← Kembali ke katalog AI Lab

Catatan sistem dan pengujian

Dokumentasi lengkap tujuh sistem Lab. Status dan tanggal mengikuti catatan pengujian yang tersedia.

Lab index

7 systems. One evidence standard.

These projects connect scholarship with working digital infrastructure. Their status will change only when testing and verifiable results justify it.

LAB007

Educational video

Hukumpreneur Studio — Video Clips

Turn a complete idea into a shareable clip.

Prototype

Problem

The question or operational constraint being addressed

Lectures and academic forums need short excerpts that preserve context.

System

The product, workflow, or capability that was built

A browser workspace for trimming local videos, composing vertical clips, adding SRT captions, and exporting video and covers.

Technology & method

The technical and analytical approach

Browser canvas, Web Audio and MediaRecorder; local templates and project settings.

Testing

How usefulness, safety, or performance is assessed

Export format depends on browser support. Keep the tab active during rendering.

Verifiable result

What a visitor can inspect or confirm

Local video editing is available. Automatic YouTube ingestion, AI highlights and cloud storage are not connected.

Experiment record

Last tested
19 October 2026
Current metric
Up to 180 seconds per exported clip.
What changed
Added BIAF, RAHRN, Hossibarani and law lecture templates.
Evidence / live system
Open Hukumpreneur Studio
Experiment notes
Review the excerpt and captions before sharing. Download results before closing the page.

Human review remains part of the system.

LAB001

Legal AI evaluation

LEX-EVAL™ Legal AI Benchmark

Legal answers need evidence, not confidence.

Pilot

Problem

The question or operational constraint being addressed

Legal AI can produce fluent answers while missing authority, doctrine, jurisdiction, or the limits of the available evidence.

System

The product, workflow, or capability that was built

A structured benchmark for comparing AI-generated legal answers through a controlled corpus, scoring rubric, evaluator workflow, and documented outputs.

Technology & method

The technical and analytical approach

Multi-model evaluation, prompt controls, answer capture, doctrinal scoring, citation review, and empirical comparison across runs.

Testing

How usefulness, safety, or performance is assessed

Pilot Round E01–E05 defines five questions for three models in one run each. See the experiment record for the reporting status of these 15 outputs.

Verifiable result

What a visitor can inspect or confirm

The benchmark framework and public sample can be opened and reviewed. Empirical validation remains in the pilot stage.

Experiment record

Last tested
05 September 2026
Current metric
Pilot protocol: 5 questions × 3 models × 1 run = 15 planned outputs.
What changed
The benchmark moved from a conceptual framework to a smaller empirical pilot before the full 180-output study.
Evidence / live system
Review LEX-EVAL™
Experiment notes
Scores will be reported only after the 15 outputs have been captured, reviewed, and checked for evaluator consistency.

Human review remains part of the system.

LAB002

Legal research system

LexNusa Legal Research Assistant

Legal research should remain traceable to authority.

Prototype

Problem

The question or operational constraint being addressed

Researchers and legal professionals need faster issue mapping and synthesis without surrendering verification, confidentiality, or professional judgment.

System

The product, workflow, or capability that was built

A human-supervised legal AI service for research, document analysis, legal evaluation, and structured pilot engagements.

Technology & method

The technical and analytical approach

Retrieval-guided workflows, structured prompts, source verification, human review, secure intake, and proposal-to-delivery operations.

Testing

How usefulness, safety, or performance is assessed

Readiness is assessed through LEX-EVAL™, sample deliverables, controlled pilot requests, and explicit review checkpoints.

Verifiable result

What a visitor can inspect or confirm

The service architecture, evidence standard, packages, pilot pathway, and sample benchmark are publicly documented.

Experiment record

Last tested
05 September 2026
Current metric
Public evidence pathway available; client pilot performance baseline not yet published.
What changed
The project expanded from a research-assistant concept into a supervised pilot service with evidence and delivery checkpoints.
Evidence / live system
Inspect the evidence standard
Experiment notes
The next meaningful measure is not traffic, but whether pilot outputs remain traceable to authoritative legal sources.

Human review remains part of the system.

LAB003

Education operations

SIMAKS School Accreditation System

Accreditation work needs one accountable workspace.

Pilot

Problem

The question or operational constraint being addressed

School accreditation evidence is often dispersed across people, files, indicators, and review cycles, making readiness difficult to monitor.

System

The product, workflow, or capability that was built

A role-based workspace where school leaders can claim a school, invite accreditation teams, manage evidence, and coordinate verification by supervisors or education authorities.

Technology & method

The technical and analytical approach

Next.js, Supabase authentication and database services, role-based access, Row Level Security, team invitations, and evidence workflows.

