Skip to main content
Compendium · Compositions · Signal Trends

Signal Trends

Items
16
of 1002 total

Full-featured composed scenes — one per row so each design has room to breathe. Click for the live render, the spec, and the source.

№ 1381

AAR Trend Dashboard

Scene STA go
AAR · 12 weeks 5 specs

Alignment is falling.

// signal: agents and specs are drifting apart on two of five

portfolio mean 82%
weekly delta 0
below floor 2 / 5
floor 85%
AAR % over time y: 50→100
FLOOR 85
W01W02W03W04W05W06W07W08W09W10W11W12
SPEC-204 auth & session SPEC-118 billing reconciler SPEC-302 event ingest SPEC-415 access policy SPEC-227 search rebuild
FALLING SIGNAL · WHAT IT MEANS 3 specs
  1. SPEC-227 -19 Agent guessing more, asking less — spec ambiguity compounding.
  2. SPEC-204 -18 New auth flow specs lagging implementation — drift forming.
  3. SPEC-415 -7 Policy edits not landing back in spec — convention erosion.
// owner: @governor · falling AAR is the leading indicator drift cares about
Compositions / signal trends

Multi-spec Agent Alignment Rate trends with falling-signal callouts and plain-language readings.

open ↗
№ 1382

GTR Trend Dashboard

Scene STA go
GTR · 16 weeks

Triggers are rising.
The question is why.

this week · GTR 29
real / total 76%
target floor 70%
REAL VS NOISE — STACKED per 100 diffs
W01W02W03W04W05W06W07W08W09W10W11W12W13W14W15W16
realcaught a defect noisefalse positive
HOW TO READ A RISING LINE
  • WHEN GTR rises AND real/total ≥ 70%
    GOVERNANCE IS CATCHING REAL PROBLEMS
  • WHEN GTR rises AND real/total < 70%
    NOISE IS RISING — TUNE THE RULES
  • WHEN GTR falls AND ratio steady
    AGENTS ARE LEARNING THE CONVENTIONS
  • WHEN GTR falls AND ratio drops
    GOVERNOR IS BURNED OUT — REVIEW COVERAGE
// GTR alone is not enough — always pair with the real/noise split
Compositions / signal trends

Governance Trigger Rate over time, with rising-signal interpretation: real problems vs. noise.

open ↗
№ 1383

SDL Diagnostic Explorer

Scene STA de
SDL · DIAGNOSTIC 10 weeks

Faster delivery.
But at what cost?

SDL · hours, spec → prod 18h (-4 wk)
W01W02W03W04W05W06W07W08W09W10
SDS security · 0-100
72 -16 over period
DRIFTING
AAR alignment · %
76 -16 over period
DROPPING
CSG violations · count
15 +13 over period
RISING
DIAGNOSTIC READING brittle weeks: 2

// reading

SDL is dropping 75% faster than the team's CSG-violation rate is being absorbed. AAR has fallen 16 points. SDS has fallen 16 points. The pipeline is faster because validation is being skipped, not because the system got better.

// action

Pin SDL at current value. Stop celebrating the speed-up. Raise AAR back to ≥85 and CSG violations back to ≤4 before allowing SDL to drop further. Speed without guardrails is not speed.

// SDL is diagnostic — never a target on its own
Compositions / signal trends

Spec-to-Delivery Latency as a diagnostic — when faster pipelines mean validation is being cut.

open ↗
№ 1384

DDL Undefined State Explainer

Scene STA go
DDL · UNDEFINED STATE

"No drift in 90 days" is not DDL = 0.
It is DDL = undefined.

A signal can only be measured when an event exists. Rendering an undefined signal as a numeric zero is a lie about how good things are.

WRONG false zero
last 90 days · DDL
0.0 hours
PERFECT — INFINITELY FAST

This reading says drift is being caught the moment it happens. It isn't. There has been no drift to catch.

RIGHT honest state
last 90 days · DDL
undefined
NO DRIFT TO MEASURE

Last drift 2026-02-18, detected in 4.2h. 92 days clean. The signal is dormant because the system is healthy.

92 DRIFT-FREE DAYS 2026-02-18 → 2026-05-21
last drift · 4.2h
now 92 days · no events · DDL undefined

// rule

A latency signal needs an event. With no event, latency has no measurable value. Render , not 0.

