MyJackfruit

Help
Browse topics · Starting

Starting

← Starting

What if the AI estimates a level I set?

Short answer

AI agents ask before changing a level you set: type your answer in Write a message… and press Send; without a yes, yours stays.

Before you start

You need to be signed in to GilbApp on a computer, with a vault you can write to, such as your own vault. Steps 1 and 2 need an agent's question in GilbApp's chat; steps 3 to 8, a Value with a level whose Source names an AI agent. The pictures show a made-up person on a test site; your screen shows your own name and address.

Words used here (6)
agent
An agent is your AI following one written procedure, its standing instructions, for one kind of work.
GilbApp
GilbApp is the app, opened in a web browser at https://gilb.com/app, where you open, fill and share your vaults.
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.
record
A record is one piece of data with its own address, such as a person, a place, a thing or a note.
short name
A short name is the one word you give one AI on one computer, such as gilb-example; it keeps that AI's key and tools apart.
field
A field is one named fact on a record, such as its title or a due date; its name comes from GilbVeda.
Steps
  1. Type your answer in Write a message….

    In GilbApp's chat, in the box that shows Write a message…, type the number of the agent's question and the number of the option you choose. The agent (your AI following one written procedure, its standing instructions, for one kind of work, such as Kai-Zen) asks before it changes a level you set, a number such as a Goal under Levels; its question starts Can I overwrite and names the level, your number and the number it would write.

    GilbApp with an agent's question in the chat asking to overwrite a Tolerable; the box "Write a message…" below it is outlined in orange.
    GilbApp with an agent's question in the chat asking to overwrite a Tolerable; the box "Write a message…" below it is outlined in orange.
    Show as text

    On the left, the chat pane shows the agent "Kai-Zen - Master Agent of Planning" at the top and its question under "Your questions:", with "1. Yes, overwrite 2. No, keep 10 (recommended)" and "Default: 2". The outlined box at the bottom reads "Write a message…". Type your answer there.

    More about this

    It asks the way it asks all its questions: under the line Your questions:, with numbered options and a line starting Default:. The answer on answering Kai-Zen's questions shows how to reply. A level counts as set by a person when its Source names a person, not an AI agent. While it waits for you, the agent stops only that part of the work and goes on with the rest. In Claude Code in Terminal you answer the same way: type your answers and press Return.

  2. Press Send.

    Press Send, or press Return. On a yes, the agent writes only that one number and names, in its source, who said yes and the date; on a no, or with no answer, it keeps your number and plans with it.

    GilbApp with an answer typed under the agent's question; "Send" to the right of the box is outlined in orange.
    GilbApp with an answer typed under the agent's question; "Send" to the right of the box is outlined in orange.
    Show as text

    The message box at the bottom of the chat pane now holds the answer, here "1. 2". At its right, the outlined button is "Send". Press it.

    More about this

    Shift and Return starts a new line without sending. While the agent is still working, the button reads Queue: your message waits and is sent when the agent is free. Then, to see on the card who set each level, or to make an AI estimate yours, go on with step 3.

  3. Press Main menu at the top left and choose Plan.

    Press Main menu, next to your profile picture at the top left, and under YOUR SECTIONS choose Plan. The button shows the name of the section you are in, such as Vaults.

    GilbApp with the "GO TO" panel open, and "Plan" under "YOUR SECTIONS" outlined in orange.
    GilbApp with the "GO TO" panel open, and "Plan" under "YOUR SECTIONS" outlined in orange.
    Show as text

    In the GO TO panel that opens from "Main menu" at the top left, look under YOUR SECTIONS on the left. The outlined row is "Plan", with the line "Your Kai-Zen plan: Functions, Values, Development Resources and Solutions." Press it to open the Plan screen.

    More about this

    A panel headed GO TO opens; Plan reads Your Kai-Zen plan: Functions, Values, Development Resources and Solutions. You can also press the Plan button on the card of a vault (the place where a person or a group keeps their records) in Vaults. The Plan screen lists your plan records and has the views Grid, Graph, Both and Card.

  4. Search and select the Value in the list.

    Type part of the Value's tag (its short name) in the search box and select the Value in the list. A Value is a perceived benefit, measured as a number.

