Reuse context across questions
Carry useful context forward without restarting from a blank prompt.
Most AI work loses value because the context is trapped in the last exchange. You write a strong explanation, make a useful decision, or extract an important source passage, then the next question starts as if none of that structure exists.
BaseHalf treats context as material you can keep applying.
Context should be portable
A reusable piece of context has three qualities:
- It can be applied independently.
- It can be combined freely with other context.
- It can be carried into the next question without rewriting it.
That is why BaseHalf breaks work into Points and Blocks. Each small piece can stand on its own, but it can also join a larger Map.
Turn context into named objects
Reusable context needs a name. A vague memory such as “the thing we decided earlier” is hard to apply. A Point called Decision: keep onboarding single-player first is much easier to reference.
Name the reusable unit by the job it performs:
- Definition: what a term should mean.
- Decision: what direction has been chosen.
- Constraint: what must remain true.
- Source: where evidence came from.
- Rubric: how options should be judged.
- Question: what still needs to be answered.
Once the unit has a name, it can be opened, revised, referenced, and combined with other units.
Use References instead of repetition
Copying the same background into every prompt is slow and fragile. It also makes the system harder to update. If the source changes, you have to remember every place where you pasted it.
A Reference solves this by connecting Points.
When a writing Point depends on a source Point, reference the source. When a plan depends on a decision Point, reference the decision. When a comparison should inform a recommendation, reference the comparison.
The AI can then work with the relevant context without turning your prompt into a pile of repeated background.
Decide what deserves to be reusable
Not every sentence needs to become durable context. Preserve the pieces that change future work:
- Definitions that must stay consistent.
- Decisions that other work depends on.
- Source passages that support a claim.
- Constraints that affect planning.
- Examples that should guide later writing.
- Open questions that will need follow-up.
If a piece of context will matter again, make it easy to find and easy to reference.
Do not preserve material only because it took effort to create. Preserve it because it changes future work. A clever paragraph that will never be reused can stay in a draft. A small decision that will shape five later tasks deserves a Point.
Combine context deliberately
Reusable context is powerful because it can be composed.
You might combine a source Point, a decision Point, and a user-research Point to draft a product note. Later, you can reuse the same decision Point with a different source to write implementation tasks. The context remains modular. The output changes because the combination changes.
This is the practical meaning of compound thinking. Each useful piece of work becomes part of the next useful piece of work.
Keep context small enough to trust
Reusable context should not become a giant archive. If a Point is too large, the model and the reader both have to guess which part matters.
Split large context into smaller Points or Blocks when:
- one section supports a claim
- one section defines a term
- one section records a decision
- one section is only useful for one workflow
- one section should be reviewed on a different cadence
Small context is easier to inspect. It is also easier to combine without dragging unrelated assumptions into the next answer.