 0.13.9 it holds place names only, in one spelling (US/Can -> United States, Canada, NYC -> New York, Chicago, IL -> Chicago, UT or ID residents only -> United States); a qualifier that names no place (Remote (see below), all remote, Remote (async)) counts as a bare Remote. Offices named next to it are in location, not here. Filter it in your sheet.

• remote_timezone (0.13.8) is the working hours a remote post asks for, as written (REMOTE (US Timezones) -> US Timezones, Remote (CET ±2h) -> CET ±2h, Remote (UTC-1 to UTC+3 only) -> UTC-1 to UTC+3), or null. Hours never leak into remote_region any more; hours that name a region give that region (US time zones -> United States, AMER or EMEA time zones -> Americas, EMEA), zone names alone (CET, PT/ET, UTC offsets) leave Worldwide, so filter on both columns if you need, say, European hours.

• These filters have the same names and meaning as in the ATS Jobs Feed, so one filter set works on both.

• Filtered-out posts are not stored, billed or remembered, so loosening the filter later picks them up. Use one stateKey per filter set.

The same fields with the same meaning come from Vellum Labs' ATS Jobs Feed (Greenhouse, Lever, Ashby, Workable, SmartRecruiters, Recruitee, Personio). Run both into one sheet to see HN posts next to the companies' own career pages; ATS Job Board Finder turns the company names from HN into board references for that feed.

4. Export Hacker News threads as Markdown for an LLM, RAG or notes

{
    "mode": "story",
    "storyIds": ["https://news.ycombinator.com/item?id=49797323"],
    "maxCommentDepth": 5,
    "maxCommentsPerStory": 500
}

Each story record has the story text and comments converted from HN's markup to Markdown, comments as a flat list with depth and parent_id, and markdown: the whole thread as one document with replies nested as blockquotes, ready to paste into Claude or ChatGPT, embed for RAG, or drop into Obsidian or Notion. Comment trees come from Algolia's items/<id> endpoint, which returns the full tree in one request, so a front page with comments is about 40 API requests, not thousands.

Other lists: front, new, best, ask (Ask HN), show (Show HN), jobs (YC job ads). maxCommentDepth: 0 keeps top-level comments only; includeComments: false exports the stories alone. AI agents can call the Actor through the Apify MCP server and read the dataset.

5. Monitor what Hacker News says about a product or topic

{ "mode": "search", "query": "postgres", "searchSort": "date", "sinceDays": 7, "stateKey": "hn-watch-postgres" }

search uses Algolia's syntax ("quoted phrase", -excluded words). With searchSort: date and a stateKey, a daily schedule returns only stories it has not exported before and stops paging as soon as it reaches already-seen stories, so you pay for new mentions only and need no deduplication downstream. Watch several topics with one stateKey each.

What this Actor cannot do

Read this before your first run. Coverage numbers are from the October 2026 thread (241 job posts after skipping 6 non-job comments).

