Register Thinking Instructions to Set User Directives to Maintain During Handoff
You can register perspectives, work principles, and response criteria that you want to consistently maintain as Thinking Instructions and use them together in explicitly executed Handoffs.
When moving to a new AI conversation, not only project context but also the work criteria the user wants to consistently maintain may be needed. For example, the principle that existing features must be preserved, answer format, review order, and prohibitions may need to be re-explained each time. Thinking Instructions is a feature for directly registering such user directives.
Thinking Instructions are not a profile created by AI guessing the user but explicit directives written and managed by the user directly.
What Changed in This Update
- Users can directly write work principles and response criteria to consistently maintain.
- When the user executes Handoff, registered Thinking Instructions can be used together in context composition.
- Managed as user-centered directives not dependent on a specific Provider.
- Editing Thinking Instructions lets the user directly update the criteria to use in subsequent Handoffs.
- AI does not automatically generate or recommend new directives by analyzing conversation content.
How to Verify
- Open Thinking Instructions in 5by Tools or related settings.
- Write the principles and prohibitions to consistently maintain in clear sentences.
- Check the saved content and edit directly when needed.
- When selecting an Anchor and executing Handoff, check that Thinking Instructions are included as intended.
Example in Actual Usage Flow
For example, to use this feature in actual work, the user opens Thinking Instructions in 5by Tools or related settings. They write the principles and prohibitions to consistently maintain in clear sentences. Then they check the saved content and edit directly when needed. When needed, they check that Thinking Instructions are included as intended when selecting an Anchor and executing Handoff. However, when using the feature, the following boundaries must be observed. Thinking Instructions must be written by the user and are not automatically extracted from conversations. Knowing this distinction helps users use the feature without confusing the system's automatic processing with their explicit choices.
Before and After the Change
Understanding this feature as simple storage, auto-summary, or auto-recommendation blurs the boundary with other features. In this update, you can register perspectives, work principles, and response criteria that you want to consistently maintain as Thinking Instructions and use them together in explicitly executed Handoffs. We separated actions the user directly executes from actions where the system shows structure, clarifying what is recorded and what does not change state.
Scope and Things to Note
- Thinking Instructions must be written by the user and are not automatically extracted from conversations.
- They are not automatically injected independently just because a new conversation was opened; they are used within the Handoff execution flow.
- Directives do not guarantee all AI response results, and Provider behavior differences may exist.
- System input and the user's actual conversation utterances are not treated as the same record.
Why This Change Matters in the Overall 5BY.AI Structure
Thinking Instructions help keep the user's work principles at the center even when moving across multiple AIs. 5BY.AI maintains a structure where the user directly declares their criteria instead of the system interpreting for them, and explicitly uses them in needed movements.
Criteria for Verifying This Update
- The user's explicit choices and the system's automatic processing should be explained separately.
- The target unit of the feature and boundaries with other features should match between public description and actual behavior.
- Screen navigation or similarity alone should not automatically determine importance, correct answers, or recommendation paths.