/agents
AI Agent Systems
Orchestration, sub-agents and long-running workflows that stay in control.
YACHENYI / BUILD
INDEPENDENT AI SOFTWARE STUDIO
Build smarter software.
Make every step count.
From agent orchestration to everyday automation, I turn complex workflows into focused, reliable software — built to be understood, improved, and used.
One independent builder · AI-native engineering · Practical by design
YOUR WORK, MADE VISIBLE
Build with clarity.
Project lifecycle↗
01Define
02Plan
03Build
04Verify
Goal → Evidence → Delivery↗
FOCUS AREAS
03domains
Agents · Automation · Web
Selected projectsVIEW ALL ↗
DDeepLoopMinderMulti-agent runtimeACTIVE
IIssuePackDeveloper workflowTOOL
AAutoScreenAgentAutomationLAB
A conceptual interface illustrating the workflow — not live operational metrics.
WHAT I BUILDPractical systems.
Practical systems.
Thoughtfully engineered.
A small studio with a focused toolkit — combining AI, automation and software engineering to solve the problems behind the interface.
/automation
Workflow Automation
Replace repetitive steps with clear, observable, reliable pipelines.
YACHENYI / BUILD
/web
Web Systems
Fast sites, internal tools and interfaces that serve real work.
YACHENYI / BUILD
/context
Context Infrastructure
Make requirements, decisions and evidence reusable across tools.
YACHENYI / BUILD
/governance
Project Governance
Turn vague goals into explicit plans, verification and state.
YACHENYI / BUILD
/delivery
Software Delivery
Build, test, document and iterate with maintainability in mind.
YACHENYI / BUILD
SELECTED WORKBuilt in the open.
Built in the open.
Made for the real world.
Projects that explore how agents, human decisions and everyday software can work better together.
All projects
01DeepLoopMinder
An AI-native multi-agent software development runtime focused on project control, durable state and external coding workers.
02IssuePack
A local-first context packaging tool that turns messy customer chats, screenshots and attachments into traceable issue packages for coding agents.
03AutoScreenAgent
An experimental Android agent that turns natural-language goals into screen actions through accessibility APIs and LLM tool calling.
HOW I WORKFrom an idea
From an idea
to a dependable system.
No unnecessary machinery. Just a clear problem, the smallest useful implementation, and proof that it works.
01 —
Understand the problem
Clarify the goal, constraints and what success actually looks like.
$ project define
✓ goal & scope
✓ success criteria
✓ goal & scope
✓ success criteria
02 —
Build the smallest useful thing
Prototype with simple components and make important state visible.
$ workflow run
→ plan → implement
✓ checkpoint saved
→ plan → implement
✓ checkpoint saved
03 —
Verify and keep improving
Use evidence, tests and feedback to make the result maintainable.
$ project verify
✓ tests completed
✓ evidence attached
✓ tests completed
✓ evidence attached
FROM THE LABNotes from
Notes from
the workbench.
Architecture decisions, experiments and notes from building AI-native software.
Read the lab01 / BUILD LOG
2026-09-20
Why this studio site is static-first
The public Yachenyi site treats crawlability, speed and machine-readable structure as architecture constraints rather than launch-time polish.
02 / BUILD LOG
2026-09-19
Agent workflows need durable project state
Long-running software work becomes more reliable when agents operate on explicit projects, tasks and artifacts instead of a single disappearing chat context.
LET'S COLLABORATEHave a problem worth solving?
Have a problem worth solving?
Let's build something useful.
A focused project starts with a good question. Tell me what isn't working, and we'll find the smallest reliable way forward.
Start a conversation