Limit	What happens	What to do
Free-form posts	Job fields are heuristics over free text. Found: company 241, location 226, role 178 (74%), remote region 127, seniority 82 (34%), parsed salary 88 of 259 (34%, 0.6). A post that is one paragraph gets company only. company is null rather than a guess when the headline starts with the role ("Senior Python Backend Engineer	REMOTE") and the opening of the post names no company ("join Reef Technologies", "Kaveronix is looking for…" are read; 0.12: 26 of 723 posts in Aug–Oct 2026 fixed, 13 of them now null instead of a role or a sentence)
Several roles in one post	One record per post, not per role. role is the role segment as written ("Multiple roles"); salaries lists every pay figure with the words before it (context, usually the role: "Staff Software Engineer - Infrastructure"). context is a hint, not a parsed title, and can carry location words. When the pay line names no role, the nearest role heading above it is used ("Engineering Manager, Data Platform -- ..." then "COMPENSATION: $180,000--$250,000"); it stays null when two role headings share one figure. The same role's pay in another currency on the same line ("45-70 USD or 180-280 PLN per hour", "$185K–$220K USD / $155K–$185K CAD") keeps the first figure's context and has alternative_to (index of that figure), so it is not counted as another role 23 of 259 October posts had 2-6 figures (129 figures in all)	Unwind salaries in your sheet for one row per role
Remote region: what still reads Worldwide	remote_region is one value per post. Since 0.13.7 it also reads regions glued to the word ("EU-remote", "(US/Remote)", "Flexible/U.S. Remote"), regions per role ("Lead (UK/EU) & SDR (Canada) | Remote" -> United Kingdom, Europe, Canada) and the first region of a long note ("REMOTE (in USA, also in Europe but only for …)" -> United States). Still Worldwide: a bare Remote, an explicit worldwide / global / anywhere, hours that name no region ("REMOTE (CET)", "Remote PT/ET Hours", "UTC-1 to UTC+3": see remote_timezone), a link to a country list ("REMOTE in select countries (listed here …)") and sentences in the headline ("all roles fully remote, but some US-only as noted"). Since 0.13.9 also Worldwide: a remote qualifier that names no known place ("REMOTE (multiple countries)", "Remote (see below)"), and places outside the built-in list of about 60 countries / regions and 90 cities are dropped ("SEA" in "NYC / SEA" is ambiguous and dropped). Check headline before relying on Worldwide.
Salary not published or ambiguous	salary is null (about 66% of posts). Ranges without a currency (145k-165k), signing bonuses and bare amounts that could be hourly or thousands ($90–$110 without "/hr") stay in salary_text only, and only when no real figure is parsed from the post. Company figures (raised $42M, $3.5B in GMV, over £50 billion) and small fees ($10) are dropped, so salary_text is null rather than misleading	Leave minAnnualSalary at 0 to keep them
"Up to $X", "$X+", one figure	Read as written: salary.bound is up_to (min null), from (max null), exact (min = max) or range. minAnnualSalary compares max, or min when there is no max	Filter on salary.bound to keep only ranges
Seniority only in the body	A role "Software Engineer" whose text says "senior" stays null	Filter markdown instead
Edited posts	Watch mode tracks post IDs, not edits: a post edited after it was exported is not exported again	Run without stateKey for a full refresh
Non-job comments	Skipped when they have no segmented headline and no way to apply (no link, e-mail, "apply", "hiring", "careers"). A bare one-line title with no contact is skipped too (1 in September 2026)	—
Deleted / flagged posts	Not returned (about 15% of a thread's top-level IDs are deleted or dead). Algolia can lag the live site by a few minutes	—
List sizes	Firebase lists return at most 500 (front, new, best) or 200 (ask, show, jobs) IDs	Use search with sinceDays (Algolia pages up to 1,000 hits per query)
Big threads	maxCommentsPerStory is depth-first, so a deep cap with a small limit keeps the first few branches only	Raise maxCommentsPerStory, or maxCommentDepth: 0 for top-level only
Rate limits	Algolia allows 10,000 requests per hour per IP. Concurrency is capped (default 5, max 10), with back-off on 429 / 5xx and Retry-After honoured	—
Watch mode in front / best	Returns stories new to you, not newly submitted ones	Use new or search sorted by date
Shared stateKey	Two schedules with one key "use up" items for each other. The store keeps the newest 50,000 IDs	One key per schedule / filter set; delete the store to reset
USD conversion unavailable	When the ECB rate service (frankfurter.dev, one request per run) is unreachable, annual_usd_* are null and a USD minAnnualSalary only matches USD posts	—
Budget cap	With Max total charge per run or maxItems, the run stops cleanly before exceeding it. In watch mode, posts not exported come in the next run	Raise the cap
Run aborted or migrated	Since 0.4 the watch state is saved on abort and migration too, so items already exported are not exported (or charged) again by the next run	—
Link posts and HN's own pages	text_markdown is empty for link posts; the linked article is not fetched. news.ycombinator.com HTML is never scraped	Fetch the article URLs with a page-to-Markdown Actor

