How we test Built with AI, proven by evidence Every SpatiumDDI build is installed from its ISO, walked four ways and broken on purpose before a person signs it off. A visual tour of ddi-pg, our proving ground, and why software written mostly with AI needs one.
What we do on purpose
Preview before commit, everywhere
Imports, migrations, CIDR planning, DNSSEC key operations — each shows the resulting diff before anything is written. Most famous DNS outages were a valid-looking change pushed straight to production; a preview is the cheapest guard against the whole class.
Fail open, not closed
The DNS and DHCP engines run as containers the control plane owns but does not gate. If the manager is down, resolution keeps working. A management plane that can take out the data plane is a liability, not a feature.
Atomic upgrades with a health gate
A new image is written to the inactive slot while the current one keeps serving, and only a healthy boot gets to keep it. Rollback is not a runbook step someone has to remember at 3am; it is the default outcome of a bad boot.
Prove parity before taking authority
An evaluation runs beside your production system with nothing dispatched to your servers. We replay your real queries at both systems and diff the answers. Cutover happens one zone at a time, with the old server kept warm.
Every mutation is auditable
Append-only, tamper-evident, and complete enough that compliance evidence collects itself. If you cannot show who changed what and when, you do not have a control — you have a hope.
Agentless wherever it is possible
We manage existing Windows DNS and DHCP over WinRM with nothing installed on your domain controllers. Every agent is a thing to patch, break and explain to a security review.
Where AI earns its place, and where it does not
We use AI heavily and say so, because the alternative — quietly using it and implying artisanal hand-craft — is the dishonest option.
What it does
- Reads existing configuration during a migration and proposes the equivalent in the new model.
- Drafts the repetitive, error-prone parts of a cutover plan so a specialist reviews rather than retypes.
- Answers operator questions about the estate in plain language, through the copilot.
What it never does
- Commit a change without a human approving the exact diff.
- Replace the specialist who owns the judgment call on your architecture.
- Get cited as a capability the product does not actually have.
AI is why our migration prices are a fraction of the incumbents’: it does the heavy lifting on the repetitive parts, and our specialists own everything that requires judgment.
Before it touches your production
SpatiumDDI is feature-frozen and a release candidate. It goes into business-critical production only after an explicit production-readiness bar is met — a hardening and QA program we run against a synthetic environment built to break it. Until then the honest path is the one we recommend to everyone: run it free, prove it beside your incumbent, and cut over when the evidence says you should.