How does superintelligence work when an AI has to decide whether two scheduled activities overlap? A study room is booked from 9:00 am to 9:40 am, and another activity starts exactly at 9:40 am. Do those sessions conflict? Under a clear start-inclusive, end-exclusive booking rule, they do not. But a booking starting at 9:20 am does overlap the first session, even if it lasts only fifteen minutes.
The mechanism is time-interval and schedule reasoning. A schedule is not merely a list of times: it contains starts, ends, durations, locations, capacities and boundary rules. An AI assistant should identify whether two intervals share actual time, distinguish adjacent events from simultaneous events, and verify the intended resource before reporting a conflict or claiming that a reservation was made.
This beginner-friendly guide uses six fictional study-room requests, a complete overlap audit and twenty-seven classification claims. All sessions are invented, use the same Singapore local clock on one unspecified day, and are not real bookings or statements about an actual room’s availability. The title “How Superintelligence Works” names an educational series; it is not a claim that present AI tools have reached artificial superintelligence.
Quick facts about intervals
- A schedule interval has a start and an end.
- This example uses a start-inclusive, end-exclusive boundary: [start, end).
- Two events that touch exactly at an end/start boundary do not overlap under this rule.
- An overlap requires a nonzero amount of shared time.
- Six fictional requests total 180 booked minutes, but occupy a 150-minute continuous union.
- The original single-room schedule has two overlapping pairs and needs changes.
1. A point in time is not a duration
A clock reading such as 09:40 marks a moment. An interval from 09:00 to 09:40 describes forty minutes of elapsed time. A start time without an end or duration cannot tell us how long the activity uses a room.
An AI that interprets every timestamp as a complete booking may invent occupied time. It needs the paired end time, or an authorised duration rule, before deciding whether resources are available.
2. Time intervals need boundaries
In our fictional desk, a booking includes its start time but excludes its end time, conventionally written [start, end). Thus a room is occupied at its start and becomes free at its listed end.
This rule makes back-to-back appointments straightforward: a session ending at 09:40 and another beginning at 09:40 do not share a minute or instant under the chosen model. Other institutions may specify buffers or different rules.
3. Touching does not necessarily mean overlapping
Suppose a lesson ends at 10:30 and another begins at 10:30. Under the half-open rule, the first lesson no longer occupies the room when the second begins.
An assistant should not flag that as an overlapping reservation just because the displayed endpoints are equal. If setup or cleaning time is required, it must be represented as a separate scheduling constraint.
4. Actual overlap is shared time
An activity from 09:00 to 09:40 and another from 09:20 to 09:35 share the period 09:20–09:35. That is fifteen minutes of overlap.
A useful conflict explanation names the overlapping interval rather than merely saying “there is a clash”. It gives the organiser enough detail to know what needs changing.
5. A general overlap test
For half-open intervals [a,b) and [c,d), with valid positive durations, they overlap if a is earlier than d and c is earlier than b. Both conditions must hold.
This test excludes cases where one interval ends exactly when the other starts. It also handles intervals of different lengths, including one short interval completely contained in a longer one.
6. Why one-sided checking can fail
An assistant might compare only whether the start of B falls within A. That detects some overlaps but can miss a long B that starts before A and continues into A.
The symmetric two-condition test is more general. It checks both endpoints rather than assuming all intervals begin in the order they were listed.
7. The fictional Cedar Study Room
Imagine one room, Room C, with a capacity of one study activity at a time. Six groups request it for one invented day. Every time is in the same Singapore local time convention, with no date transition or daylight-saving change.
The requests are proposed intervals, not confirmed room reservations. They do not contain information about tutors, students, equipment or permissions. This keeps the arithmetic separate from real operational availability.
8. The six proposed booking intervals
- A — 09:00–09:40, duration 40 minutes.
- B — 09:40–10:15, duration 35 minutes.
- C — 10:00–10:30, duration 30 minutes.
