GitHub · new popular AI repos ↗
Newer repositories in this topic ranked by total stars; this is discovery, not measured star growth.
20 observed items · Last fetched 10/29/2026TRANSPARENT DISCOVERY
The source pool
These configured sources feed the live radar. The scheduled collection job checks them and stores compact signal metadata once for all readers. The resource library is a separate directory of places to explore.
For the reader-facing claim checklist, see How we check AI news →.
Primary Reddit discussions leave public discovery seven days after AINewsIQ first collects them. Saved briefs remain available. A Reddit link cited as evidence does not expire the story it supports.
4active sourcesPaper · discussion · repositories
Hugging Face Daily Papers ↗
Recent community submitted research papers.
276 observed items · Last fetched 10/29/2026Hacker News · AI ↗
Recent matching Hacker News stories with at least 20 points. Search: AI.
105 observed items · Last fetched 10/29/2026Hacker News · LLM ↗
Recent matching Hacker News stories with at least 20 points. Search: LLM.
48 observed items · Last fetched 10/29/2026How the radar lenses place signals
Every lens keeps the topic quadrants fixed. The five rings change meaning when you select a lens, and the labels beside the map show the current bands. Closer to the center means newer, more measured attention, a stronger interest match, a fuller stored source trail, or a stronger documented evidence trail, depending on the selected lens. The 24-hour and seven-day windows roll from the viewing time; neither resets at midnight.
TRUST LENS / V1Show the evidence trail, not a verdict.
Example: an automated X-discovered article with a fetched primary page and original link earns 65 points. No editor review is assumed.
65/100SOURCE EVIDENCE
30305
30 linked article30 fetched page5 original link0 editor review
Origin contributes
Original paper
+45
Original repository
+40
Linked article
+30
Community discussion
+5
Context can add
+30Fetched primary page
+15Feed text or abstract instead
+15Editor reviewed the link
+5Original non-discussion link
BELOW 2525–4950–6465–7980+
Source and review steps add points; visible source concerns subtract 5–20 points. The score records available context, not accuracy, peer review, or independent corroboration.
For you rubric
This lens is available when you follow a topic or have saved a briefed signal in the current feed. From center outward: followed topic plus at least two matching saved-article tags; followed topic plus one tag; followed topic or at least two tags; one tag; no match. It uses only those explicit interests. A match is not a judgment of importance, quality, or benefit to you.
Source trail rubric
From center outward: an editor reviewed the link for relevance and AINewsIQ fetched the source page; fetched source page and stored brief; brief made from a feed excerpt or paper abstract; original link without a stored brief; community lead without a stored brief. This is a record of available context and review steps. Editorial review of a link, a fetched page, and a source-based brief do not independently verify its claims.
Trust lens rubric v1
This is a source-evidence score, not a truth score. Original paper: 45 points; original repository: 40; linked article: 30; community discussion: 5. A stored brief from a fetched primary page adds 30 points; a feed excerpt or abstract adds 15. Editor review of the link adds 15, and an original non-discussion link adds 5. The five bands are 80+, 65–79, 50–64, 25–49, and below 25. An open source concern subtracts 5–20 points: 5 for a pending check, 8 for an unverified date, failed brief, or review need, 12 for unreadable text or an older source date, and 20 for an unmatched claim. Unknown peer review, lab claims, and author credentials earn no points. Popularity never adds trust. A high score means we can show more of the source trail, not that the findings have been independently verified.
The inner node gauge shows a freshness index based on elapsed time since the source or discovery post, reaching zero after seven days. It refreshes while the page is open and currently uses the viewer’s device clock. Node sizes remain equal because there is no stored editorial significance rating. Potential impact needs a separate, claim-linked rubric for consequences such as cost, security, or regulatory change; we do not infer it from attention or from the presence of a brief.
How does radar attention work?
The Time lens places sources in labeled publication-age bands. The Attention lens uses a source-specific index, capped at 100: Hugging Face likes or upvotes, Hacker News points, observed engagement on linked X posts, or GitHub stars gained between daily checks. A repository needs two daily star observations before growth can be measured. Lifetime stars remain context, not a recent-attention score. Observed Bluesky, YouTube, Reddit, and other platform counts remain visible context until a scale is published for each; they do not receive a guessed Attention index. The index is not comparable across platforms and does not measure credibility, editorial significance, or your personal relevance.
ATTENTION / SOURCE-SPECIFICFour scales. Four different measures.
Each reference level reaches 100 on its own logarithmic display scale. A missing observation stays unmeasured.
Hugging Face
70 upvotes
Hacker News
350 points
X post
2,500 weighted interactions
GitHub
100 stars gained/day
The bars illustrate each scale’s reference point, not current signal counts. X uses likes plus four times reposts on the strongest observed linked post. GitHub requires two daily observations to show growth. Scores across platforms are not directly comparable.
Attention index v1 rubric
Each source type has its own logarithmic display scale: 70 Hugging Face likes or paper upvotes, 350 Hacker News points, 2,500 weighted X interactions (likes plus four times reposts on the strongest observed linked post), or 100 GitHub stars gained per day between checks reaches 100. Counts above those reference levels stay at 100. The underlying counts appear in each node preview. These are display thresholds, not benchmarks of quality.
Linked discovery posts are sampled evidence, not a complete count of how often a URL was shared. Their metrics are not independently verified by AINewsIQ. A future link-mention count would need a documented search window, query, and coverage limit.
How are new signals shown?
New discoveries can appear while their source check is still open. A clickable flag explains an unreadable page, uncertain date, or claim that needs checking. The original source and discovery links remain available so readers can inspect the evidence. Confirmed dead links and signals removed by an editor leave public discovery. Social attention never establishes truth.