Hacker News Agent

Find the Hacker News conversations worth your attention.

Kanso searches Hacker News for the topics you choose, gathers the context around each conversation, and tells you which ones are worth your time and why. It researches and recommends. Kanso does not post to Hacker News.

Free to start · No credit card required. No Hacker News account needed. See how it works

Hacker News opportunity
Research and drafting notes only. Hacker News guidelines say: "Don't post generated text or AI-edited text." Read the thread yourself, decide whether to take part, and write anything you post entirely in your own words. Disclose your affiliation with the project if you mention it.
Existing conversation on Hacker News
Ask HN: What do you use to track stock for a small retail shop?
Hacker News comment · comment opportunity · posted 6h ago
Comment opportunitystock forecastingKeyword in the item
Why Kanso recommends this

The comment describes the exact problem your product addresses, a small team unsure when to reorder, and the thread is still recent.

Research notes: write your own reply in your own words
Relevant aspect

How small teams decide when to reorder stock.

Points to address
  • Explain how you set a reorder point for a slow-moving item
  • Say what you would track before buying software at all
  • Be open that you work on inventory software
Caution

Disclose that you work on a related product if you mention it, and keep it about the question rather than the product.

These are research notes, not a draft reply. There is nothing here to copy or paste. Read the thread yourself and write your reply entirely in your own words.

Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
The problem

Finding the right conversation is slow work

Hacker News is full of useful discussion. The hard part is finding the few threads that matter to you, quickly enough to be useful.

Conversations move quickly

Relevant discussions appear and fade within hours, and the front page changes all day.

The lists are broad

The front page, newest, best, Ask HN and Show HN are not about your topic. Finding the part that is takes scanning all of them.

Search is manual

Searching for each topic, for both stories and comments, means running the same searches again and again.

You see the same things twice

Without a record of what you have already looked at, the same threads keep coming back.

Context takes time

A comment only makes sense with its parent, the story and the replies around it, and your own company's position.

Deciding is the hard part

Whether to join a conversation, and what you could usefully add, takes judgment that is easy to rush.

The difference

Without Kanso, and with it

Without Kanso
  • Searching Hacker News for each of your topics
  • Scanning the front page, newest, Ask HN and Show HN
  • Filtering out what you have already seen
  • Opening each thread to read the comment, its parent and the story
  • Working out whether it is relevant to your company
  • Deciding what is worth acting on
With Kanso
  • Your keywords searched across stories and comments, and five Hacker News lists scanned
  • Items that do not mention your topics filtered out
  • Items you have already been shown, or dismissed, left out
  • The thread's context gathered for you
  • A reason given for why each conversation is relevant, and what to watch out for
  • A recommendation of what to do next, waiting in your Agents Feed

You decide what to do. Kanso does not post to Hacker News.

What it is

From a flood of threads to the few worth your time

The Hacker News Agent is a research assistant, not a posting tool. It does the searching, filtering and reading, and leaves every decision, and every word on Hacker News, to you.

01

Discover

Searches your keywords and scans five Hacker News lists.

02

Filter

Keeps what mentions your topics and skips what it has seen.

03

Gather context

Reads the thread, and uses your company context.

04

Evaluate

Judges whether the conversation is worth your time.

05

Recommend

Explains why and what to do next.

How it works

From your keywords to a recommendation in six steps

STEP 01

Choose your keywords

You tell Kanso what to look for.

  • Up to eight words or short phrases, such as a product category or a problem you solve
  • Add or remove them any time. Each change is saved straight away
  • Kanso needs at least one keyword to recommend anything, because the keyword check comes first
  • There is nothing to connect: no Hacker News account and no login
Hacker News Agent
Keywords

These keywords define what Hacker News discovery searches for. Stories and comments are only considered when they mention one of these words or phrases.

inventory managementretail analyticsstock forecasting
Add a keyword or short phrase
Up to 8 keywords, 64 characters each.
Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
STEP 02

Discover

Kanso reads Hacker News's public data from the last 48 hours.

