CHECKING FOR UPDATESAutomatic check every 5 seconds
The Spark DadaTracker icon
THE SPARK · EVIDENCE CONTROL PLANE

The Spark DadaTracker

v1.0.0 PROTOTYPEPRODUCTION

Evidence-backed tracking, roadmaps, progress, migration, installations, questionnaires, release gates, and artifacts for The Spark and F.O.U.R.

Production tracker is live · roadmap sources are connected
The Spark33%Current evidence-based portfolio estimate
F.O.U.R. parent weight90%36.891875% exact official model
InstaDater parent weight10%Top-level lifecycle service
Tracker stateLIVEPlanning, evidence, and roadmap views
All 25 tracker sections
04 · Current Forgey AIO

Forgey AIO — current training, goals, models, and lifecycle

Current Forgey AIO evidence captured 7/22/2026, 4:25:39 AM. Foundry and wider system percentages are intentionally excluded from this page.

1,400Dataset v3 records
5 / 5Expert adapters passed
noneCurrent evaluation expert
PASSCurrent training state
CURRENT Dataset v3
1,400

1101 train · 178 validation · 121 test

CURRENT HELD-OUT EVALUATION

No active expert

Evaluation state: STOPPED_UNEXPECTEDLY

Training state: PASS

Last verified 7/22/2026, 4:25:39 AM

Current expert training

ExpertRoleStatusRecordsValidation loss
Forgey CoreIdentity, safety, honesty, uncertainty and runtime truthPASS
222 / 222 records100%
1.5559
Forgey CodingProgramming, debugging, PowerShell and testsPASS
212 / 212 records100%
2.5375
Forgey FoundryComponents, lifecycle, evidence and provenancePASS
268 / 268 records100%
2.7923
Forgey PlanningSequencing, estimates, pause points and recoveryPASS
277 / 277 records100%
3.1176
Forgey VisionImages, screenshots, diagrams and visual groundingPASS
122 / 122 records100%
2.7153

Current Forgey AIO goals

GoalWeightProgressPurpose
P0 · Safety, ownership and approvals14%
Safety, ownership and approvals progress55%
Hard rules, evidence, privacy, backup and rollback
P1 · Reliable core intelligence13%
Reliable core intelligence progress38%
Reasoning, honesty, instruction following and evaluations
P2 · Memory and personalization9%
Memory and personalization progress12%
Separated, reviewable and consent-aware memory
P3 · Computer and coding tools10%
Computer and coding tools progress32%
Commands, files, Git, tests and app building
P4 · Vision and media8%
Vision and media progress34%
Images, interfaces, diagrams, video and design
P5 · Voice and natural interaction4%
Voice and natural interaction progress5%
Voice, approved speakers, calls and continuity
P6 · Desktop and unified interface7%
Desktop and unified interface progress18%
Visible control center, tools and approvals
P7 · Forgey Utilities5%
Forgey Utilities progress6%
Clock, timer, alarms, notes, weather and more
P8 · Ecosystem and watchers9%
Ecosystem and watchers progress24%
Foundry, Global, Master Engine and profiles
P9 · Internet, cloud and communications4%
Internet, cloud and communications progress16%
Sources, storage, email and messaging
P10 · Automation and organization5%
Automation and organization progress22%
Crash-safe recurring work and planning
P11 · High-risk domains2%
High-risk domains progress3%
Finance, health, law and security safeguards
P12 · Training and model development8%
Training and model development progress62%
Curated data, QLoRA, evaluation and model growth
P13 · F.O.U.R. ecosystem2%
F.O.U.R. ecosystem progress20%
Shared architecture across product families

Current model ladder

Forgey 7B
68%
Forgey 7B progress68%

Target 88%

Held-out evaluation incomplete; standalone candidate is not released

Evidence-backed progress formula
  • PASS 16% — Governance and frozen Dataset v3
  • PASS 12% — Verified base model and reproducible environment
  • PASS 28% — Five expert-adapter training runs
  • PASS 7% — Five adapter reload-integrity checks
  • PASS 5% — Standalone candidate package and hash verification
  • PENDING 12% — Held-out evaluation for every expert
  • PENDING 10% — F.O.U.R. and Forgey Foundry integration proof
  • PENDING 10% — Release, signing, install and rollback acceptance
Forgey 32B
12%
Forgey 32B progress12%

Target 25%

Unified native multimodal plan; 14B coder is teacher only

Future larger AIO
4%
Future larger AIO progress4%

Target 10%

Requires stronger base, data, hardware and repeated gates

The Forgey AIO lifecycle flowchart is now grouped with every other visual in the Flowcharts section.
Current release boundaryCandidate adapter package exists; do not release, sign, merge, quantize, convert to GGUF, or register it until every held-out evaluation and release gate passes.
Current evidence sources
  • D:\05_Forgey_AI\05_Training\05_Runs\forgey-7b-expert-training-v3-20260720\result.json
  • D:\05_Forgey_AI\05_Training\01_Datasets\03_Approved\Frozen\forgey-dataset-v3\manifest.json
  • D:\05_Forgey_AI\05_Training\06_Reports\0027_Forgey_AIO_Fully_Expanded_AI_Lifecycle_Flowchart_2026-07-21.md
  • D:\05_Forgey_AI\05_Training\10_Packages\Forgey_7B\Forgey_7B_v3_Standalone_Candidate_20260722T041400Z-build-report.json
Family · Weight 25%

Forgey AIO

AI training, datasets, expert adapters, evaluation, and future desktop app.

CONNECTED!CONNECTED! · v1.0.0 · automaticCURRENT STATUS · MANUAL
Blue Current 41.5%Green 38%
42%BLOCKED
Current Status42%Manually maintained from accepted evidence · provisional · 62% confidence
Roadmap connectionCONNECTED!1 file(s) registered automatically
Roadmap registrationAutomatic recursive registration
Roadmap sourceArtifacts/Current/Approved_Roadmaps/Batch_05/0018_FOUR_Forgey_AIO_Roadmap_v1.0.0.txt
Current-status ownershipManual · Codex / ChatGPT / Doonchy evidence review
Evidenceprovisional · 62% confidence
Official weight25% of F.O.U.R.
Exact Blue Current41.5%
Rounded display42%
Contribution to parent10.375 percentage points
Current-status sourceStep 5 evidence matrix and official portfolio reconciliation v1.0.0
Version truthUnreleased / evidence snapshot
Last verified7/21/2026, 7:31:55 AM
Progress formulaweight × Blue Current
Next actionReview current status when accepted evidence changes
Superseded weight18% · preserved as history
Superseded Blue28% · replaced by Step 5 evidence
The connected Batch 05 roadmap preserves training goals, datasets, adapters, experts, model ladder, lifecycle, SQLite3 memory, and local-model work. No source or model blobs were moved.