Frontier is an open AI intelligence stream. It continuously collects public sources, deduplicates related events, ranks their relevance, and publishes a quality-gated feed in English and Simplified Chinese.
Read the latest AI news at frontiermemo.com
Frontier uses AI-assisted classification, translation, curation, and synthesis. Important claims should always be verified with the linked original publisher.
Frontier is available as a standalone Agent Skill. Give an agent this prompt:
Read https://raw.githubusercontent.com/Alex-yang00/frontier/main/skills/frontier/SKILL.md,
then use Frontier to get the latest important news in the AI industry.
Every candidate receives a score from 0 to 100:
score = clamp(0, 100,
source_quality
+ popularity
+ freshness
+ high_signal_terms
+ AI_relevance
)
| Component | Range | Calculation |
|---|---|---|
| Source quality | 13-30 | Established first-party and specialist sources receive higher prior weights; unknown sources default to 15. |
| Popularity | 0-25 | Text: min(25, points // 20 + comments // 15). Video: min(25, floor(8 x log10(views / 1,000))) for at least 1,000 views. |
| Freshness | 0-25 | max(0, 25 - floor(age_hours / 4)). The score falls by one point every four hours. |
| High-signal terms | 0-20 | Weighted terms such as release, launch, funding, acquisition, security, benchmark, and open source. |
| AI relevance | 0-15 | round(model_relevance x 15), where model relevance is constrained to 0-1. |
Source quality is a prior, not an editorial verdict. Popularity is capped so a single network cannot dominate, freshness decays automatically, and the final published shortlist must also pass bilingual editorial review, source-diversity checks, deduplication, and section-level quality gates.
Frontier's code is MIT licensed. Frontier-authored editorial fields are available under CC BY 4.0; third-party material remains subject to publisher rights. See DATA_LICENSE.md for details.