Hacker News hiring thread parsed into an Airtable job list

Hacker News hiring thread parsed into an Airtable job list

An n8n workflow that turns the monthly "Ask HN: Who is hiring?" thread into a tidy job list. It finds the latest thread, loads every job post under it, cleans the text, and asks an OpenAI model to pull out the company, title, location, pay and links. Each post becomes one row in an Airtable table.

What is inside

• workflow.json: the workflow as n8n exports it, 20 nodes. Fifteen of them run and five are sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• Search for Who is hiring posts posted a copy of a browser request to the search backend of Hacker News Search. It carried the site's app ID, browser headers, and an API key that the file did not hold. It now calls the documented search API at hn.algolia.com/api/v1/search_by_date, which needs no key. It asks for stories that match Ask HN: Who is hiring, newest first, 30 at a time. The reply still holds title, created_at, updated_at and story_id, as we checked on 12 November 2026.
• OpenAI Chat Model used gpt-4o-mini. OpenAI's deprecation page, read on 12 November 2026, shuts that model down on 14 January 2027 and names gpt-6-luna. The node uses gpt-6-luna, and the note about the data structure says so.
• In Structured Output Parser, the type of each field was written as ["string", null], with null unquoted. It is ["string", "null"] now, which is what JSON Schema expects.
• Two node names had typos. Trun into structured data is Turn into structured data, and HI API: Get the individual job post is HN API: Get the individual job post.
• The sticky notes that told you to copy a browser request, and that linked a sample Airtable base, are rewritten.
• The Airtable base and table IDs are placeholders, <YOUR_AIRTABLE_BASE_ID> and <YOUR_JOBS_TABLE_ID>. The credential links are gone, and the file's name is ours.
• Nothing else. The cleaning code, the prompt and the field list are as the file had them.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key.
• An Airtable account with a personal access token and a base that holds a table with these text fields: type, title, salary, company, location, apply_url, company_url and description.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Open Write results to airtable. Add an Airtable credential and replace the two placeholders with your base and table. A base ID starts with app and a table ID with tbl, and both are in the Airtable address.
3. Add an OpenAI credential and select it in OpenAI Chat Model.
4. Press Test workflow. Limit for testing (optional) keeps the first 5 posts, so a first run is small.
5. Check the new rows. Switch the limit off, or raise it, when you want the whole thread.

What happens in a run

1. When clicking ‘Test workflow’ starts the run by hand.
2. Search for Who is hiring posts asks Hacker News Search for the hiring threads. Split Out makes one item per hit, Get relevant data keeps title, createdAt, updatedAt and storyId, and Get latest post keeps the ones created in the last 30 days.
3. HN API: Get Main Post loads the thread from the official Hacker News API. Split out children (jobs) makes one item per reply ID, and HN API: Get the individual job post loads each reply.
4. Extract text keeps the reply text. Clean text removes the HTML tags and entities and puts web links on their own lines. Limit for testing (optional) keeps the first 5.
5. Turn into structured data sends each post to the model with the message "Extract the JSON data". Structured Output Parser shapes the answer as company, title, location, type, work_location, salary, description, apply_url and company_url. Any of them can be empty.
6. Write results to airtable adds one row per post with eight of those fields. work_location is not saved.

What to change

• The thread to follow: the query in Search for Who is hiring posts. Another phrase finds another kind of post.
• The fields: the schema in Structured Output Parser, and the columns in Write results to airtable.
• The test limit: Limit for testing (optional).
• The model: OpenAI Chat Model.
• Where the rows go: replace the Airtable node with a sheet or a database.

Good to know

• A thread has hundreds of replies. Each one is a request to Hacker News and a call to the model, so a full run takes a while and costs more than a test.
• The thread is posted once a month. A run more than 30 days after the latest one finds no thread.
• Running it again adds the same posts as new rows. Nothing checks for repeats.
• The model reads free text, so it can miss a field or put the wrong thing in one. Read a sample before you rely on the list.
• The posts are public and often hold contact details. Keep the table to the use you have in mind, and check the site's terms for anything wider.
• Data: the text of each post goes to OpenAI, and the rows go to Airtable.
• Cost: OpenAI charges per post.