Testing

How usefulness, safety, or performance is assessed

Access is tested by role: participants see authorised school data, while verification and administrative views remain restricted to approved users.

Verifiable result

What a visitor can inspect or confirm

A public product entry point and account flow are available. Institutional onboarding and role verification remain pilot activities.

Experiment record

Last tested
05 September 2026
Current metric
Four access roles defined: school leader, accreditation team, supervisor, and education authority.
What changed
Public self-registration was extended with role selection, school claims, team invitations, and external verification boundaries.
Evidence / live system
Open SIMAKS
Experiment notes
Pilot validation must confirm that each role can complete its work without seeing records outside its authorised school scope.

Human review remains part of the system.

LAB004

Digital learning

Raja Ali Haji Learning Platform

A classical intellectual tradition, made learnable.

Production

Problem

The question or operational constraint being addressed

Raja Ali Haji's ideas are intellectually significant but remain difficult for many learners to access through a coherent, guided pathway.

System

The product, workflow, or capability that was built

An eight-module learning experience combining readings, podcasts, reflection, essays, quizzes, saved progress, and verifiable certification.

Technology & method

The technical and analytical approach

Next.js learning routes, participant authentication, Supabase-backed progress, assessment workflows, certificate validation, and cohort administration.

Testing

How usefulness, safety, or performance is assessed

Completion requires all mandatory module activities, a minimum quiz score of 70 in each module, and a final reflection in Module 8.

Verifiable result

What a visitor can inspect or confirm

All eight modules are publicly available, while participant progress and certification operate through authenticated accounts.

Experiment record

Last tested
05 September 2026
Current metric
8 learning modules available; certificate threshold requires ≥70 on every required module quiz.
What changed
The programme progressed from public learning content to authenticated progress, assessment, reflection, and certificate validation.
Evidence / live system
Review the learning programme
Experiment notes
Learning impact will be evaluated through baseline–endline change, completion, quiz results, reflection quality, and participant feedback.

Human review remains part of the system.

LAB005

Agent architecture

AI Agent Hossibarani Digital Ecosystem

One ecosystem needs coordinated intelligence and clear limits.

Prototype

Problem

The question or operational constraint being addressed

A growing network of websites, research programmes, learning platforms, legal technology, and institutional initiatives creates fragmented operational knowledge.

System

The product, workflow, or capability that was built

A control-centre architecture for specialised agents that can inspect, diagnose, score, and recommend improvements across the ecosystem.

Technology & method

The technical and analytical approach

Role-separated agents, restricted control-centre access, audit trails, structured diagnostics, and integrations with development and analytics services.

Testing

How usefulness, safety, or performance is assessed

The current phase is deliberately read-only: agents may audit GitHub, Vercel, website health, and SEO but may not commit, merge, deploy, or alter production automatically.

Verifiable result

What a visitor can inspect or confirm

The access boundary and control-centre foundation are established. Autonomous production actions remain disabled by design.

Experiment record

Last tested
05 September 2026
Current metric
Read-only operating boundary active; 0 autonomous commit, merge, or production deployment permissions.
What changed
The concept was narrowed into a governed control centre with specialised diagnostics and explicit production restrictions.
Evidence / live system
View the control centre
Experiment notes
Permission expansion will be considered only after audit quality, traceability, and safe human approval paths are demonstrated.

Human review remains part of the system.

LAB006

Programme automation

Automated Pilot Cohort Management

Learning support should respond before participants fall behind.

Pilot

Problem

The question or operational constraint being addressed

A four-week learning cohort requires consistent reminders, progress follow-up, certificate eligibility checks, and credible final reporting.

System

The product, workflow, or capability that was built

An operational layer for weekly communication, lagging-participant reminders, completion validation, evaluation, and executive cohort reporting.

Technology & method

The technical and analytical approach

Scheduled workflows, progress rules, communication templates, assessment data, certificate logic, administrative dashboards, and PDF reporting.

Testing

How usefulness, safety, or performance is assessed

The pilot tracks module completion, quiz thresholds, reflection submission, weekly progress, baseline–endline change, and participant evaluation.

Verifiable result

What a visitor can inspect or confirm

Participant and administrator workflows are implemented for the first Raja Ali Haji Institute pilot cohort and await live cohort validation.

Experiment record

Last tested
05 September 2026
Current metric
4-week cohort workflow with weekly monitoring and certificate eligibility checks prepared for pilot validation.
What changed
The learning platform gained weekly messages, lagging-participant reminders, eligibility rules, evaluation, and executive reporting.
Evidence / live system
See the pilot programme
Experiment notes
The first live cohort will establish delivery, engagement, completion, reminder-response, and baseline–endline benchmarks.

Human review remains part of the system.