GQ-005 · Where the Field Is Moving

Which categories, techniques, and projects are gaining real momentum, what mechanisms explain the change, and what does the acceleration predict next?

SISO Open Source Foundation

experimental

scoped

Not declared

Define the minimum public-safe question program before scheduling research.

gls:release:1ee81aa0-42cc-4bcd-a87b-4f34baf9cc13 · registry-seed-2026-08-01

Not released — the selected Release has no public answer artifact.

corpus

Where to direct research attention and which techniques to adopt early versus wait out.

Every predicted movement carries a date and a scoring rule fixed in advance, so it can later be marked right or wrong without renegotiation. · Each movement names an observed delta with the measurement window it came from, not a general impression of momentum. · Causal hypotheses are stated separately from the observations, so a correct observation with a wrong mechanism is visible as such. · Leading indicators are things that can be checked without waiting for the outcome they predict.

Dated predictions, once scored, do no better than a naive 'current trends continue' baseline. · Movements are only ever identified after they are widely known, meaning the map lags rather than leads. · The observed deltas are dominated by measurement artifacts — a source changing its reporting rather than the field changing. · Confidence ratings show no correlation with which predictions later score correct, making them decorative.

No movement map exists yet and no prediction has been made, so nothing can be scored. This contract sets the bar in advance, which is the only time it can be set honestly. · The baseline covers primary language on one curated GitHub corpus. A categories x time axis IS constructible — 55,467 repos join between momentum.sqlite (3 days) and the curated taxonomy in identity.sqlite — but it does not yet yield a signal. Measured 2026-08-04: the top category by star gain (agent-extension-pack, 2.06/repo against a 0.045 baseline) draws 92.1% of its gain from ONE repository, and only 29 repos in the whole corpus moved at all across the three days. The window is too short, not the taxonomy too weak — which is the opposite of the awesome_sections failure and needs more observation days, not a different source. · Star velocity is unusable FROM THE PEOPLE GRAPH — one distinct momentum_day (2026-07-11), star_delta null on 406,297 of 463,230 rows. CORRECTED 2026-08-04: a separate database, foundry-data/domains/github/momentum.sqlite, holds 170,062 repo_snapshot rows across THREE consecutive days (2026-07-09/10/11, ~56,687 repos each). Real deltas are computable there — 56,687 repos appear on all three days and the top mover gained 1,112 stars. So a velocity-based movement signal IS available; it was the graph that could not support it, not the machine. Three days is still a short window and the observations stop on 2026-07-11, so it cannot yet score a dated prediction.

A dated prediction reaches its date and must be scored. · A tracked source changes how it reports, which can create an apparent delta with no real movement. · A prediction is contradicted before its date, which is information that should not wait for the scoring window.

frozen Foundry momentum snapshots · category and capability movement · release and adoption signals · counter-signals that separate durable emergence from short-lived attention

A ranked movement map with observed deltas, causal hypotheses, confidence, leading indicators, and dated predictions that can later be scored.

Re-answer on a fixed cadence and whenever a leading indicator crosses a pre-registered materiality threshold.

public metadata only