YY is an independent AI forensic data analysis and research laboratory. We build inspectable software and run fixed-scope services for datasets, operational records, technical claims, and evidence-bearing workflows.

Our work starts from the actual constraints: the data, the machine, the failure modes, and what can be measured. We publish concrete artifacts rather than speculative capability claims, and we state where each conclusion stops.

Open the public Proof Desk to inspect reproducible case files and clearly labeled fictional or synthetic service demonstrations before buying software or hiring YY.

YY currently offers two fixed-scope paid pilots for operations and quality teams. Both require a fit check before payment; neither is broad consulting, legal advice, or a compliance opinion.

• Calibration Register Cleanup is a $149 pilot for one existing gage list and up to 25 supported calibration-certificate PDFs. YY returns a source-linked register and review queue within three business days after accepted intake and payment.
• Association Minutes is a $19 pilot for one authorized recording up to 60 minutes plus the meeting materials. YY returns a recording-checked DOCX and PDF draft by the end of the second business day after accepted intake and payment.

Looking for Dataset Preflight ($39), YY's automated dataset-integrity check? It is a product, not a service; see Paid products below or go straight to the listing.

Review exact scope, public samples, and fit-check instructions.

Dataset Preflight is a paid, local Python tool that turns a dataset directory into a deterministic review packet: canonical JSON, standalone HTML, and a SHA-256 manifest. It surfaces exact and narrowly normalized duplicates, optional provenance coverage, conservative secret-risk signals, and explicit CI thresholds without uploading the dataset or making runtime network requests.

• Product and exact scope
• Inspectable sample report
• Engineering note: inspect a training dataset before upload
• Buy the $39 founding release
• One organization, internal use, all 0.x updates

Dataset Preflight is not a security scanner, audit, certification, legal clearance, or training-readiness verdict.

AgentSafe is a paid, local Python tool for producing deterministic approval-to-execution evidence for OpenAI Agents SDK 0.19.x FunctionTool workflows. Its founding release checks six specific invariants and publishes an inspectable sample report before purchase.

• Product and measured scope
• Inspectable sample report
• Founding license: $49 once, one organization, internal use, all 0.x updates

AgentSafe is not a security audit or certification. Dynamic approval callbacks and other tool types are outside the founding release's measured scope.

YY IDE explores an agent-first development loop: the agent holds the implementation state while the person guides constraints, reviews evidence, and intervenes where judgment matters. The page is a clearly labeled working preview, not a download or availability claim.

Workload Receipt is an Apache-2.0 Python package for turning observed AI usage metadata into a deterministic, content-free receipt. It writes a canonical JSON summary, a Markdown report, and a SHA-256 manifest, keeps unobserved measurements explicit as evidence gaps, and verifies the artifacts later.

python -m pip install yy-workload-receipt

• Product page and synthetic sample
• PyPI package
• Source and v0.1.0 release

Workload Receipt is not a billing system, cost estimator, security scanner, compliance certification, or hardware recommendation.

Corpus Evidence is YY's first open-source release. It creates deterministic, local corpus manifests, records optional provenance declarations, and checks the result for drift later.

python -m pip install yy-corpus-evidence

• Project page
• PyPI package
• Source and documentation

ClipCheck is an Apache-2.0 command-line preflight for short-form video. It runs local ffprobe/ffmpeg checks for media properties such as duration, displayed aspect ratio, audio sample rate, loudness, true peak, and caption-sidecar presence, and can emit JSON evidence for CI.

• Project page and measured scope
• Source and v0.1.0 release

• Deterministic before convenient.
• Local and inspectable by default.
• Explicit about unknowns and limitations.
• Claims backed by reproducible artifacts.

Learn more at yycore.ai.