A human-directed operating layer for reliable AI-assisted work across models, tools, devices and long-running projects.

Korako is being built to reduce human-postman work between otherwise capable systems while keeping consequential authority with the human.

Goal → Plan → Authority → Execution → Evidence → Verification → Continuation

Mira is the conversational interface. A private orchestration core coordinates the underlying work. Korako Yolawani is the public product and the human journey around both.

Status date: 2026-09-10
Stage: active prototype / Golden Demo validation
Production-ready: no
Canonical implementation: private

The current focus is proving one useful Mira journey end to end rather than expanding feature count.

Latest internally verified progress

• the current private Golden Demo line passes 413 automated tests with 0 failures and 1 environment-dependent skip;
• TypeScript validation and a production Webpack build pass;
• the production runtime responds successfully on the local application root;
• deterministic calculator, current weather, Slovenian cinema schedule lookup, Slovenia job search and Tirol job search have passed live runtime smoke checks;
• consequential email sending still stops at a Human Gate and remains preview-only without approval;
• local/free-first speech-to-text is available in the current development runtime.

Still in validation

• natural voice completion and browser/TTS acceptance;
• live LPP departure retrieval under the bounded runtime path;
• repeated end-to-end Mira voice → useful action → independent verification evidence;
• external-user and third-party reproduction evidence.

Next gate: browser/voice acceptance → canonical product merge → external tester evidence → sanitized public proof update.

Follow PROOF-001 — Voice → Verified Action →

Explore the public Korako repository →

Korako comes from the Slovenian idea of korak / koraki — a step / steps. The product is built around helping a person move through complex work one meaningful step at a time while preserving context, direction and control.

Yolawani represents the wider human journey around those steps — the goals, decisions, projects and experiences that give each step meaning. It is used as the distinctive second part of the brand rather than as a literal translation from another language.

A life is not one task. It is a journey made of steps. Korako helps the human move through those steps without losing the journey.

I deliberately distinguish implemented, tested, verified, in validation, experimental and planned work rather than presenting concept-stage ideas as finished products.

• Human-directed: consequential decisions remain human-owned.
• Evidence-first: PREPARED != EXECUTED != VERIFIED.
• AI-native: AI is used heavily for research, implementation assistance, critique, documentation and testing.
• Multi-tool: models and tools are capabilities that can be routed rather than the identity of the product.
• Recovery-aware: permissions, reversibility, checkpoints and failure recovery are designed in.
• Free/local-first: local or already-owned capability is preferred where it is sufficient; paid escalation should never happen silently.

My focus is the intersection of product vision, systems decomposition, AI-native workflows, acceptance criteria, prototyping, testing direction and orchestration.

The public Korako GitHub is intentionally a diligence surface, not a source dump.

Public: product thesis, high-level architecture, sanitized evidence contracts, reproducibility targets, limitations and review protocol.
Private: canonical runtime, credentials, machine configuration, security-sensitive boundaries, operational logs and unpublished IP-sensitive material.

• a technical co-founder / lead engineer for Korako Yolawani;
• independent reliability, security and architecture review;
• pilot partners with measurable multi-step workflows;
• AI/product collaboration;
• early-stage mentor and investor conversations grounded in evidence.