SentraMind

Checklist

What this POC claims, and what it does not

A buyer checklist for separating current POC behaviour, disclosed stand-ins, and design-partner roadmap from a production claim.

1. Is this sentence about current POC behaviour?

A finished answer maps the sentence to something a reviewer can exercise in this build, and it names the maturity label beside it. Platform Core and GenAI Prompt DLP are the modules that can carry the label available for this POC. If the sentence needs AI usage observability, it is labelled coming next. If it needs any other crew direction, it is labelled design-partner roadmap and it is not written as a result.

Ask. Is this sentence about behaviour the current POC can exercise, and which maturity label sits beside it?

Why. A sentence that needs a roadmap module is a sentence about a direction, and the label has to say so.

Weak answer. The demo looked finished, so the sentence can be written as current behaviour.

2. Which stand-in is doing the work in this step?

Name the stand-in in the same sentence as the step. The names to reach for are the simulated endpoint fleet, software officer keys, the software machine CA, the synthetic log source, the emulated checkpoint acknowledgement, the local password account, and the encoder stand-in with an unmeasured prototype head. The architecture drawing can show the intended component. The sentence still has to say which stand-in ran.

Ask. Which stand-in performs this step, and where is that label visible to the reader?

Why. A simulated fleet, a software key, a development CA, and a generated log are named stand-ins, and a diagram does not promote them.

Weak answer. The architecture diagram shows the real component, so the label can be omitted.

3. Does the sentence contain a number we have not measured?

The finished answer contains no latency percentile, no throughput, no GPU figure, and no model-quality score. Incident counts from a real estate are absent for the same reason. Where the product carries a latency field, the honest value in this phase is empty. An estimate written to fill the gap is a measured number that was never measured.

Ask. Does this sentence contain a latency, throughput, or quality figure, and was that figure measured?

Why. Those figures are left empty on purpose, because a number from a handful of development requests would be read as a benchmark.

Weak answer. Use a reasonable estimate so the sentence has a number in it.

4. Does “shipped” here mean the grid cell or a contract?

If the word appears, the finished answer points at the coverage-matrix legend and at the cell. In that legend, a shipped detection class is one that runs in the deterministic tier today, a simulated cell is a real code path exercised by a harness, and a roadmap cell is specified and unbuilt. The word does not, by itself, state that a buyer can purchase a production estate.

Ask. When this sentence says shipped, does it mean a cell in the coverage matrix, or a commitment someone can buy?

Why. The coverage-matrix legend gives the word its meaning, and the word does not travel into a contract on its own.

Weak answer. Shipped means the customer can buy it.

5. Are we implying a customer, a duration, or a certificate?

Read the sentence aloud and mark any organisation, any length of time in service, and any approval certificate. The finished answer has none of the three. It describes a mechanism, a fixture, and a limit. A composite story about a deployment that did not happen is still one of those three claims.

Ask. Does this sentence imply a customer, a time in service, or a certificate?

Why. Those claims wait for evidence the owner has approved, and this POC does not supply them.

Weak answer. A composite story reads more clearly than the mechanism alone.

6. Would enforcement and the viewer agree?

The finished answer assumes one deterministic evaluator for the effective-policy view, for dry-run, and for the verdict path. Precedence inside it is the exception, then the rollout stage, then the rule action. If the sentence needs the viewer to show one action and enforcement to take another, the sentence is describing a defect. Cut it or rewrite it until both callers would return the same action for the same snapshot.

Ask. Would the effective-policy viewer and the enforcement path return the same action for this sentence?

Why. The design is one evaluator for both callers, so a sentence that needs them to disagree is wrong.

Weak answer. The viewer is only illustrative, and enforcement can differ.

7. Is model use on the verdict path?

The finished answer keeps the verdict path on the byte cap, the policy snapshot, and the deterministic tier. A generative model does not sit on that path. The encoder tier, when the sentence mentions it, is the CPU stand-in with a synthetic prototype head, and its precision on a real corpus is unmeasured. Absence of a model is a stated condition, and the sentence does not fill it with a quiet exception for difficult prompts.

Ask. Does this sentence put a generative model on the verdict path?

Why. The verdict path is deterministic first, and a generative model there would contradict the design.

Weak answer. Only the hard prompts would be sent to a generative model.

8. What will we show beside the claim?

Place three things in reach of the sentence: the limit that bounds it, the stand-in label if one is in the path, and the synthetic fixture that produced the screen. A reviewer should be able to find all three without asking. The benefit line can stay. It does not replace them.

Ask. What limit, label, and fixture will be shown beside this claim?

Why. A claim without the limit, the stand-in label, and the fixture is a brochure line.

Weak answer. The benefit statement is enough on its own.

How to read this

How to read this
Read it asWhich meansDo not read it as
Current behaviour What the POC evaluation can run A roadmap module that looked finished
A number Only one that was measured A reasonable estimate
Beside the claim The limit, the label, and the fixture A benefit line on its own

Product statements on this page follow the POC maturity ladder. A coming-next or roadmap item is not available for the current evaluation.