REQ-0142
Password reset completes within 5 seconds
Should Timeout ≤ 5 s, enforced Is Configuration declares 30 s
High risk Observed Observed — found in the implementation
You own the code. Do you own the understanding?
A map of your components, their dependencies, what is claimed about them and the evidence behind each claim — built from an analysis run by a coding agent you already trust.
No repository access. No source-code upload to 1ADK. Someone with access to the code runs one instruction — you, your developer, or your agency. Nothing to install, and 1ADK gets no access to anything.
Orbita Billing
Requirements, and what was actually found
REQ-0142
Password reset completes within 5 seconds
Should Timeout ≤ 5 s, enforced Is Configuration declares 30 s
High risk Observed Observed — found in the implementation
REQ-0143
Reset emails are delivered through a queue
Should Asynchronous delivery Is Redis queue, one worker
Confirmed Verified Verified — confirmed by evidence
REQ-0151
Payment webhooks are idempotent
Should Repeat delivery is safe Is Unknown — nobody has confirmed this
Unknown Declared Declared — stated by documentation
Should
Timeout ≤ 5 s, enforced and measured
Is
Configuration declares 30 s; runtime not measured
Not a diagram somebody drew from memory. A picture assembled from what an agent found in your own repository, with the uncertain parts left visibly uncertain.
Orbita Billing — a demonstration project, rendered live by the 1ADK design system. Every figure below is that project's, not a customer's.
Code can belong to the company while understanding of the system leaves with a person.
That is not a documentation problem. It is an ownership problem, and it only becomes visible at the moment it is most expensive.
How it works
The analysis runs in a tool you already trust, on your side. What reaches 1ADK is a description of your system that you have seen and approved first.
You get a set of instructions for the coding agent you already use. Nothing is installed, nothing is connected.
Claude Code, Codex or a local model does the reading, in the environment you choose, with the access it already has.
The names of the parts, what each one is for, what it connects to, and what that conclusion is based on. Not your source files.
1ADK builds the picture, and the next analysis shows you what moved.
What will be sent
exampleNot in this transfer
Evidence
For every statement it can support, 1ADK shows the file, configuration or document it came from. What it cannot support, it marks unconfirmed rather than filling in with something plausible.
The password reset endpoint takes 30 seconds, not the five the requirement asks for.
Every part of that record points back at something real: the line of configuration, the requirement it contradicts, and the reason the colour is what it is. Nothing on the map asserts anything it cannot show you.
See it on the mapUnknown — no limiter found in the code; absence not proven
Unknown — outside the part of the system that was scanned
Unknown — nothing measured it while it was running
Not found is not the same as proven absent. An unknown comes with the reason it is unknown, which is the part you can act on — and it is left on the map rather than quietly filled in.
Two sources disagree
When your documentation and your system disagree, you see both and which is which. 1ADK does not pick one and move on.
What changed
Run the analysis again and the picture does not simply refresh. It tells you what arrived, what moved, what went away — and which of yesterday's answers it can no longer stand behind.
The same demonstration project at two scans. 1ADK makes no claim here about how often you would run it, or about what any plan includes.
The boundary
The analysis runs on your side. 1ADK receives a reviewed description of your system and never opens a connection to your machine.
The honest limit If you run the analysis with a cloud coding agent, that agent processes your code under its own terms. 1ADK's boundary is the 1ADK service. Choose a local model and the analysis can stay entirely inside your own infrastructure.
When this matters
The condition is always there — understanding of the system lives in people rather than in the company. These are the moments it stops being theoretical.
And when it is not for you If you are technical, you want a single answer today and you will not need it again — your coding agent on its own is enough. 1ADK is for the case where the answer has to outlast the person who asked it, and be checkable by someone who did not.
The person who knows why it was built that way is working their notice. The code stays; the reasoning is about to walk out with them.
Moving contractors means handing your system to someone who has never seen it, with no independent account of what they are inheriting.
The product works and ships. What is missing is anyone who can say what it now contains, or which parts nobody has checked.
Their first months are spent reconstructing what already exists, from the people who are already busy.
The first map is free. Someone with access to the code runs one instruction — you, your developer, or your agency.
Do I have to give 1ADK access to anything?
No. Not your repository, not SSH, not production, not your .env.
Who runs the instruction?
Anyone with access to the code — you, an in-house developer, your agency, or the person who is leaving.
Do I need to install something?
No. The instruction is copied into a coding agent the runner already has.