Built and operating
Habitat
A personal AI operating environment where tools, agents, memory, approvals, logs, backups, and recovery paths work as one governed system.
Labs / AI as the newest chapter
The same questions now move into AI.
Labs are where I test product beliefs under current AI pressure: agency, evidence, boundaries, memory, judgment, and human control. Built systems, working prototypes, and concepts are deliberately shown at different weight.
Maturity, made visible
Three labs. Three different levels of proof.
Working prototype
KIREI
A photo-to-evidence product scanner that resolves identity, looks for ingredient sources, exposes uncertainty, and provides personal-fit context.
Concept / early workflow
JAVIS
A job-intelligence direction focused on role evidence, market signals, effort decisions, and human agency rather than application volume.
01 / Built and operating
Habitat
Most AI tooling gives an assistant more reach without making boundaries, approvals, memory, logs, or recovery equally visible. Habitat treats governance as the product.
Local agent environment, persistent memory and skills, scheduled routines, watchdogs, private context indexes, and explicit approval habits.
How useful autonomy changes when work is observable, external actions are gated, and recovery is designed before failure.
01Human authority
02Permission gates
03Bounded agent work
04Observable evidence
05Recovery path
Hard approvalExternal sendPrivate contextIrreversible change
Operating evidenceProposeAct within scopeLog and recover
02 / Working prototype
KIREI
Beauty and household products are hard to judge when packaging, barcodes, ingredient lists, claims, local availability, and personal context are disconnected.
Photo upload or camera input, OCR and barcode extraction, identity resolution, ingredient-source fallbacks, evidence provenance, and personal-fit context.
When identity or ingredients cannot be supported, the prototype asks for a clearer back-label image or manual ingredients instead of inventing a rating.
Prototype only. Evidence coverage and accuracy are still being improved. It is not medical guidance or a validated consumer product.
01Photo or barcode
02Product identity
03Source evidence
04Ingredient context
05Human decision
OCR + barcodeIdentity confidenceSource provenanceVisible uncertainty
03 / Concept direction
JAVIS should improve career judgment, not application volume.
The early direction connects role discovery, market signals, company context, effort decisions, and reflection. It remains a concept until a working workflow can prove that the evidence actually changes a person's decision.
Guardrail
AI is the newest chapter, not the origin story.
The labs are credible because they continue the professional work: trust and delegation, signal-to-direction synthesis, decision-load reduction, and leadership systems.