  • Scans the first five stories of the top, new, best, Ask HN and Show HN lists
  • Searches stories and comments for each of your keywords
  • Reads a few top comments under the stories it finds
  • If one source is unavailable, the run carries on with the rest
Hacker News discovery
Where Kanso looksLast 48 hours
  • Official lists
    Top, new, best, Ask HN and Show HN: the first five stories of each
  • Keyword search
    Stories and comments that mention one of your keywords
  • Comments under stories
    A few top comments and replies under the discovered stories

Read from Hacker News's public data. No Hacker News account and no login.

Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
STEP 03

Filter and deduplicate

Only what is about your topics, and new to you, goes forward.

  • An item has to mention one of your keywords, or the title of its story has to
  • Each item is looked at once per run, however many ways it was found
  • Anything Kanso has already evaluated for your project is not shown again, even if you dismissed it
  • Up to fifteen items per run go on to the next step
Filtering and deduplication
  • Ask HN: What do you use to track stock for a small retail shop?Goes forward
    Story · Mentions "inventory management"
  • A comment about reorder points in the same threadGoes forward
    Comment · Mentions "stock forecasting"
  • Show HN: A tiny terminal text editorSkipped
    Story · No keyword match, so it goes no further
  • An item Kanso already evaluated for this projectSkipped
    Story · Already evaluated, so it is not shown again
Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
STEP 04

Gather context

A comment means little without what surrounds it.

  • Reads the comment, its parent, the story and the first replies
  • Brings in your company context when it is ready
  • Adds a few web search results where they are available
  • Treats everything from Hacker News or the web as untrusted data, never as instructions
Context for one item
From the Hacker News thread
  • The comment: We outgrew spreadsheets once we had two locations. Reorder points were the part we kept getting wrong.
  • Its parent comment: Small shops often lose track of which items are about to run out.
  • The story: Ask HN: What do you use to track stock for a small retail shop?
  • First replies: Two short replies to the comment, read for context.
Your company context
Acme Inc. makes inventory software for small retail shops. Used when your website context is ready.
Web evidence
A few web search results about reorder points, kept only if they are cited.

Everything that comes from Hacker News or the web is treated as untrusted data. Only your own company context is trusted.

Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
STEP 05

Evaluate and recommend

One AI evaluation per thread, with the whole context in view.

  • Decides whether the conversation is worth your time, and says why
  • Names the part of it that matters to you, and what to be careful about
  • Flags anything that needs fact-checking and shows the evidence it used
  • Can conclude that a thread is not worth your time, and keeps it out of your feed
What an evaluation can produce
Comment opportunity
An existing conversation worth joining: why, which part matters, and points to address in your own words.
Story opportunity
Submission idea, not a finished submission
Title idea (check it against the source before submitting)
A practical guide to reorder points for small shops
Source URL
You would need to supply your own source for this submission.
Framing notes
Explain the method step by step and say how you tested it.
Evaluated, not recommended
Show HN: A tiny terminal text editor
The thread is about a text editor. Your product has nothing relevant to add, so Kanso did not recommend it. Kept quietly so the same thread is not evaluated again.
Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
STEP 06

Review and decide

The recommendation waits in your Agents Feed. What happens next is up to you.

  • Open the conversation on Hacker News and read it yourself
  • Notes only: the points a reply could address, or an idea for a submission, never finished text
  • Ask Kanso to re-evaluate the same item, dismiss it, or mark it complete
  • Kanso never posts. Anything you write, you write yourself, in your own words
Agents Feed
Hacker News Agent1 opportunity
Ask HN: What do you use to track stock for a small retail shop?
Hacker News comment · comment opportunity · posted 6h ago
Not provided
  • Posting or submitting to Hacker News
  • Commenting or replying
  • Voting
  • Logging in to a Hacker News account
  • Writing the finished comment for you
Illustrative example. The company, keywords, thread and evidence shown are invented, not real data.
Capabilities

What the Hacker News Agent does

Hacker News discovery

Scans the top, new, best, Ask HN and Show HN lists, reading Hacker News's own public data. No account is needed.