Input

Field	Type	Default	Notes
mode	string	front	front, new, best, ask, show, jobs, search, who-is-hiring, story.
query	string	""	Search terms for search mode (Algolia syntax).
searchSort	string	date	date (newest first) or relevance.
storyIds	string[]	[]	HN item IDs or item URLs for story and who-is-hiring modes.
maxItems	integer	100	Cap on stories (or job posts in who-is-hiring). Caps cost.
sinceDays	integer	0	Only items newer than N days (0 = no filter). In who-is-hiring mode, all hiring threads in the window.
includeComments	boolean	true	Fetch comment trees and render thread Markdown.
maxCommentDepth	integer	3	0 = top-level comments only.
maxCommentsPerStory	integer	200	Depth-first cap per story.
maxConcurrency	integer	5	Parallel API requests (max 10).
stateKey	string	""	Watch mode. Name of a key-value store that remembers exported IDs; runs return only unseen items.
keywords	string[]	[]	who-is-hiring only. Keep posts whose company, role or text contains one of these words.
locationContains	string[]	[]	who-is-hiring only. Keep posts whose location, remote region or first line contains one of these.
remoteOnly	boolean	false	who-is-hiring only. Keep posts that say REMOTE.
postedWithinDays	integer	0	who-is-hiring only. Keep posts added in the last N days (also reads last month's thread when needed).
seniority	string[]	[]	who-is-hiring only. Keep posts whose role maps to these levels (intern ... executive, none).
employmentTypes	string[]	[]	who-is-hiring only. Keep posts offering one of these types (full-time ... internship, unknown).
minAnnualSalary	integer	0	who-is-hiring only. Keep posts whose disclosed pay reaches this yearly amount; undisclosed pay is dropped.
salaryCurrency	string	USD	Currency of minAnnualSalary. USD converts other currencies with the day's ECB rate.

Output fields

One dataset item per job post or story, downloadable as JSON, CSV, Excel or HTML.

• Job records (type: "job"): besides the example above, story_id, story_title, author, onsite, hybrid, remote_region, years_experience_min (first "N+ years of experience" in the body), employment_types (every type the post states; employment_type keeps the first headline type for 0.9 integrations), yc_batch (W24 ...), tags (unclassified headline segments: work-mode flags, tech stack, "Visa sponsorship available"), fetched_at. salary also carries bound (range / exact / up_to / from), summary, source, annual_min, annual_max. salaries is every pay figure in the post in order (same shape as salary, plus context, the words before the amount on its line); salary stays the first figure from the headline, so 0.6 integrations keep working. visa is sponsored, not-sponsored or null. urls lists the post's links, then links typed without a scheme ("acme.com/careers", returned as https://acme.com/careers).
• Story records (type: "story"): id, hn_url, title, url, domain, story_type (story, ask, show, job, poll), points, author, created_at, comment_count, text_markdown, comments[] (id, parent_id, author, created_at, depth, text_markdown), comments_stored, markdown, fetched_at.

SUMMARY.json in the run's key-value store holds the counts: items stored, skipped (duplicates and non-job comments), skippedByFilter, skippedSeenInPreviousRuns, API requests and retries, charged events, duration.

Pricing

Pay-per-event. Platform usage (compute, storage) is included.

Event	Price	When
job-parsed	$0.004 per post	Once per Who-is-Hiring post written to the dataset. Seen, filtered-out and non-job comments are free
story-exported	$0.003 per story	Once per story written to the dataset, comments included. Deleted, dead and seen stories are never charged
Actor start	$0.01 per run	Apify's standard start event

Examples: a full monthly hiring thread (about 250 posts) costs about $1.01; a daily watch schedule about $1 a month in total plus $0.30 for the 30 run starts; the front page with comments (20 stories) $0.07. Set Max total charge per run to hard-cap spend; the Actor stops before exceeding it.

Data sources

• Hacker News API (Firebase): lists and items. "There is currently no rate limit."
• Algolia HN Search API: search, date search and whole comment trees in one request. 10,000 requests per hour per IP.
• frankfurter.dev: ECB reference rates for annual_usd_*, one request per who-is-hiring run.

Only these public APIs are used; news.ycombinator.com pages are never scraped (its robots.txt asks for a 30-second crawl delay).

