Hacker News

@hackernews what are the top 5 stories on the frontpage right now about AI or machine learning?

Here are the top 5 AI/ML stories currently on Hacker News frontpage: 1) 'New technique reduces LLM inference costs by 40%' (342 points, 89 comments), 2) 'Why transformer attention is all you need – revisited' (278 points, 54 comments), 3) 'Open-source alternative to OpenAI's Whisper' (201 points, 67 comments), 4) 'Machine learning for protein folding: lessons from AlphaFold' (189 points, 31 comments), 5) 'Training neural networks on a single GPU in 2024' (156 points, 42 comments). Each entry includes the current point count and comment thread size for context.

This example shows the MCP's ability to fetch and filter frontpage content in real time. No authentication required. The frontpage changes constantly, so results reflect a snapshot at query time. Useful for monitoring trending technical discussions without leaving your workspace.

@hackernews find recent posts about Rust performance benchmarks from the last week, limit to 3 results

Found 3 recent posts matching 'Rust performance benchmarks': 1) 'Benchmarking Rust vs C++ for high-frequency trading' (posted 2 days ago, 145 points, 78 comments) – discusses latency comparisons in production systems. 2) 'Why Rust's async runtime is faster than you think' (posted 4 days ago, 203 points, 112 comments) – deep dive into tokio internals. 3) 'Memory allocation patterns in Rust: a performance study' (posted 6 days ago, 87 points, 34 comments) – compares allocator strategies. Each result includes submission time, engagement metrics, and a brief excerpt from the discussion.

Demonstrates full-text search with time-based filtering. The MCP queries Hacker News's Algolia-powered search API, which indexes posts with a slight delay (usually under an hour). Search quality depends on keyword specificity—broad terms return more noise. Rate limits apply to search queries during high-traffic periods.

@hackernews pull the comment thread for item 39847362 and summarize the main technical disagreements in the discussion

Retrieved item 39847362: 'Show HN: I built a distributed task queue in Go' (187 points, 94 comments). Main technical disagreements: 1) Architecture choice – several commenters argue a centralized coordinator creates a single point of failure; author defends with failover strategy details. 2) Comparison to existing tools – debate over whether this adds value beyond Celery or Temporal; author highlights lower operational overhead. 3) Consistency guarantees – discussion splits on whether at-least-once delivery is sufficient for the stated use case. The thread also includes 12 comments praising the documentation quality and 8 offering feature suggestions for the roadmap.

Shows how the MCP retrieves specific items by ID and enables AI synthesis of discussion threads. The response is truncated to prevent context overflow—deeply nested threads may be summarized rather than returned in full. Useful for extracting signal from long comment chains when evaluating tools or techniques mentioned in your team's work.

When HN search beats manual browsing for market research

A 3-person SaaS team runs a Friday competitive intel sync where they review what competitors shipped and what the market is talking about. The Hacker News MCP wins here because the search tool lets you query by keyword and date range in one shot—no tab-hopping through Show HN threads or manually filtering launch posts. You can pull all mentions of a competitor's name or a technology keyword from the past week, then feed those item IDs into the get-item tool to grab comment threads without context overflow. The threshold: if your vertical rarely hits HN (say, enterprise HR software), this MCP returns empty results and wastes time. But if you're in dev tools, AI infra, or consumer tech, it's faster than a Slack channel full of links.

Using HN threads to answer edge-case product questions

A solo founder or 2-person support team fields technical questions about integrations or edge cases their docs don't cover. The Hacker News MCP is useful when your product has been discussed on HN and those threads contain real-world troubleshooting or workaround advice from users. You search for your product name, pull the top 5 threads, and scan comments for the specific error message or workflow question. The get-item tool limits response size so you don't blow your context window on a 400-comment flame war. The trade-off: this only works if your product has HN traction—if you're pre-launch or in a non-tech vertical, the corpus is empty. For dev-facing products with active HN presence, it's a faster alternative to GitHub issue archaeology.

When daily HN scraping feeds editorial calendar planning

A 4-person content or marketing team at a B2B startup needs to spot emerging topics before they're saturated. The Hacker News MCP's today-posts and latest-posts tools let you pull fresh submissions each morning and filter by upvote threshold or keyword. You run a daily query for your category (say, 'open source' or 'LLM'), export the top 10 posts, and use that as input for a weekly editorial meeting. The frontpage tool is less useful here because it's too broad—you want the firehose with filters, not just what hit the top. The boundary: if your team already has a social listening tool or RSS setup, this MCP is redundant. But if you're bootstrapped and HN is your primary signal source, it's cheaper than a Feedly subscription and easier to automate.