XKP AI (XKPAI) / Resources

XKP AI guides.
Ideas in context.

Original XKP AI (XKPAI) guides on context, scenario thinking and measured feedback, with a worked example of reviewing an energy signal from observation to a recorded next step.

Read the worked exampleExplore the platform

Review an energy signal,
with its context intact.

Solar output has fallen. Is it a change in weather, an equipment issue, or incomplete information? This worked example follows the question through the XKP AI design approach.

Illustrative example. All readings, source names and review times below are synthetic. This is a manual review exercise, not a live XKP AI integration, a validated model or a customer result.

Input: preserve the observation

Keep the original source and measurement unit with each value. Here, output comes from an inverter log, irradiance from a weather sensor, and weather notes from an operator. Device status is a separate feed.

On a narrow screen, scroll the table sideways to compare every field.

Sample observations · 20 October 2026 · UTC+08:00
TimeOutput
kW · inverter A		Irradiance
W/m² · sensor W1		Weather
operator note		Device status
inverter A status log
10:0042720Partly cloudyRunning · received 10:00
10:0528480Cloud cover increasingRunning · received 10:05
10:1027475Cloud cover persistsUnknown · no fresh status

These are point observations, not interval averages. The 10:10 output reading is available, but the last device-status message is from 10:05. Missing status must remain visible.

Check: is the comparison valid?

Confirm the time zone, units and asset identity before comparing records. Check timestamps for delays or duplicates, and confirm that the weather sensor describes the same location. Do not treat an old “running” status as a fresh observation.

Compare: keep explanations open

From 10:00 to 10:05, output and irradiance both fall by about one third. That is consistent with less sunlight, but does not prove the cause. Equipment conditions, sensor quality or timing differences still need review; three readings cannot establish a fault or forecast.

Review: bring in a person

An assigned reviewer checks the local operating log, recent maintenance and any alerts, then requests a fresh device-status message. If context is still missing, the reviewer records that limit instead of turning the comparison into an instruction.

Output: record the next step

The useful output is a review record: what changed, the possible explanations, the missing evidence and who will check it. This exercise does not change equipment settings or issue automatic control commands.

A review record that can be revisited

Observation
Output fell alongside irradiance; current device status is missing.
Assessment
Weather is one possible explanation. Equipment condition remains unconfirmed.
Next step & owner
Assigned reviewer requests fresh status and checks the operating log at 10:15.
Feedback point
At 10:30, attach the next valid readings and log check. Mark the explanation supported, revised or still unresolved.
Current outcome
Pending review. No result or performance improvement is assumed.

Use the platform explorer to see the same observe, model and coordinate structure, or read the guides below for the reasoning behind each step.

Showing 3 concept guides.

Conceptual context illustration
context / Concept guide · 2 min read

Context is part of the signal

A useful observation carries its environment with it.

Read the guide

A number needs a setting

A measurement is an observation at a particular time, from a particular source, in a particular environment. Removing those details may make it easier to compare values, but it can also remove the reason a value matters.

Consider a change in equipment output. Its meaning depends on operating mode, recent conditions, nearby events and the expected range. The same change may be ordinary in one setting and deserve review in another.

Preserve the details that change the interpretation

Useful context begins with the basics: source, timestamp, units and the conditions under which a signal was captured. It also includes relationships. Which components affect this signal? Who is responsible for reviewing it? Which assumptions are being made?

A practical interface does not need to show everything at once. It should make the relevant detail available at the moment someone asks why a pattern appears or whether it applies to the current situation.

Make uncertainty readable

Missing information is itself part of the picture. A delayed source, a changing boundary or an incomplete observation should be visible rather than silently folded into a confident-looking display.

The design goal is a shared basis for discussion. Keep the observation close to its context, and a person has a stronger starting point for asking what changed and what might follow.

A useful starting question

What details would change how we interpret this observation?

Conceptual modeling illustration
modeling / Concept guide · 2 min read

Scenarios make assumptions visible

Explore a possible change before making it real.

Read the guide

Begin with a question, not a prediction

A scenario is a structured way to ask what might happen if conditions change. It is useful when a team needs to compare possibilities, understand constraints or find the assumptions that most affect an outcome.

For a physical system, a useful question is usually specific: how might a change in operating conditions alter the observed pattern? Which nearby components could be affected? What information would improve the comparison?

Keep a visible boundary

Every model simplifies the environment. A clear scenario shows which conditions it includes, which relationships it assumes and what it leaves outside the boundary. That explanation helps people decide where the result is informative.

Comparisons are easier to interpret when the baseline stays visible. Show what changed between scenarios and what remained the same. Small, deliberate changes can be more informative than a large set of unexplained alternatives.

Leave room for review

A modeled result is a perspective, with limits. It can support discussion without becoming an instruction. A person should be able to inspect the assumptions and compare the scenario with local knowledge.

Useful review asks where a result appears sensitive, what evidence is missing and what a next observation might resolve. The output of scenario thinking may be a better question as often as a preferred option.

A useful starting question

Which assumption changes the comparison most, and how could we check it?

Conceptual learning illustration
learning / Concept guide · 2 min read

Feedback completes the loop

Connect a considered action to what changed afterward.

Read the guide

An action needs a review point

A decision becomes easier to learn from when its original context is recorded. What was observed? Which options were considered? What outcome would tell a team that its understanding had improved?

Without a review point, new measurements can become a separate stream of information. With one, they can help revisit the question that prompted an action in the first place.

Measure outcomes in context

A change after an action does not automatically explain its cause. Other operating conditions may have changed at the same time. Feedback should retain timing, relevant surrounding signals and the limits of what can be concluded.

That makes it possible to distinguish a useful observation from an overconfident explanation. A learning loop should accommodate surprises, incomplete results and the possibility that the original assumption needs revision.

Carry the learning into the next cycle

Useful feedback changes how the next question is asked. It may refine a scenario, adjust a review threshold or reveal that a different source of information is needed.

The aim is a visible practice: observe, interpret, coordinate, review. Each cycle adds context to the next, while people remain able to understand and question how the system arrived at its current view.

A useful starting question

What did we observe afterward, and what does it change about our original assumption?

A SHARED VOCABULARY

Make the conversation easier.

Four concepts recur across the intelligence loop.

Signal

An observation that describes a condition or change in an environment.

Context

The timing, source, relationships and constraints that make a signal meaningful.

Scenario

A structured exploration of how a system might change under explicit assumptions.

Feedback

Measured outcomes used to revisit a question and refine the next cycle.

Build with context

Put the idea in your context.

Explore how a clear intelligence loop could help describe your environment.

Prepare a project brief