// when to alarm

Alarm when DDL becomes defined again — drift has reappeared. The transition matters, not the dormant value.

// what to report

"92 days drift-free, last DDL 4.2h" is true and useful. "DDL = 0" is false and harmful.

// honesty about what you cannot measure is part of the framework
Compositions / signal trends

Why "no drift in 90 days" is the highest DDL score — not zero. Explains the undefined state honestly.

open ↗
№ 1385

ICS Quarter Trend

Scene STA de
ICS · 13 WEEKS portfolio rolling mean

A quarter of investment in spec quality.

start mean54
end mean78
delta+24
weeks above gate10 / 13
ICS · MEAN + Q1-Q3 BAND y: 30→100
GATE 60 Q opens Lint rule shipped AAR program Q closes
W01W02W03W04W05W06W07W08W09W10W11W12W13
// ICS lifts when a team invests; it doesn't drift up on its own
Compositions / signal trends

Quarterly ICS rolling average with confidence band shading and gate-cross annotations.

open ↗
№ 1386

CCS Convention Quality

Scene STA co
CCS · 8 WEEKS 8 conventions

Conventions decay if you don't tend them.

aggregate

85%

code convention trend score delta verdict
  1. CONV-014 logical CSS only 68% -24 DECAYING
  2. CONV-045 tokens not hex 52% -20 DECAYING
  3. CONV-001 kebab-case files 94% -4 WATCH
  4. CONV-031 no barrel files 91% -4 WATCH
  5. CONV-058 conventional commits 99% +1 HEALTHY
  6. CONV-022 no any in TS 94% +6 HEALTHY
  7. CONV-067 spec frontmatter 92% +10 HEALTHY
  8. CONV-082 test colocated 89% +13 HEALTHY
// a decaying convention is a convention nobody is enforcing — fix the lint, not the score
Compositions / signal trends

Convention Coverage Score across the codebase — which conventions are healthy, which are decaying.

open ↗
№ 1387

SDS STRIDE Scorecard

Scene STA go
SDS · STRIDE 6 categories

Security documentation by threat class.

aggregate SDS

77

S Spoofing
score 88 ▲ +2
T Tampering
score 76 ▼ -3
R Repudiation
score 54 ▼ -1
◀ weakest link · prioritise this
I Info disclosure
score 82 ▲ +4
D Denial of service
score 68 ▲ +1
E Elevation of privilege
score 91 — 0
// an 80 aggregate hides a 54 on Repudiation — always read the breakdown
Compositions / signal trends

Security Documentation Score broken down by STRIDE category, with per-category trend arrows.

open ↗
№ 1388

CSG Violations Rising

Scene STA go
◆ ALERT · CSG VIOLATIONS RISING 8-week trend

14× the baseline.

violations are not a code problem — they are an un-promoted pattern.

baseline3/wk
current42/wk
multiplier14×
peak42/wk
TOP VIOLATING PATTERN 38 of 42

// pattern

await fetchWithRetry(url, { retries: 3 })

// repeated in

  • · src/services/billing.ts:124
  • · src/services/auth.ts:67
  • · src/services/events.ts:208
  • · …35 more sites
RECOMMENDED ACTION go pillar
  1. 1.Promote — open CONV-NNN: retry semantics for network calls.
  2. 2.Write rule — add no-bare-fetch to the linter with autofix.
  3. 3.Re-baseline — autofix all 38 sites, close the alert.
  4. est. effort · 1 day · @author + @gatekeeper
// rising CSG is rarely a bad team — usually a missing convention. promote, don't punish
Compositions / signal trends

CSG-violation count climbing — alert banner plus recommended action: promote the pattern to a convention.

open ↗
№ 1389

SHS Corpus Heatmap

Scene STA do
SHS · CORPUS 60 specs

Structural health of the spec corpus.

