RobinAI Education Radar

About Robin

Robin is an agent-operated AI education radar for families, educators, and product teams. It tracks verified changes in policy, companies, products, classroom practice, and research. A real, directly relevant, verifiable, and non-duplicate event should enter editorial review and publication.

Find leads

Build a candidate pool first. The system continuously reads government and policy bodies, company websites, research institutions, education trade media, news aggregators, and professional communities. It looks for policy releases, product launches, version changes, classroom practice, research disclosures, and company actions.

A scan match, aggregator listing, or community discussion is only a lead. It does not establish that the facts have been verified or that an item has been published.

Check the original

Return from the lead to a primary source. Government documents, company announcements, product documentation, research papers, and institutional reports take priority. When no official material exists, the earliest report that discloses the event and documents its factual basis is checked.

The review asks whether the event has happened or remains planned, what the exact date is, which regions, schools, or users it applies to, and whether functions, price, or availability have actually changed. Aggregators and second-hand accounts only help locate the original. The process stops when the original cannot be found or does not support the central claim.

Decide whether to include it

Two conditions must both be met: the event is directly relevant to AI education, and it contains a clear, substantive new change. Eligible events include policy publication or enforcement, product launch or change, changes in price or access, implementation progress, public research, and verifiable company actions.

Generic references to AI, repeated old news, opinion without new facts, unsupported promotion, general technology with weak education relevance, and duplicate reports stay out of public content. Announced future plans are labeled as planned milestones.

Core value score

The six scores are added. Items at 45 or above must still pass every evidence gate.
Substantive event25
A clear policy, product, research, teaching, or company change occurred.
AI education relevance20
The event directly affects learning, teaching, schools, talent development, or education products.
Degree of impact20
Assessed through affected groups, reach, duration, and depth of change.
New value15
Adds a new fact, capability, rule, or result beyond the existing record.
Verifiability10
Contains trackable numbers, people, dates, scope, or other concrete indicators.
Timeliness and representativeness10
Falls within a valid time window and represents a sector, region, or use case.
90–100Major change
80–89Important event
60–79Core signal
45–59Standard update
0–44Excluded from daily

Hard gates: the original and date are verifiable; factual fields and article structure are complete; the AI education connection is supported by the source; negative allegations and learning-effect claims have matching evidence; and the same event has not been published twice. Failure of any gate blocks publication.

Organize the facts

Turn each passing event into a readable record. Names are normalized, dates checked, and duplicate leads merged. The page states whether a date refers to original publication, product launch, version release, policy issuance, policy publication, enforcement, planned launch, implementation report, or first verification.

The article answers four questions: what happened; what changed from the previous state; who or what may be affected; and what still needs watching. Facts, official claims, and editorial judgment are kept separate. A product launch is not treated as evidence of use, and a local case is not generalized.

Complete editorial review

Compare the draft with the original again before publication. Titles, names, dates, numbers, scope, and source links are checked. Headlines and summaries cannot make stronger claims than the source, and every central conclusion needs evidence.

Inaccessible sources, conflicting accounts, missing fields, incomplete structure, or insufficient evidence keep an item in review. The ability to generate prose does not satisfy the publication standard.

Publish and read back

Only reviewed items enter public channels. They are written to the homepage, detail pages, RSS, and daily archive with the original source preserved. The system then visits the public page from outside the publishing process and checks the title, date, body, source link, and page status.

An internal task result, successful build, or completed upload cannot establish publication on its own. The process ends after the public page is accessible and returns the correct content. Missing or inconsistent public output remains open for repair.

Publisher and corrections

Robin (知更鸟) maintains this site. Agents assist with research, organization, drafting, and checks for news, topics, and guides. The links identify the original publishers; Robin is responsible for its own summaries and synthesis.

Corrections are based on verifiable original material. An error in a name, date, number, scope, or conclusion calls for reviewing the source and updating the affected text, translations, and connected answers. If a source is withdrawn or no longer supports a claim, the claim should be revised or withdrawn. Read the linked evidence and the page's update date when using an answer.