Verified

• 0.13.11 (2026-11-06): August to November 2026 threads (837 job posts): 12 fields changed in 11 posts, all checked by hand. company 6: three posts "Backed by Sequoia Capital, Peregrine is the operational AI platform …" -> Peregrine, "Company: esumry.com Location: West Austin TX …" -> esumry.com, "whitecircle.com/careers" -> whitecircle.com, null -> Sumble ("Sumble is the newco …"). salary_text 6 -> null, all company figures, not pay ("$70m" raised, "$42M" raised, "$30" trillion AUM, "$100"B cash flow, "$3.5"B GMV). Unit tests 53.
• 0.13.9 (2026-11-04): August to November 2026 threads (837 job posts): remote_region changed in 132, all checked by hand; no other field changed. Distinct place values 160+ -> 64 (non-places such as "VoiceAI", "YC 19", "Customer Success Engineer; structure", "Pre, seed", "and", "all" are gone; "US Canada" / "US & Canada" / "US/Can" all read United States, Canada). Worldwide 86 -> 119 (33 posts whose qualifier named no place). Unit tests 47, eslint clean.
• 0.13.8 (2026-11-03): August to November 2026 threads (837 job posts): 56 posts got remote_timezone, all checked by hand; remote_region changed in 50 ("US Timezones" -> United States, "EU time zone" -> Europe, "North America, Europe, UTC, 8 to UTC+2" -> North America, Europe, "United States, Eastern Central Time Zones" -> United States, "CET ±2h" / "PT, ET Hours" / "UTC, 1 to UTC+3" -> Worldwide). No other field changed. Worldwide 69 -> 86. Unit tests 45, eslint clean.
• 0.13.7 (2026-11-02): August to October 2026 threads (718 job posts): 7 posts changed, all checked by hand, only remote_region: Worldwide -> the region named in the headline (Fliplet "(UK/EU) & … (Canada)", Langfuse "EU-remote", FusionAuth ×2 "REMOTE (in USA, …", SentiLink "(UNITED STATES ONLY)", OneChronos "Flexible/U.S. Remote", 3Gimbals "(US/Remote)"). A time-zone preference ("Remote / EU/NA timezones") is not read as a region. Worldwide 70 -> 63. Unit tests 41, eslint clean.
• 0.13.6 (2026-10-31): August to October 2026 threads (718 job posts): 80 posts changed, all checked by hand. remote_region changed in 78: 23 posts that said Worldwide now name the region from the remote qualifier or a remote segment ("REMOTE (US) or NYC", "Remote, USA only", "REMOTE (US/Canada/Brazil/Poland/UK/India)"), and offices or stray segments mixed into the region are gone ("United States, Santa Clara Brooklyn" -> "United States", "PMF; United States" -> "United States"). location changed in 3 ("Denver, CO" -> "Denver, CO, USA"; a "Technical Trainer" role had been read as the location). Worldwide 93 -> 70. Unit tests 41, eslint clean.
• 0.13.5 (2026-10-30): August to October 2026 threads (718 job posts): 44 posts had a body-only salary_text with no parsed figure; 42 of them were company figures or fees and are now null (all checked by hand against the post text), the other 2 are real pay ("160k-240k", "$1.5K USD/mo") and are unchanged. No other field changed. Unit tests 41, eslint clean.
• 0.13.0 (2026-10-26): August to October 2026 threads (723 job posts, 295 with salary_text): 19 posts changed, all checked by hand. Removed from salary_text: funding rounds (Thunder Compute $17.5M, Solace $207M, Tasklet $42M, XP Health $50M), "$1" / "$10" fees, a split range (OneChronos "$180,000 / $250,000 / $150,000" -> "$180,000--$250,000 / $150,000--$200,000"); Reef's PLN figure now shown. Headline hints unchanged. Unit tests 39, eslint clean.