Healthy · 90-100 11 18%
OK · 70-89 31 52%
Watch · 50-69 12 20%
Critical · <50 6 10%
HEATMAP · 10 × 6 column-major by id
100 51
101 82
102 61
103 91
104 88
105 77
106 89
107 84
108 86
109 43
110 64
111 81
112 30
113 60
114 81
115 44
116 91
117 78
118 87
119 79
120 65
121 75
122 83
123 80
124 83
125 91
126 83
127 93
128 83
129 40
130 68
131 88
132 81
133 85
134 71
135 43
136 44
137 87
138 76
139 63
140 53
141 82
142 90
143 77
144 61
145 95
146 98
147 82
148 81
149 89
150 85
151 60
152 61
153 83
154 59
155 97
156 86
157 91
158 94
159 98
legend: 90-100 70-89 50-69 <50
worst SPEC-112 30
// clusters of red usually share a co-author — review specs by team, not by id
Compositions / signal trends

Spec Health Score heatmap — every spec by structural completeness, colour-coded by tier.

open ↗
№ 1390

SCR Dependency Radar

Scene STA go
SCR · 4-AXIS period vs prior

Supply chain has two soft axes.

composite SCR

80

-2 · WATCH

RADAR · 4-AXIS
FRSHADITLICNPROV
AXIS BREAKDOWN cur · prev · Δ
  • freshness 72 84 -12
  • audit 88 80 +8
  • license 94 92 +2
  • provenance 66 70 -4
WORST DEPENDENCIES · ACT ON THESE 4 flagged
  1. @org/legacy-codec@1.2.0 FRSH 28 upgrade to 3.x or replace
  2. crypto-shim@0.9.x PROV 32 no signed provenance — drop
  3. image-utils@4.0.1 FRSH 41 14 months stale · audit
  4. sso-bridge@2.1.0 PROV 44 sigstore key rotated — re-verify
// SCR composite of 80 with PROV at 66 is not "fine" — read the axes
Compositions / signal trends

Supply Chain Resilience trend — dependency vitality across freshness, audit, license, and provenance axes.

open ↗
№ 1391

Combined Signal Hero

Scene STA go
SIGNALS · NOW read together, never alone

Four signals.
One reading.

system state

WATCH

AAR drifting

AAR Agent Alignment Rate
value · weekly delta
83 % −4

agents are guessing more — spec ambiguity rising.

WATCH
GTR Governance Trigger Rate
value · weekly delta
14 /100 +3

rising with high real/total — governor is catching defects.

HEALTHY
SDL Spec→Delivery Latency
value · weekly delta
36 h −12

within 24-72h band — speed is healthy alongside other signals.

HEALTHY
DDL Drift Detection Latency
value · weekly delta
0d
no events

92 days drift-free — signal dormant, system healthy.

DORMANT
// one tile is a number. four tiles is a story
Compositions / signal trends

Full-bleed dashboard hero stacking AAR, GTR, SDL and DDL into one read-at-a-glance scene.

open ↗
№ 1392

Signal Interpretation Cheatsheet

Scene STA go
CHEATSHEET print me · pin above the desk

What does this trend mean?

kinds: leadingpredicts laggingmeasures diagnosticnever a target
signal ▲ rising ▼ falling — flat
  1. AAR Agent Alignment Rate leading
    specs landed, agents agree drift forming — review specs steady — no change to act on
  2. GTR Governance Trigger Rate leading
    ambiguous — read with real/total ratio either learning or burnout governance in steady-state
  3. SDL Spec→Delivery Latency diagnostic
    pipeline slowing — find the bottleneck only good with healthy AAR/SDS predictable cadence
  4. DDL Drift Detection Latency diagnostic
    detection getting slower detection getting faster undefined when no drift = healthy
  5. ICS Intent Completion Score lagging
    team investing in spec quality specs being rushed no investment, no decay
  6. CCS Convention Coverage Score lagging
    conventions taking hold conventions decaying — fix lint steady — neither growing nor shrinking
  7. SDS Security Documentation Score lagging
    threat model improving threats outpacing documentation audit the worst STRIDE cell
  8. CSG Convention Self-Governance leading
    pattern needs promotion to convention lint enforcing — agents adapting baseline holding
  9. SHS Spec Health Score lagging
    corpus structurally improving new specs rough — set a template depends on coverage
  10. SCR Supply Chain Resilience lagging
    deps audited and current staleness or provenance gap check axis breakdown anyway
// a trend without a reading is a wallpaper
Compositions / signal trends