- D — 10:30–11:00, duration 30 minutes.
- E — 09:20–09:35, duration 15 minutes.
- F — 11:00–11:30, duration 30 minutes.
- One shared room, with capacity for only one activity at a time.
- Every interval includes its start and excludes its end.
- There is no assumed setup or cleaning buffer.
9. Check the duration of A
A begins at 09:00 and ends at 09:40. Its duration is forty minutes. We calculate duration by subtracting start from end, not by counting both endpoints as separate minutes.
The start-inclusive/end-exclusive convention avoids adding one extra minute to every interval. An AI that reports forty-one minutes has confused discrete labelled minute marks with elapsed duration.
10. Check B’s duration
B begins at 09:40 and ends at 10:15. It spans twenty minutes from 09:40 to 10:00 and fifteen more to 10:15, for thirty-five minutes total.
The change of clock hour does not reset the interval. A model performing only 15 minus 40 would give a negative result because it failed to account for the hour boundary.
11. Why A and B do not conflict
A’s end is exactly B’s start, 09:40. Because A excludes its end and B includes its start, there is no shared time in which both are active.
That makes them compatible under the stated one-room rule. A different real policy might require a five-minute transition, but no such buffer was included in this toy case.
12. Why A and E do conflict
E runs from 09:20 to 09:35, entirely within A’s 09:00–09:40 interval. Their shared time lasts fifteen minutes.
Two simultaneous activities cannot occupy the same one-activity room under the defined policy. The conflict is established by the interval overlap, not by any guess about the groups’ identities.
13. Why B and C do conflict
B covers 09:40–10:15, while C covers 10:00–10:30. Their common interval is 10:00–10:15, lasting fifteen minutes.
The collision continues across several clock labels, but it has one contiguous fifteen-minute shared period. An AI should not count every overlapping minute as a different conflicting pair of bookings.
14. Why C and D do not conflict
C ends at 10:30, exactly when D begins. Under the defined half-open intervals, C is no longer active at the instant D starts.
The two requests are compatible on this dimension. An assistant that flags them because the endpoints are the same has mistaken adjacency for a true overlap.
15. Why D and F do not conflict
D covers 10:30–11:00, and F covers 11:00–11:30. Their common interval is empty under the chosen boundary rule.
This result would change only if an additional rule required room preparation, cleaning or another buffer. Such a rule may be sensible in real life but cannot be invented as though the fictional policy already included it.
16. Why E does not conflict with B
E ends at 09:35, while B begins at 09:40. There is a five-minute gap between them. E’s overlap with A does not imply it conflicts with every later session.
A conflict graph should include an edge only for pairs with real shared time. The fact that E is involved in one conflict is not a blanket statement that it is unsuitable for the entire schedule.
17. Count all conflicting pairs
The two actual overlapping pairs are A–E and B–C. Each shares fifteen minutes. Other pairs either are separated or touch without sharing time.
The conflict count is two pairs, not four activities as if every pair among A, B, C and E conflicted. The pair structure matters when trying to repair the schedule efficiently.
18. Examine 09:25
At 09:25, A is active because 09:25 lies inside 09:00–09:40. E is also active because 09:25 lies inside 09:20–09:35. Thus two requests require the room simultaneously.
That exceeds the one-activity capacity. A useful scheduling checker should show both names and the relevant overlap rather than only a red calendar icon.
19. Examine 10:05
At 10:05, B is active, since it runs until 10:15, and C is active, since it began at 10:00. Both require the room.
Again the concurrency is two. A classroom assistant can determine this from exact interval membership without needing any information about whether the activities are popular or educationally valuable.
20. Examine exact endpoints
At exactly 10:30, C has ended and D has begun, so only D is active among those two requests. At exactly 11:00, D has ended and F has begun, so only F is active among that pair.
Checking equality cases helps catch off-by-one and inclusive-endpoint errors. The result depends on the documented interval convention rather than the visual closeness of calendar blocks.