Your keywords

Searches stories and comments for up to eight words or short phrases that you choose and can change at any time.

Relevance filtering

Only items that mention your topics go forward, so you are not shown the rest of Hacker News.

Deduplication

Each item is looked at once, and anything it has already evaluated for you is not shown again.

Thread context

Reads the comment, its parent, the story and the first replies, and brings in your company context.

Reasoned recommendations

Each recommendation says why the conversation matters, what to be careful about, and what needs checking.

Two kinds of opportunity

A conversation worth joining, or an idea for a submission of your own. Or a judgment that it is not worth your time.

Agents Feed delivery

Recommendations wait in your feed. Open one on Hacker News, ask Kanso to re-evaluate it, dismiss it, or mark it complete.

Does not post to Hacker News

It never posts, comments, replies or votes, and never logs in. What you write, and whether you write anything, is up to you.

When it runs

Start a run with Run HN Agent whenever you like. A daily scheduled run is built in and currently paused.

The boundary

Kanso researches. You take part.

Hacker News has no way for Kanso to post, and Kanso is built not to try. It is a clear line, and it is deliberate.

What Kanso does

  • Searches Hacker News for your keywords
  • Scans five Hacker News lists
  • Filters out what does not match, and what it has already shown you
  • Gathers the context around a conversation
  • Explains why a conversation is relevant, and what to be careful about
  • Recommends what to do next
  • Re-evaluates an item when you ask

What Kanso does not do

  • Post or submit to Hacker News
  • Comment or reply
  • Vote
  • Log in to or manage a Hacker News account
  • Schedule or automate anything on Hacker News
  • Write the finished comment or submission for you
  • Track karma, votes, replies or results

Why notes and not drafts

Hacker News's guidelines say not to post generated text or AI-edited text. So every recommendation carries a reminder of that, and gives you research notes to work from instead of words to paste. You read the thread, you decide whether to take part, and you disclose your connection to your company if you mention it.

Compared

Researching Hacker News by hand, and with Kanso

Manual Hacker News researchKanso Hacker News Agent
Finding conversationsBrowse the lists and search by handScans five Hacker News lists and searches your keywords across stories and comments from the last 48 hours
Keeping track of what you have seenMemory or notesRemembers every item it has evaluated for your project, and does not show it again
Thread contextOpen each thread and read itGathers the comment, its parent, the story and the first replies
Your company's contextIn your headUses your company context when it is ready
Deciding what is relevantYour own judgmentA recommendation with its reasoning and cautions, which you can have re-evaluated
Running itWhenever you have timeRuns when you start it. Scheduled runs are built in and currently paused
Writing a comment or submissionIn your own wordsResearch notes only. You write anything you post, in your own words
Posting to Hacker NewsYou post it yourselfNot provided. Kanso does not post, comment, reply or vote
A Hacker News accountYour own accountNot required. Kanso never logs in
Measuring resultsHacker News's own pagesNot provided. Kanso does not track karma, votes, replies or results

This compares doing the research by hand with using Kanso for it. It makes no claim about results: Kanso does not measure traffic, votes or replies, and what you do with a recommendation is up to you.

Part of Kanso, not a separate tool

The Hacker News Agent works from the same company context as Kanso's other agents. Start with your website, and see the pricing page for which plans include it.

See pricing →

Got questions?

Hacker News Agent questions

Use cases

Where the Hacker News Agent is used

These use cases name this agent in their workflow. Each one says what it produces and where you stay in control.

Browse all use cases

Find the conversations worth your attention.

Kanso does the searching, filtering and reading, and tells you what is worth a look. You decide what to do, and Kanso never posts to Hacker News. Enter your website to start.

Get started free · No credit card required