One headline score, and the eleven indicators behind it: what each means in plain words, the recorded reason for any change, and how stable the picture has been across the published quarters.
The Black Hole Index gives each platform one number, B: how much of a user’s work the platform holds (capture) divided by how easily all of it could leave (escape). Above 1, more holds than can leave. B is a structural ratio, not a price, a time or a probability; its five zones are reading bands, and only B = 1 is definitional. B is only an entry point. This page shows the primary indicators it is built from, what each means in plain words, and how stable the picture has been across the published quarters.
Release Q4-2026 · compared with Q3-2026
184
platforms scored
135
black hole
32
event horizon
13
transition zone
2
growing gravity
2
useful tool
Stability. Of the 184 platforms present in both quarters, none changed its B and 184 kept their zone (100%). Rank order between the two quarters: Spearman ρ = 1.00. A score moves only when a recorded reassessment moves one of the indicators; each observed change carries its recorded reason on the platform’s page, and the modelled Q4-2025 back-projection is not a change.
capture 2.86 ÷ escape 0.17 · since Q3-2026: ±0.00, same zone, same rank
B across published quarters
Q4-2025Q1-2026Q2-2026Q3-2026Q4-2026
Range 17.03–17.03 over 5 published quarters; the hollow point is a modelled back-projection, not an observed score.
What holds a user in
The six capture indicators. The more of a user’s data, memory, actions, processes and relationships sit inside the platform, the higher they score.
Data Depth (d)
9 / 10
Volume and density of observable user and organizational context
Rubric anchor 7 · Full: Complete work context: emails, documents, calendar, contacts, meeting notes. Example: M365 Copilot. Scored between anchors 7 and 10 (Org exhaust).
Memory (m)
9 / 10
Persistent non-portable state — embeddings, profiles, learned policies
Rubric anchor 7 · Persistent: Deep cross-session memory: references all past conversations, learned work style. Scored between anchors 7 and 10 (Model).
Action (a)
9 / 10
Ability to initiate and complete real-world actions autonomously
Rubric anchor 7 · Execution: Plans and executes multi-step task sequences. Sends emails, schedules meetings, creates PRs. Scored between anchors 7 and 10 (Autonomous).
Process (p)
10 / 10
Degree of embedding inside daily workflow chains
Rubric anchor 10 · Critical path: Business processes cannot complete without this system. Downtime equals revenue loss.
Rubric anchor 7 · Cross-side: Strong cross-side network effects: more users attract more developers attract more users. Scored between anchors 7 and 10 (Standard).
Closeness (c)
10 / 10
Frequency of contact and cognitive offloading habit formation
Rubric anchor 7 · Enterprise: Enterprise-wide deployment. Procurement contracts, training, IT infrastructure aligned. Scored between anchors 7 and 10 (Rebuilt).
Momentum (t)
5 / 10
Capability growth velocity — Arthur increasing returns
Rubric anchor 5 · Market pace: Keeping pace with competitors. No structural advantage.
What changed, and why
No recorded change of B for WeChat within the published quarters (Q4-2025 to Q4-2026). History: platform page or the history endpoint.
Scores are the sealed Q4-2026 release, read through the same pointer as /api/bhi; indicator definitions and the full five-anchor rubric are on Parameter Explorer, the formula on Methodology, and the sensitivity of B to scoring noise on Robustness. No confidence interval is shown because none has been derived. Data: CC-BY-4.0, /api/indicators.