AN INDEPENDENT WORKSHOP
I’m Logan Rourke. I set up phones and laptops for maps and reading without a signal, and build small AI workflows with checks you can see.
Practical tools, made for your task. Offline device setups · AI workflows with human review.
SELECTED PROJECTS
Different interests. The same instinct:
make something useful, then make it better.
OFFLINE & RESILIENCE
A pocket phone and a Linux Mint laptop, put back to work. Field Library is growing into a searchable collection of saved references; its portable release is still being checked.
Explore the offline kit
ART & CREATIVE TOOLS
sillypsyben is my art brand: odd characters, their own little world, and tools that help turn drawings into things you can hold.
Explore sillypsybenShop
SYSTEMS & AUTOMATION
A dashboard for service health and project context. ORSON’s manual pilot is installed. The newer web-panel upgrade passed 145 checks and a no-write dry run; installation and device checks are still ahead.
Inside the ORSON buildLET’S MAKE SOMETHING USEFUL
Small, practical builds.
A clear idea of what “working” means.
PAID OFFLINE DEVICE SETUPS
For people preparing for patchy coverage, travel, or time away from the internet.
Put a compatible phone or laptop you already own to work with downloaded maps and an agreed reading library. You get a connection-off check and simple notes for using and updating the setup.
Start here: Send your device model, region, and the references you want. I’ll check what it can support and suggest a scope.
VERIFIED AI WORKFLOWS
For independent creators and small teams spending time on the same task each week.
A small workflow that produces a draft, checks agreed rules, and sends the result to you for review. You get the workflow, sample inputs, the checks, and handoff notes.
Examples we can scope together:
Start here: Send one example input and the result you need. We’ll choose a useful first task and define what a passing result looks like.
How I work: Agree a small scope and a pass/fail check, build it, verify it, then hand it over with notes.
Before work starts, we agree the price, delivery, where the tools will run, any account or AI usage costs, and the support included. AI output stays subject to human review.
Tell me what you want to make easier. One concrete example is a good place to start.
Email meONGOING EXPLORATIONS
More threads I’m following:
better research, better tools, stranger worlds.
PERSONAL TOOLS & EXPERIMENTS
Choose useful work, prepare the task, and send it down the right path. Keep the lessons for the next time.
SAVED CONTEXT
Reviewed notes, sources, decisions, and corrections: context for ORSON and the research memory preflight.
Use memory, research, and lab status from the Home Lab Control Plane to suggest tasks.
I decide what is worth pursuing and what a useful result should look like.
Shape the prompt, requirements, and checks, then route the task to the appropriate workflow.
From Prompt OS · choose a path
ENGINEERING
Propose code and run separate checks before review.
Output: a checked code candidate.
RESEARCH
Reuse reviewed work, refresh it, or research the gaps.
Output: findings with their sources.
MEDIA · PLANNED
Plan storyboards and prepare scene prompts and reference images. Route approved shots to suitable generation tools, prioritizing free and already-included options.
Planned output: prepared scenes and generated media for review.
ALL PATHS · RESULTS HELD FOR REVIEW
Check the output and its evidence. Revise, reject, or approve it. A generated result does not approve itself.
Only the output I approve moves forward.
I choose which lessons and corrections to promote into reviewed notes for future tasks.
Recorded on : a preflight read reviewed memory and requested a refresh because a claim still needed checking. Read the example and its limits ↗
Muse now has a separate read-only connection to two approved project-context documents. It is a small starting point, not an automatic handoff across the lab. Read the Muse connection note.
For paid work, I offer offline device setups and small AI workflows with human review.
NOTES, SOURCES & CORRECTIONS
Keeping sources, decisions, and lessons together, so I can pick up a project without starting from scratch.
In development · Recorded examplesRESEARCH & PLANNING
Prompt OS helps prepare tasks and research briefs. Research OS keeps the sources and brings saved work into the review. The connected handoff is still being developed.
Manual use · Integration in developmentRECORDED IN JULY 2026
A handoff stopped at a required status check. A later automated replay found seven filename mismatches against its evidence list.
Read the incident and its limitsAI-ASSISTED ENGINEERING
The Agentic Engineering Operating System: AI workers propose code, and separate checks decide whether it can move to review.
Development & testingSCRIPTED TEST · 10 SEPTEMBER 2026
Two successful attempts reached the review stage. All seven intended negative attempts were kept from acceptance, including a worker’s incorrect claim that its output was clean.
Read the test and its limitsAI VIDEO EXPERIMENTS
Short AI-generated videos exploring impossible places and strange phenomena.
Creative exploration
THE PERSON AT THE WORKBENCH
I’m Logan Rourke. In automotive operations, I appraise vehicles and improve day-to-day workflows. That work shapes how I build: understand the task, find the friction, and check that the change helps.
I make art as sillypsyben, and Rourke Labs is the workshop behind it. Today I build creative tools, small automations, and useful setups for older devices.
The tinkering started with jailbreaking my iPod Touch in middle school, followed by console-game modding in high school. I’ve also repaired jewelry as an apprentice goldsmith. Different workbenches, the same enjoyment of making things work.
GOOD THINGS START WITH A CONVERSATION
Tell me the task, the tools you use, and what a useful result would look like.
I’m also interested in collaborating on offline reference tools, creative workflows, and ways to check AI-generated work.
Working in U.S. Eastern time.
FROM THE WORKBENCH
What worked, what broke,
and what the checks actually showed.
THE WORKSHOP · PUBLISHED
A new business milestone, small practical projects, and a clear scope before work starts.
FIELD JACKET · PUBLISHED · UPDATED
The visual direction in a short AI-generated workshop clip. Fit, construction and a physical sample remain ahead.
AI COLLABORATION · PUBLISHED · UPDATED
Persistent context serving and a confirmed readback. The daily status document and ongoing freshness checks remain ahead.
ORSON · PUBLISHED · UPDATED
The pilot is installed. The separate web-panel upgrade passed 145 checks and a no-write dry run.
FIELD LIBRARY · PUBLISHED · UPDATED
An eight-document Mint demo, a real selection screen, and the checks still ahead for a customer release.
WORKFLOW CHECKS · PUBLISHED · UPDATED
Recorded test results, a fresh document review, and the difference between a fix and its adoption.
LOCAL AI · PUBLISHED · UPDATED
A supervised Goose/Qwen conversation: three replies with consistent supplied context and successful cleanup. Live tool integration remains unverified.
FAILURE MECHANISMS · PUBLISHED · UPDATED
Three lessons about file ownership, backup signals, and testing an alarm.
