Everyday work beyond code

The same working habits apply to a note, a comparison, a plan, or a research question: give the agent the material it needs, choose a useful artifact, inspect what it returns, and keep the decision that helps the next task.

All events, documents, source excerpts, and outputs on this page are fictional teaching material. The examples are constructed, not a transcript of agent runs. They make no claims about a real venue, reading group, or service. You can follow the reasoning without running code or installing a skill.

Write a note from supplied facts

Material: A volunteer's source note says: “Reading-group welcome at 18:00 in Room B. Setup volunteers arrive at 17:30. Facilitator unconfirmed. No registration link supplied.”

Assignment:

Draft a short invitation for attendees using this note. Keep the time and room exact. Put missing information in a separate editor's note. Use a friendly, direct voice and return a draft for review.

Illustrated result:

Join us for the reading-group welcome at 18:00 in Room B.

The editor's note can say: “The source does not supply a date, facilitator, or registration link. Confirm any of those needed for the final invitation.” The volunteer arrival time belongs in a volunteer message; placing it in the attendee invitation could make the arrival instruction unclear.

Check each factual phrase against the supplied note. Then review the voice and usefulness for this audience. These are different judgments: accurate details can still be presented poorly, and an inviting tone can still contain an invented fact.

Keep the accepted draft and the reason it fits. If the person prefers this style for future group invitations, give that preference a scope in the context brief. The taste exercise shows how to make that record concrete.

Compare documents without deciding the policy

Material: Two fictional planning drafts contain:

• Draft A: “Volunteer setup at 17:00 in Room A. Registration closes on Monday.”
• Draft B: “Volunteer setup at 17:30 in Room B.”

Neither is identified as the approved plan. Draft B's omission of registration could reflect a deliberate change or an incomplete draft.

Assignment:

Compare these drafts. Return changed details, details present in only one draft, and decisions needed before an invitation can be finalized. Preserve the wording that supports each difference. Do not choose which draft governs.

Illustrated result:

The next decision is which draft or correction supplies the current plan. “Registration no longer has a deadline” would go beyond the comparison's evidence.

Check that every substantive difference appears in the return and that the two source identities remain visible. Once the owner selects the governing details, record that decision separately. A later summary should be able to distinguish the document difference from the accepted plan.

Build a plan around constraints

Material: For a separate fictional setup exercise, the room opens at 17:30 and attendees arrive at 18:00. Preparing chairs takes an estimated 20 minutes for one volunteer; preparing materials takes an estimated 20 minutes for one volunteer. The two tasks can happen independently. Two volunteers are expected, but their attendance still needs confirmation.

Assignment:

Propose a setup plan using those constraints. Show task ownership and times, identify assumptions, and explain what changes if only one volunteer arrives. Return the plan and the confirmation needed before relying on it.

Illustrated result: With two volunteers present, one prepares chairs from 17:30 to 17:50 while the other prepares materials during the same interval. That leaves ten minutes for a final check before 18:00.

With one volunteer performing both estimated tasks sequentially, setup takes 40 minutes and finishes at 18:10. The current plan would miss the arrival time. Someone must change a constraint: obtain help, reduce the setup, arrange earlier access, or change the start. The agent can explain those options; the person needs to choose an available one.

Check the arithmetic, shared resources, and claimed task independence. The times are estimates supplied for the exercise, not measured durations. Before using a real plan, confirm access and attendance and decide what margin the uncertainty needs.

The useful artifact is a conditional plan with a clear missing confirmation. Formatting an optimistic schedule does not make the assumptions true.

Research a question that may remain unanswered

Material: The fictional question is whether Room B has a projector available for the welcome. Three supplied excerpts say:

• S1, venue overview: “A screen is installed in the main hall.”
• S2, last year's event note: “We used a portable projector at the spring gathering.”
• S3, current room booking: “Room B includes tables and chairs.”

Assignment:

Using only these excerpts, assess whether a projector is confirmed for Room B. Link each conclusion to a source label, distinguish current evidence from inference, and return the narrowest question that would resolve the gap.

Illustrated result: The supplied material does not confirm a projector for Room B. S1 concerns a screen in another room. S2 records historical use of portable equipment without current availability. S3 lists furniture but says nothing about a projector. Their combined silence still leaves the availability question open.

The useful next question is: “Does the current Room B booking include a working projector for this event, and does it require a separate reservation?” Return that as a draft question for the organizer to use.

For real research, retain the source location, identity, date, and relevant passage. A summary should let someone inspect the support behind the claim. These constructed excerpts demonstrate the reasoning; no external retrieval or venue check was performed.

Choose what comes back

Each assignment ends at a different useful state. The invitation returns a draft and missing facts. The comparison returns differences and an unresolved owner decision. The plan returns a conditional schedule. The research question returns an evidence gap and a way to resolve it.

Choose that return before starting. The work brief can be a few sentences for everyday tasks. Add context when prior decisions matter, and ask for a bounded investigation when the next attempt needs better information.

For something on your desk today, supply the smallest relevant source set and ask for one artifact you can check. Review it against the original request before deciding whether the next step is editing, confirming a fact, making a choice, or handing off the work.