• 0.11.0 (2026-10-24): October and September 2026 threads (515 top-level comments, 267 figures): 4 posts state one role's pay in two or three currencies (Reef Technologies twice, Mozilla, vCluster Labs); their 5 extra figures now carry alternative_to and the first figure's context (Reef's PLN figure had context: "or"). No other output changed. Unit tests 38, eslint clean.
• 0.10.0 (2026-10-23): October 2026 thread (259 job posts): employment_types set on 180, two or three types on 6 (all checked by hand: "Full time or part time or contract", "Full-Time, Internship", "Full-time / Contract" ...), 0 posts where it disagrees with employment_type. Unit tests 37, eslint clean.
• 0.9.0 (2026-10-21): context names the role from a heading line above the pay line when the pay line has none (OneChronos-style posts: role paragraph, then "ESTIMATED COMPENSATION: ..."); null when two headings share one figure. A part-period figure restated as yearly ("$65k ($130k annualized)" for a 6-month contract) is read as the yearly $130k; "annualized revenue" is not pay. Measured on the October and September 2026 threads: 3 posts changed, no other output changed.
• 0.8.0 (2026-10-20): October 2026 thread: salaries 129 -> 141 figures, posts with a salary 87 -> 89 (DuckDuckGo 7 roles written "Role - $178,500 USD + equity", Middesk 5 ranges that had been dropped because "(2x/week in office)" was read as a weekly rate). September thread: 112 -> 125 figures, and Attendi's €6,000 - €7,000 is now monthly (was weekly). Every changed figure checked by hand, 0 new false positives; no figure that 0.7 parsed was lost. Unit tests 35, eslint clean.
• 0.7.0 (2026-10-19): October 2026 thread (259 job posts): 23 posts list 2-6 pay figures, 129 figures in salaries, every multi-figure post checked by hand (Baton 6 roles, Solace 5, Rebar 5, Vistulo 6 hourly PLN). RSU / PSU amounts ("up to $76,026.50 $HOST RSUs") are skipped. minAnnualSalary: 200000 keeps 47 posts (44 when only the first figure counted). salary changed on 1 post ($180,000--$250,000, was null); September thread: 1 change ($90–$110/hour, was $110/hour). Unit tests 34, eslint clean.
• 0.6.0 (2026-10-18): October 2026 thread, 259 job posts: parsed salary 77 -> 88, 11 new all checked by hand (5 exact, 2 up_to, 4 hourly ranges), 0 false positives; funding ($42M), signing bonus (€10k) and fixed-fee ($12K) amounts stay null. Unit tests 32, eslint clean.
• 0.5.0 (2026-10-17): October 2026 thread (266 top-level comments) with keywords: [typescript, rust], remoteOnly: true, postedWithinDays: 10: 9 posts stored, 250 filtered out (not billed), 3 API requests. Unit tests 31, eslint clean.
• 0.4.0 (2026-10-09): parser measured on the September and October 2026 threads. October: 247 top-level comments, 6 non-job comments skipped, role found in 178 of 241 posts (was 165 of 247), wrong locations (tech stack, body text, "100–140k€") removed. September: role 185 of 248 (was 176 of 252). Unit tests 30, eslint clean.
• 0.3.0 (2026-10-01): who-is-hiring with maxItems: 500: 3 API requests, 254 posts, 1.7 s. With seniority: [senior, staff, lead, principal] and minAnnualSalary: 100000: 26 posts stored, 228 filtered out.
• 0.1.0 (2026-09-22): front prefill (20 stories, comments depth 2, 50 per story): 41 API requests, 0 failed, 4.3 s. Pay-per-event path tested with a run budget: stops gracefully before the item that would exceed it.

