Airffic InstituteAERIAL PHYSICAL AI

Aerial Physical AI Institute

Methodology: what we chose, how, and how deeply we read

Research is worth what its method is worth. We state the limits too.

Updated: · 2 min readHow we researched →

Sir Airffic GoIf you find an error, tell us through the link at the bottom of the page. Fixing is part of research.

Selection

  • Papers: by theme (state estimation, detection, learning control, sensors, swarms, foundation models, counter-UAS and more), we chose influential work later used in products or research.
  • News: from 6 November 2006 to 6 November 2026, we chose products, funding, acquisitions, regulation, defense and partnerships that changed the market's direction, merging follow-up stories into one record.
  • Companies: we chose firms with different strategic models across seven layers, from sensors to airspace management.

Reading depth is stated separately

Seeing a search result is not reading. We separately mark full text read, abstract only, excerpt only, and title-and-date only.

Summaries are rewritten for a 14-year-old, so check originals for precision.

Fact and interpretation are separated

What happened is written as fact; what it means is written as interpretation, in a separate field.

The entry-strategy page includes interpretation from Airffic's perspective. It is not a neutral third-party assessment.

What we do not know (limitations)

  • We publish no estimates of market size or company revenue, margins or share, because reliable public figures are scarce.
  • Military and defense information is limited to what is public; real performance and full contracts are unknown.
  • Announcements by private companies are their own accounts, not third-party verification.
  • Sources are mainly English and Japanese; Chinese and other-language sources are under-read.

Any tool can continue the work

All text and data are stored as plain JSON. Claude, ChatGPT, Grok or a person can edit the same files and rebuild the site.

Where work stopped is always recorded in the repository's PROGRESS.md.