    The "Plan" screen with the list of plan records, with the search box above the list outlined in orange.
    The "Plan" screen with the list of plan records, with the search box above the list outlined in orange.
    Show as text

    Look above the list on the Plan screen, under the type filters. The outlined box starts with "RECORDS" and reads "Search in current filters…". Click in it, type the tag of the record you want, and select it in the list.

    More about this

    The search box shows Search in current filters… while a filter is on, and Search your vault… otherwise. The button Value in the row of kinds shows only Values. Under Levels, the words after <- at the end of a line say who set that number.

  5. Press Edit.

    At the top right of the Value's card, press Edit. The button then reads Done.

    The Value's card, not in edit mode, with a "Goal" line ending in the AI agent's name after "<-", and the "Edit" button at the top right outlined in orange.
    The Value's card, not in edit mode, with a "Goal" line ending in the AI agent's name after "<-", and the "Edit" button at the top right outlined in orange.
    Show as text

    The Value's card is open and scrolled to "Levels": the "Status" and "Tolerable" lines end with a person's name after "<-", and the "Goal" line with the AI agent's name. At the top right, the outlined word is "Edit". Press it.

    More about this

    The card says Editing. Changes save as you leave a field. There is no Save button. Beside Done, a + does what + Add Information at the foot of the card does; you do not need it. The card shows Edit only to a person who may write in its vault. Without Edit, a click on a number or on a Source line changes it in place; Edit is clearer, because every box of the row is then shown.

  6. In the row's Source, press the red round (−) beside the agent's name.

    On the row whose Source names the AI agent, press the red round (−) at the left of the agent's name. The row's Source is then empty.

    The Value card in edit mode scrolled to "Levels", with the red round (−) beside the AI agent's name on the "Goal" row outlined in orange.
    The Value card in edit mode scrolled to "Levels", with the red round (−) beside the AI agent's name on the "Goal" row outlined in orange.
    Show as text

    In edit mode, find the "Goal" row under "Levels". At its right, above "Source:" and the AI agent's name, the outlined control is the small red round button with a minus. Press it to remove the agent as the source; it is not the "×" at the end of the row.

    More about this

    Do this before you type: a number typed while the agent is still named keeps the agent as the one who set it, so it still counts as the agent's estimate. On a row, the Source sits after the number box, before Note, e.g. estimate. Below the name, a box that shows Source can name someone else or a document; you need it only in step 7.

  7. Type your number in the row and press Enter.

    In the number box of the same row, type your number and press Enter. GilbApp then names you, by your own card, in the row's Source, so the number counts as set by you.

    The Value card in edit mode with the "Goal" row's source now empty, and the number box on the "Goal" row outlined in orange.
    The Value card in edit mode with the "Goal" row's source now empty, and the number box on the "Goal" row outlined in orange.
    Show as text

    The "Goal" row's source is gone; an empty "Source" box shows instead. The outlined box is the row's number box, still holding the agent's number. Type your own number and press Enter; your name becomes the source.

    More about this

    If you keep the agent's number, typing it again saves nothing. Then, and also if Source stays empty after Enter, type your name in the box that shows Source and select yourself; with no match the screen says No stakeholder or document matches. From then on, an AI agent asks you before it changes this number (step 1).

  8. Press Done when you have finished.

    Press Done at the top right of the card. What you typed is already saved.

    The Value's card in edit mode, scrolled to the top, with its "Illustration" showing the new Goal on the arrow, and "Done" at the top right outlined in orange.
    The Value's card in edit mode, scrolled to the top, with its "Illustration" showing the new Goal on the arrow, and "Done" at the top right outlined in orange.
    Show as text

    The card is scrolled to the top. The "Illustration" shows the new Goal on the arrow. At the top right, beside the box with the tag, the outlined word is "Done". Press it when you have finished.

    More about this

    Under Levels, the line now ends with <- and your name.

Deeper

Also asked:

What if the AI guesses a number I set myself? It keeps your number and plans with it; it changes it only with your yes. The numbers it sets itself are marked as its estimates, and you may change any of them.

Here a level is one of the numbers under Levels, such as Status, Tolerable or Goal (on a Development Resource, what the work uses up, Budget). It is not the Level that says which part of the plan a record belongs to.

