What this is:a genuinely live digest — every refresh fetches real, current items from Hacker News (Algolia search API) and freshly submitted papers from arXiv's cs.AI and cs.LG categories. Sort by significance or recency, bookmark items, and browse your local archive of past refreshes. Optionally add your own OpenAI, Anthropic, or Groq API key below to get a 1-2 sentence AI summary and significance score per item — entirely optional, and your key is never sent to our servers.
Your key is held only in this tab's memory — it is never persisted to localStorage/cookies, never logged, and never sent to EngineersUniverse's servers. When you click "Summarize" on an item below, your browser calls Groq's API directly with your key. Reload the page and the key is gone.
This tool builds a live, scannable digest of what's happening in AI right now, pulled fresh from public sources every time you refresh — not a static, pre-written article. It works as a genuinely useful browsing tool with zero setup, and adds an optional layer of AI-generated summaries and significance scores if you supply your own API key from OpenAI, Anthropic, or Groq.
A server-side route on this site fetches two categories of public, no-signup sources on every refresh. The first is Hacker News, via the public Hacker News Algolia Search API (hn.algolia.com) — the digest runs several targeted searches (LLM, GPT, Claude, Gemini, OpenAI, Anthropic, machine learning, neural network, transformer AI) restricted to stories from the last 7 days, merges and deduplicates the results, and then re-checks that each title or story text actually mentions an AI-relevant term before it's allowed into the digest, since Algolia's own relevance ranking is fuzzy and typo-tolerant enough to occasionally surface loosely related stories. The second is arXiv's public API (export.arxiv.org), pulling the most recently submitted papers from the cs.AI (Artificial Intelligence) and cs.LG (Machine Learning) categories directly — these categories are inherently AI-relevant by definition, so no keyword filtering is needed there. Reddit's public JSON endpoints were evaluated during development but excluded: Reddit currently returns HTTP 403 to unauthenticated requests (including from this site's own server), making it too unreliable as a live source right now.
Hacker News's Algolia API sends an Access-Control-Allow-Origin: * header, meaning a browser could call it directly with no proxy at all. arXiv's API does not send that header, so a browser fetch to it is blocked by CORS before the response is ever readable — it has to be fetched server-side and handed to the browser in a normalized shape instead. This site's /api/ai-news-digest route does both fetches server-side for a consistent reason: it lets the browser make one simple request and get back one normalized list of items, regardless of which upstream API happened to support direct browser access.
Without ever touching an LLM, the digest is already useful: the "Recent" sort is a straightforward reverse-chronological feed across both sources, and the "Top" sort ranks Hacker News items by a simple engagement score (points plus twice the comment count, a common proxy for how much a story resonated) while arXiv papers — which have no public community score — are ranked by recency instead, exactly as a plain significance heuristic should work without needing a model to read anything. Source filters (Hacker News / arXiv cs.AI / arXiv cs.LG) and a keyword filter box let you narrow the list further, all client-side and instant.
If you paste your own API key into the panel above, an item-level "Summarize" button appears. Clicking it sends that single item's title and excerpt directly from your browser to your chosen provider's chat completion endpoint (OpenAI's /v1/chat/completions, Anthropic's /v1/messages, or Groq's OpenAI-compatible /openai/v1/chat/completions) asking for a 1-2 sentence summary and a 1-10 significance score. That request goes straight from your browser to the provider's API — it never passes through EngineersUniverse's servers, is never logged by this site, and the key itself is only ever held in this tab's in-memory React state, never written to localStorage, cookies, or any file. Reloading the page clears it. Because you're calling the provider with your own key, usage is billed to your own account at that provider's normal per-token rate (Groq's free tier is typically the cheapest way to try this feature — see the Token Counter & API Cost Calculator linked below to estimate cost before summarizing many items).
Starring an item saves a copy of it to your browser's localStorage under this site's domain — it is not synced to any account or server, so favorites are specific to the browser and device you're using, and clearing site data will remove them. Similarly, every successful refresh is automatically saved as a timestamped snapshot in a local archive (capped at the most recent 8 refreshes) so you can look back at what the digest showed you earlier, even after the live feed has moved on. Both features work with zero setup and no API key.
No. The live digest — real items from Hacker News and arXiv, sortable by recency or engagement/significance, with favorites and an archive — works fully without any API key. The API key is only needed for the optional per-item AI summary and significance score.
No. The key lives only in this page's browser memory for as long as the tab stays open, and the only network calls it's used in go directly from your browser to your chosen provider's own API domain (api.openai.com, api.anthropic.com, or api.groq.com). This site's backend never receives, logs, or stores it.
It comes down to CORS headers. Hacker News's Algolia search API explicitly allows cross-origin browser requests, so it could technically be called directly. arXiv's API does not send that permission header, so browsers block reading its response directly — this site's server fetches it instead and returns a normalized result so the browser only ever has to make one simple request.
The Hacker News searches are deliberately over-inclusive at the API level (Algolia's search is fuzzy/typo-tolerant), so this tool applies a second keyword check afterward requiring the title or story text to actually contain an AI-relevant term. This trades a small amount of recall for a materially more relevant, less noisy digest.
arXiv papers don't have a public community engagement signal the way Hacker News stories have points and comments, so there's no fair apples-to-apples significance score to compare them on. "Top" sort therefore ranks by real engagement where that signal exists (Hacker News) and treats papers as unscored, effectively separating the two — switch the source filter to arXiv cs.AI or cs.LG to browse papers on their own, sorted by recency.
No — both are stored in this browser's localStorage on this specific device, with no server-side account behind them. They will not appear if you open this tool in a different browser or device, and clearing your browser's site data will remove them.
Try our AI Engineering Studio
More calculators, simulators, and guides for this discipline.