What does this trend mean? — a printable cheatsheet pairing each direction with its real-world reading.

open ↗
№ 1393

Rising / Falling Flag Bank

Scene STA go
FLAGS · NOW 10 signals · one glance

Morning coffee. One look.

good4
bad4
neutral2
AAR

drift forming in 2 specs

GTR

real/total 78% — catching defects

SDL

within band but watch AAR

DDL

undefined · 92 days clean

ICS

+24 points this quarter

CCS

CONV-045 decaying fast

SDS

audit the R cell

CSG

42 violations — promote pattern

SHS

corpus mean +6 points

SCR

PROV axis lost 4 points

// if every flag is good, you're not measuring honestly
Compositions / signal trends

At-a-glance flag bank across all ten DoCoDeGo signals — direction arrow + tone + one-line reading.

open ↗
№ 1394

Threshold Violation Timeline

Scene STA go
THRESHOLD · TIMELINE Q1+Q2 2026

Six crossings. Six actions.

total6
resolved3
open2
accepted1
datesignalthresholdtriggeractionmttaoutcome
  1. 02-04 AAR < 85% SPEC-204 dropped to 82 spec rewritten · examples added 6h RESOLVED
  2. 02-22 CSG > 10/wk retry pattern emerged promoted to CONV-091 · lint shipped 2d RESOLVED
  3. 03-11 SDS < 70 R cell fell to 54 threat model session · still scheduled OPEN
  4. 03-29 SDL > 96h pipeline congested parallel review lanes opened 1d RESOLVED
  5. 04-15 SCR < 75 PROV axis fell to 66 accepted · sigstore migration Q3 4h ACCEPTED
  6. 05-02 AAR < 85% SPEC-227 dropped to 75 review scheduled · in progress OPEN
// every crossing is a story — owner: @governor
Compositions / signal trends

Timeline of moments a signal crossed a threshold — what triggered, what was done, what changed.

open ↗
№ 1395

Cross-Correlation Panel

Scene STA go
CROSS-CORRELATION AAR · CONV-UPDATES

Stop updating conventions.
AAR drops in two weeks.

pearson r

-0.81

strong negative

DUAL SERIES · 12 WEEKS left: AAR · right: conv/wk
2-WK LAG INFLECTION
W01W02W03W04W05W06W07W08W09W10W11W12
AAR% · left axis conv updatescount · right axis

// observation

Convention updates fell to 0 in W06. AAR began falling in W07 and bottomed in W10. Updates resumed in W09, AAR recovered by W12. The lag is consistently 1-2 weeks across the corpus.

// rule

Convention cadence is a leading indicator for AAR. Watch the bar chart, not the line — the line is already too late.

// pair signals when you want to see cause, not just effect
Compositions / signal trends

AAR plotted against convention update cadence — the leading-indicator pair that predicts spec health.

open ↗
№ 1396

Signal Hierarchy Diagram

Scene STA go
TAXONOMY three jobs · ten signals

Signals do three jobs.
Don't confuse them.

LEADING predicts
  • AAR agent alignment
  • GTR governance triggers
  • CSG convention self-gov

// does

predicts what will happen next. moves first.

// never

never a backward-looking grade.

DIAGNOSTIC explains
NEVER A TARGET
  • SDL spec→delivery latency
  • DDL drift detection latency

// does

explains a result. read together with leading + lagging.

// never

never a target. optimising it cuts validation.

LAGGING measures
  • ICS intent completion
  • CCS convention coverage
  • SDS security documentation
  • SHS spec health
  • SCR supply chain resilience

// does

measures what already happened. settles slowly.

// never

never the early-warning signal.

leading moves first diagnostic explains lagging confirms
Compositions / signal trends

Leading vs lagging vs diagnostic — the framework taxonomy of signals rendered as a layered diagram.

open ↗
Alpha · Honest about it

The framework is real.
The community is forming now.

DoCoDeGo is in Alpha. The framework is documented, the practices are battle-tested at small scale, and the next release is being shaped in public.

If it produces anything, it should produce engineers and teams who think more clearly about what they are building and why.

Two doors in
Join the Discord

Discord is where specs are debated, the framework gets sharper, and decisions land in writing. The conversation is the artefact.