Changelog

• 0.13.11 (2026-11-06): company is a name, not a label line, a careers URL or a sentence: "Company: X Location: Y" -> X, "acme.com/careers" -> "acme.com" (a bare domain such as "Fly.io" stays: it is the name), a long headline "Backed by …, Peregrine is the …" -> "Peregrine". A salary_text taken from a one-paragraph headline with no parsed pay now keeps only pieces that could be pay, as body text has since 0.13.5 ("raised roughly $70m" -> null).
• 0.13.10 (2026-11-05): urls also holds links an author typed without https:// ("vvd.world/careers", "www.psi.ch"), which HN does not turn into links, as https:// URLs after the real links. Only a domain with a path or a www. host counts; "Node.js/Express", a bare "example.com" in prose and e-mail domains do not. On the Aug–Nov 2026 threads (837 posts) 24 posts gained a link and 14 that had none now have one; no other field changed.
• 0.13.9 (2026-11-04): remote_region is rebuilt from known place names only (countries, regions, US coasts, provinces and about 90 cities, one spelling each: US/USA/U.S. -> United States, NYC -> New York, SF -> San Francisco, NL -> Netherlands). Words that are not places are dropped; two-letter US state codes stand for United States (never for themselves: "CA" can be California or Canada); a qualifier with no place at all falls back to the headline's regions, else Worldwide, like a bare Remote. Places are joined with , (was sometimes & or and).
• 0.13.8 (2026-11-03): new remote_timezone (working hours as written). Time-zone clauses ("US Timezones", "Pacific Time hours", "CET/CEST overlap", "UTC-8 to UTC+2", "2h overlap with US Pacific") are split out of the remote qualifier, location and tags before remote_region is read; regions they name count ("Europe / US Eastern Time zone" -> Europe, United States), zone names and offsets do not. A qualifier that is only hours is the whole answer, so other segments ("+ Quarterly Meet-ups", the office "Berlin") no longer become the region. "Remote / EU/NA timezones" now reads Europe, North America (was Worldwide in 0.13.7) to match "US time zones" -> United States.
• 0.13.7 (2026-11-02): a remote post that would read Worldwide takes the regions its headline names next to "remote" ("EU-remote", "US/Remote", "Remote/US", "REMOTE (in USA …") or alone in parentheses ("(UK/EU)", "(Canada)", "(UNITED STATES ONLY)"). An explicit worldwide / global / anywhere keeps Worldwide; a match followed by "timezones" / "hours" is a preference and is skipped.
• 0.13.6 (2026-10-31): remote_region is read from the remote qualifier ("REMOTE (US)", "Remote (EU time zone)") when there is one, so offices in other segments and stray segments no longer mix in; without a qualifier, a short remote segment kept in tags ("Remote, USA only") is read before falling back to Worldwide (a preference such as "PT/ET hours preferred" is no restriction and stays Worldwide). A place after "in" keeps up to two ", Region" parts ("Denver, CO, USA"). "Trainer" and "Instructor" are role words.
• 0.13.5 (2026-10-30): a body-only salary_text with no parsed pay figure keeps only pieces that could be pay (thousands, not followed by M / B / T or "million", not after raised / funding / valuation / revenue / GMV / market / over). Funding rounds ("$42M / $60M"), company scale ("$3.5B in annual GMV" was shown as "$3.5", "£50 billion" as "£50") and fees ("$5") now give null; "160k-240k" and "$1.5K USD/mo" stay.
• 0.13.4 (2026-10-29): when no headline segment is a location, the place named in a headline phrase is used: "… at NLnet foundation in Amsterdam" -> location "Amsterdam", "On-site (hybrid) in Stockholm, Sweden (Visa …)" -> "Stockholm, Sweden". A remote post that also names an office reads "Remote / Bremen, Germany" (remote_region is still taken from the remote part only). Only after an on-site / hybrid / office / based / located word or "at
• 0.13.3 (2026-10-28): a company name ending in a lowercase organisation word ("Rust developer at NLnet foundation in Amsterdam": company "NLnet foundation", role "Rust developer") is read instead of the whole headline. Words: foundation, lab(s), inc, ltd, llc, gmbh, group, institute, collective, studio(s), ventures, technologies. August to October 2026 threads: 1 of 719 posts changed, nothing else.
