What is the one job you wish you could hand over? Let’s start there: one task, one deployment, one handoff you can use.
A proposed remote setup service using Hermes Agent, the open-source software from Nous Research. All Light’s contribution is configuration, a working integration, testing, and clear instructions.
$500 proposed one-time setup
Pilot in development. Not taking payments. Delivery readiness and customer demand are not yet validated.
A specific handoff
One compatible computer and one customer-owned assistant profile.
One approved model connection, with a usage limit.
One agreed workflow, demonstrated on the customer’s example.
Restart, cancellation, failure, and access-revocation checks.
Plain-language instructions and a proposed seven-day setup-defect correction window.
Separate costs: hardware, model/API usage, third-party subscriptions, ongoing support, and additional workflows.
Remote access is authorized and revocable. The customer signs in to their own accounts; credentials are not handed over in chat.
Your example, your limits, and a shared picture of what useful means.
02
Build the first version
Something concrete to use, shaped around that one job.
03
Check it works
Test the happy path, the failures, and the evidence behind the answer.
04
Try it together
You confirm it helps. We address the gaps without handing you busywork.
05
Make it yours
A working handoff, clear instructions, and an agreed next step.
WHAT WE’RE BUILDINGV1 · In development
The first Garden
Less passing context. More making progress.
Garden is our agent-workspace project: one place to define a task, coordinate model workers, inspect what they produce, and keep the decisions that matter.
The first workflow we’re testing
01
A clear brief
The task, its limits, and what a good result looks like.
02
A bounded worker
The right model proposes an output within those limits.
03
A separate check
Evidence is checked. Uncertainty stays visible.
04
A durable record
The work can be inspected and picked up again.
Starting small, on purpose. The first proof is a local-model proposal and a replayable handoff. Garden is not yet a released product or a replacement for your existing AI subscriptions.
VERIFICATION NOTES
A clear status. A useful next step.
WORK EXCHANGE · ILLUSTRATIVE EXAMPLE
A handoff without erasing the earlier idea.
Imagine two versions of a project brief. Importing the newer one should not quietly replace the older one. The local tool retains their separate records and includes the original bytes in its export.
KeepBoth source versions
CheckFile integrity on export
Carry forwardFiles and their provenance
This explains the tested local workflow. It is not a live demonstration or an automatic connection to another AI account.
What these checks mean
Work Exchange’s 33 passing tests cover the local import/export implementation. Garden’s recorded draft is a dated integration proof. CourtsidePass’s page response shows local availability, not verified sports data. None establishes customer demand or a finished commercial product.
What comes before selling
A defined buyer and problem, a usable demonstration, verified delivery, and a measurable pilot. For Hexclaw, we first reproduce the setup on our own equipment and count delivery time, costs, and support work.
Our first proposed creator experiment starts with the founder’s own comedy footage: select complete moments, preserve the timing, and create clear vertical edits.
We’ll measure editing time and cost first, then audience retention and actual enquiries or earnings after an approved release. Creator partnerships and engagement gains are not claimed here.
Planned · No results yet
Care, made practical
You should know what you can trust.
Evidence before claims
A demo, a passing test, and a production result are different things. We keep that distinction visible.
Room for different models
We’re working toward a common handoff format, with local models first and supported provider connections next.
Human judgment, by design
Workers propose. Checks produce evidence. People retain control over publishing, spending, and consequential changes.
Start a conversation
What are you trying to build?
Tell us about the work, the friction, and what a useful first result would be.