Working together

I take on a few engagements at a time.

What you get is one person with an unusually good harness – me, orchestrating a fleet of AI agents inside your codebase and your working practice. There is no bench of engineers behind me, and every judgment you see funnels through one head. That is the real shape of it, and it is worth knowing before you write.

Email me – hello@skylarkcreations.com
Status
Open to a small number of engagements
Based
Colorado. Engagements run remote and async-first
Shape
One human orchestrating a fleet of AI agents, inside your codebase
Billing
Engaged attention, not machine time
Start
An email, then a call once there is a decision to make

How it runs

What the working relationship looks like

Async by default

No standing meetings. A call gets scheduled when a decision needs one; everything else arrives as artifacts in your inbox. One line from you can re-sort my whole week.

Verified turnarounds, usually same or next day

You email late about a bug; the root cause, the fix, and the deploy are done by morning. On one engagement, 6–11 verified merges per day was the normal rate over a two-week stretch. That is what one decision-maker plus a fleet of agents is for.

Nothing you say evaporates

Every round of feedback gets broken into numbered points, and each point gets a written disposition: a faithful paraphrase, a ruling with its reasoning, and a pointer to where it landed. When I disagree I argue it in writing rather than quietly dropping it.

Spec and a kill-gate before code

We decide and spec first, and the opening milestone is designed to prove or kill the core technical bet in days – before there is anything expensive to be wrong about.

A second model tries to refute the work before you see it

Release gate: a different AI model family, with no stake in the reasoning that produced the work, is prompted to attack it. Whatever survives gets re-verified against live code. This gate has caught a bypass in a fix that had already been reviewed.

Scope triage happens in the open

When a client sent a batch of additions before signing, roughly 80% were confirmations or cheap adds and got absorbed at no charge; exactly one item was named as real scope and priced. They saw the split.

What it costs

Billed on engaged attention

Not machine time. Adding agent capacity never raises your invoice – the leverage is mine to supply, not yours to fund. Ramp-up and admin are not billed. A quiet week legitimately bills near zero, and if an engagement is idling because the person I need is consumed elsewhere, I will tell you it is idling rather than manufacture activity to fill an invoice.

Scope lives in a running ledger rather than a fixed deliverable list, so a change in direction gets absorbed instead of triggering a change order.

Fit

Who this isn’t for

Named plainly, so nobody wastes a month finding out. If one of these is the shape you need, there are good shops for it – I am not one.

Standing-meeting cultures. Calls here are decision events, not a daily ritual.

Fixed-scope builds thrown over a wall. Scope lives in a running ledger, and direction changes get absorbed rather than renegotiated.

Staff augmentation. I am not a seat to fill on someone else’s sprint board.

Teams with nobody to react. The loop runs on your reactions – it works best when someone on your side can take a pull request, argue a finding, and merge.

How to start

Send an email describing what you are trying to build or fix. You do not need a spec – a paragraph is enough to tell whether this is a fit. If it is, the first milestone is designed to prove or kill the core technical bet before there is anything expensive to be wrong about.

hello@skylarkcreations.com

Or read how the studio runs first – it is the same operating system your engagement would run on.