• 0.13.2 (2026-10-27): an amount followed by "+ 20% bonus", "+ 10% commission" or "+ benefits" is read as that fixed amount (bound: "exact"), not as "or more" (from). "$250k+" is still from. August to October 2026 threads: 1 of 719 posts changed ("Salary mid point: $165k + 20% bonus target").
• 0.13.0 (2026-10-26): salary_text for posts whose pay is only in the body is rebuilt from the parsed salaries (up to 3 figures), so funding rounds ("raised $17.5M"), fees ("$10") and split ranges ("$180,000--$250,000" shown as "$180,000 / $250,000") no longer appear in it. A pay hint in the headline is kept as written.
• 0.12.0 (2026-10-25): company is no longer a role name when the headline starts with the role ("Senior Python Backend Engineer | REMOTE", "Role: …"); the name comes from "join X" / "X is hiring" in the body or the headline sentence, else null.
• 0.11.0 (2026-10-24): salaries[].alternative_to: a figure that restates the previous one in another currency, joined only by "or", "/", "|" or "," on the same line, points at it and takes its context (Reef's "180-280 PLN" had context: "or", Mozilla's and vCluster's CAD / EUR bands had none). Filters are unchanged: minAnnualSalary with salaryCurrency: "PLN" already matched the PLN figure.
• 0.10.0 (2026-10-23): employment_types (every employment type the post states, same values and rules as the ATS Jobs Feed, including temporary) and the filter employmentTypes (same name as the feed; unknown keeps posts without a type; filtered posts are not billed).
• 0.9.0 (2026-10-21): context names the role from a heading line above the pay line; "($130k annualized)" is read as the yearly figure.
• 0.8.0 (2026-10-20): more salaries parsed: a lone amount after a spaced dash or a colon ("Senior Backend Engineer - $178,500 USD + equity", "Salary: $150,000"), yearly-sized or with an explicit rate, never after "bonus", "raised", "fee" and similar words. A pay interval is only read from the words that belong to the amount, so "($148K – $255K) — hybrid (2x/week in office)" is a yearly range (was dropped) and "€6,000 - €7,000 | 3 days/week" is monthly (was weekly). context drops an intro sentence on the same line as the first role.
• 0.7.0 (2026-10-19): salaries: one entry per pay figure, with context (the role written before it), for posts that list several roles. minAnnualSalary matches when any role reaches it. Also parsed: ranges written $180,000--$250,000, lone amounts in parentheses ("Junior SWE ($300k + equity)"); equity, RSU, PSU and bonus figures are never read as pay.
• 0.6.0 (2026-10-18): salary bounds are explicit: salary.bound is range, exact, up_to or from. Newly parsed: "$180k+" (from), "up to ~ $750k", a lone amount in a headline segment ("| £120k |", "| $181K base + 15% bonus |"), and rates written "/hr" ("$23-$34 USD/hour", "$50/HR"). The README wrongly said "up to $X" was read as min = max; it has always been min: null.
• 0.5.0 (2026-10-17): the ATS Jobs Feed's filters for who-is-hiring: keywords, locationContains, remoteOnly, postedWithinDays (same names and meaning, filtered posts not billed). With postedWithinDays, the previous month's thread is also read while the window reaches into it.
• 0.4.0 (2026-10-09): non-job top-level comments (complaints, job seekers, replies) are skipped and not billed; em-dash headlines are split; plural and multi-role segments ("Software Engineers", "Multiple roles") are read as the role wherever they sit; tech-stack segments go to tags instead of location; body text glued to the headline is no longer taken as role or location; FT / PT and "100–140k€" are recognised. Watch mode fix: the second run of a new watch no longer exports the whole previous month's thread (it re-checks only threads the watch already read). Watch state is saved when a run is aborted or migrated.
• 0.3.0 (2026-10-01): comparable job fields shared with the ATS Jobs Feed: seniority, years_experience_min, remote_region, structured salary with annualised and USD bounds; the raw hint moved to salary_text. Filters seniority, minAnnualSalary, salaryCurrency.
• 0.2.0 (2026-09-24): watch mode (stateKey); who-is-hiring re-checks the previous month's thread; date-sorted search stops at the first fully-seen page.
• 0.1.0 (2026-09-22): initial release. Nine modes, Markdown threads, Who-is-Hiring parser.

Support

Open an issue on the Actor's Issues tab. Issues are answered within one business day; fixes ship as new builds without changing the input schema.

Made by Vellum Labs.