An AI agent that follows the Kai-Zen method does not wait for you for a number. It sets Status, Tolerable and Goal (on a Development Resource, Budget) itself, as well as it can, marks each as its estimate in its source, and tells you in one line; you may change any of them later. In a Value Kick-out Table, which picks one of several options, it makes its own estimates too, costs included, if no estimates are given. It may change its own estimates when needed, and keeps them marked as estimates.

A level you or another person set is changed by an AI agent only with a yes from that person or from you. Without a yes it keeps the number and plans with it, in a Value Decision Table, which compares Solutions, in a Value Kick-out Table and in every other way. It may still move a Value you wrote to the part of the plan where it belongs, such as from Product to Stakeholder, with your words and numbers unchanged, and lists each move for your approval at the end. At the end it also lists every Value you set that is hard to reach.

This is a rule the agents follow, not a lock: the Levels in GilbApp ask nothing and show no AI mark, and anyone who may write in the vault can change a number there. A line names who set it only after <-, and typing over a number does not change who is named (steps 6 and 7).

The numbers under Impacts, the change a Solution is expected to bring, do show it. A number an AI agent estimated shows AI estimate · and the agent's name, with Accept when you may write there: press it, or type your own number, to make the number yours. A number another person set shows set by and their name, and before you replace it the box asks set this number. Replace it with yours? after their name, with Replace and Keep theirs. The answer on writing a Solution in your plan covers impacts.

How to set the levels yourself is in the answer on setting a Goal and a Tolerable level; a new reading is in the answer on updating a Value's Status.

This answer is written for a computer screen.

Source: GilbApp af0f04ee838a4d55887d2ff1f72edd49a4ff7ca0 (https://gilb.com/app/version.json, read 2026-11-09). Chat: AgentChat.tsx:1046 (Write a message…), :1062-1064 (Send, Queue while busy), :710-714 (Return sends, Shift+Return does not), :512-519 (a message typed while busy waits in the queue). Steps 3 to 5 and 8: identical to to 3 and 9 (gilbAppMenuRows.ts:158-166; RegisterScopedSearch.tsx:73-74; cardMode.tsx:233-305; AddInformation.tsx:33). Levels: planValue.ts:1981-2034 (a level line ends <- and its source), PlanValueLines.tsx:470-481 (the source's name), :115-116 and :2051-2054 (a typed number names the person only when the row's Source is empty), :1872-1877 (the same number typed again saves nothing), :2546-2554 (the row's Source), :221-236 and :684-720 (in place without Edit); PlanSource.tsx:37, :171-241 (Source:, Remove, picker Source, No stakeholder or document matches.); useSelfSource.ts:20-38 (your own card); RemoveControl.tsx:26-51 (red round). No ask and no AI mark on levels: useReplaceGate and SetByLine are used only by impact and cost cells (PlanImpacts.tsx:577, :932, :1414; PlanCellValueSide.tsx:136, :356; PlanCellResourceSide.tsx:151; PlanCostCell.tsx:177; PlanTable.tsx:484), not by PlanValueLines.tsx. Impacts: PlanHumanSet.tsx:92-156 (AI estimate · <name>, Accept only for a person who may edit and whose own card is known, set by <name>), :258-281 (<Name> set this number. Replace it with yours?, Replace, Keep theirs); humanSet.ts:16-42 (mayOverwrite: an AI's number is the person's to change, a person's number asks), :23-25 (GilbApp has no AI write path; the refusal belongs where the write lands). Standards (2026-11-09): Kai-Zen.ValueRules.Estimate, .HumanSet (Can I overwrite <field> '<current>' with '<new>'?; yes: write only that, name who said yes and the date in its source; no or no answer: keep it and plan with it; not by VKT, VDT or any other way; an AI's own number may be changed and stays marked as an estimate; a Value at the wrong level still moves); Kai-Zen.ResourceRules.Estimate (Budget); Kai-Zen.VDTRules.HumanSet; Kai-Zen.VKTRules.Estimate, .HumanSet; Kai-Zen.AskRules.Decide (tell in one line; may change later), .Stop (a change to a human-set Value stops that part, asked per Form), .Form, .EndApproval (every human-set Value that is hard to reach); Agent.Outcome.Gate (moved with words and numbers unchanged, listed for the end approval).

Checked against GilbApp d529620.