MyJackfruit

Help
Browse topics · Agents and AI

Agents and AI

← Agents and AI

How does Kai-Zen keep the whole project's Values in view?

Short answer

Agent.Architect, an agent Kai-Zen hands work to, adds Values that matter across the whole project, such as security, to the decision tables of their level.

Before you start

Words used here (3)
agent
An agent is your AI following one written procedure, its standing instructions, for one kind of work.
Kai-Zen
Kai-Zen is the planning agent: it finds out what you want, chooses the way, hands the work to other agents and reports back.
project
A project is work to bring about a change you want, such as a new app or a better way of doing something.
Deeper

An agent is your AI following one written procedure, its standing instructions, for one kind of work. Agent.Architect's standing instructions say it keeps the Values that span the whole project in every decision. A Value is a perceived benefit, measured as a number. A Value is one of these when it still matters after any single function or activity of the project is taken away, such as how secure the app is, or a user's privacy.

Agent.Architect writes these Values at the Product level of the plan. It writes one at the Stakeholder level instead only when a stakeholder holds that Value as their own, such as a user's privacy; a stakeholder is a person, group or system that cares whether a Value goes up or down, and the answer on the levels of a plan explains the levels. Each such Value sits at one level only and is never copied to another. Agent.Architect adds it as a row to each Value Decision Table (VDT) and each Value Kick-out Table (VKT) whose rows are that level's Values, and scores it like the others. A VDT ranks options by how much they move the level's Values for their cost; a VKT picks one of several options. An option that drops one of these Values below its Tolerable (the survival limit) loses on that alone.

Agent.Architect also names the threats to the project: who could do what harm, and how. For each threat it writes a Security Value, such as the share of attack attempts that fail, or the time it takes to notice one, with a Scale, the ruler it is measured on, a Meter, which says how its number is found, and a Tolerable. It writes each one at the level its other Values go to, and adds it as a row like them.

Before a Solution is chosen, Agent.Architect checks that five ways to get it were each on the table; in its own words: build new, reuse our own, borrow, share, and wrap what exists. A Solution is how a Function (what a system does) gets done. Agent.Architect stores the check as one record about that Solution, with one line for each way: what it considered, what it found and why. The answer on reusing a Solution or a Constraint covers it.

Agent.Architect plans the rounds of delivery so that every Value is first lifted above its Tolerable, and only then moved toward its Goal (the number you plan to reach). When it meets a choice it would ask a person about, it runs a VKT on it.

Kai-Zen hands work to Agent.Architect when it challenges the Solutions while it sharpens the plan: one Solution, its alternative and their VKT at a time, in each of the two rounds of challenges. Kai-Zen sharpens a new project's plan after it has written its levels, sharpens it again with what was learned after each round of delivery, and sharpens it after you add or change a spec, a requirement written in your plan. The answer on challenging a Solution explains sharpening.

GilbApp has no mark for these Values: on the Plan screen they are Values like any other. It has no view for the check of the five ways either.

This answer is written for a computer screen.

Source: Standards read 2026-11-10: Agent.Architect (K agent-architect-06gd2d2akhyx3b1th9qrx09q6m), 11 steps: Role (You keep the Values that span the whole project in every decision.), Test ("A Value is yours if it still matters after you delete any single function or activity."), Rows ("Write your Values at Product level. Write one at Stakeholder level instead only when a stakeholder holds it as their own, for example a user's privacy. Each Value sits at one level only and is never copied to another. Add it as a row in each VDT and VKT whose rows are that level's Values, and score it like the others. A means that drops one below Tolerable loses on its own."), Options (five ways; t:SolutionReuseCheck, one line each what you considered, what you found and why), Path (first lift every Value above Tolerable, then move toward Goal), Security ("who could do what harm, and how ... a Security Value with Scale, Meter and Tolerable, for example the share of attack attempts that fail, or the time to notice one"), Choices (When you face a choice you would ask a human about, run Kai-Zen.VKT.); Cutover, ChangeLog, Commands, Sign not used (not about Values). Kai-Zen.Sharpen (K 06gdk6x12htfn3v2czbajncqdm) Challenge and Repeat ("Hand each Solution to its own hand-off: one Solution, its alternative and its VKT per call"; both hand off to Agent.Architect); Kai-Zen.New.Sharpen, Kai-Zen.Cycle.Improve (when Kai-Zen sharpens). Kai-Zen.New, .WriteLevel, .VDT, .ProductLevelRules read: none hands off to Agent.Architect. GilbApp af0f04ee838a4d55887d2ff1f72edd49a4ff7ca0 (version.json read 2026-11-10), the code shows only code comments (no mark, no view). GilbVeda public read: t:SolutionReuseCheck 200; t:ArchitectValue, t:SecurityValue, t:ProjectWideValue 404.

Checked against GilbApp d529620.