21. Calculate requested activity-minutes
The six durations are 40, 35, 30, 30, 15 and 30 minutes. Adding them gives 40 + 35 + 30 + 30 + 15 + 30 = 180 requested activity-minutes.
This sum counts each activity’s duration independently, including time when multiple groups would share the room in conflict. It is not necessarily the number of distinct clock minutes during which the room is occupied.
22. Calculate the union of occupied time
A covers 09:00–09:40, followed immediately by B to 10:15. C overlaps B and extends occupation to 10:30. D then continues to 11:00, and F to 11:30. Thus the union of A, B, C, D and F continuously covers 09:00–11:30.
E lies inside A and adds no new occupied time to that union. The elapsed duration from 09:00 to 11:30 is 150 minutes. That is the distinct clock-time occupancy across all six requests.
23. Reconcile duplicate booking-minutes
The sum of individual activity durations is 180 minutes, while the union occupies 150 minutes. The difference is thirty activity-minutes that were counted more than once when adding individual durations.
Those thirty arise from the two fifteen-minute overlaps A–E and B–C. This is a special result of the particular fixture; for more complex overlapping schedules, the concurrency profile and inclusion–exclusion need care.
24. The maximum concurrency is two
The room has two overlapping activities during 09:20–09:35 and 10:00–10:15. No three requests share a time in the toy ledger. The maximum number of simultaneously active requests is therefore two.
Because the room capacity is one, the original plan is infeasible even though its overall clock-time union looks continuous and fits within 150 minutes. A conflict can occur without any empty time.
25. Reconciled schedule scorecard
- A duration: 40 minutes.
- B duration: 35 minutes.
- C duration: 30 minutes.
- D duration: 30 minutes.
- E duration: 15 minutes.
- F duration: 30 minutes.
- Total individual activity duration: 180 minutes.
- Distinct occupied clock time: 150 minutes.
- Overlap A–E: 15 minutes.
- Overlap B–C: 15 minutes.
- Excess booking-minutes: 30.
- Conflicting pairs: 2.
- Maximum simultaneous activities: 2.
- Room capacity in the fixture: 1.
26. Why the original schedule cannot all run
At 09:25, A and E both require the same single-capacity room; at 10:05, B and C do too. These two conflicts are enough to show that the proposed requests cannot all be hosted in the room at their stated times.
Mathematical infeasibility is different from a confirmed real booking failure. The fictional office has not called a provider, checked an actual calendar or reserved anything.
27. A conflict-free alternative
One possible sequential reschedule is A 09:00–09:40, B 09:40–10:15, C 10:15–10:45, D 10:45–11:15, E 11:15–11:30, and F 11:30–12:00.
The six activities now run consecutively, with their original durations unchanged, no overlaps and no gaps. The total continuous span is exactly 180 minutes, from 09:00 to 12:00. This demonstrates feasibility under the toy room-capacity rule.
28. A feasible plan is not a confirmed appointment
The reschedule above is one proposed arrangement, not evidence that a particular tutor or student can attend the new times. We have no data about their availability or school deadlines.
A tool-using assistant must still check participants, real room calendars, authority and provider outcomes before describing any of those proposed slots as booked.
29. Why the smallest edit is not always obvious
An organiser may prefer to move only the two short conflicting activities, but their new positions could create conflicts with existing sessions. A scheduling method should inspect the resulting whole timetable after each change.
Our sequential example changes several start times and is straightforward to verify. It is not claimed to minimise disruption, travel time or participant inconvenience; those objectives are absent from the fixture.
30. What if a cleaning buffer is required?
Suppose a real policy required five minutes between activities. Then sessions ending and starting at the same moment would no longer be acceptable back-to-back assignments. The room’s usable schedule would need extra gaps or longer blocks.
This would be a new constraint and might change the minimum finishing time. An AI should make it explicit instead of inserting hidden setup minutes into every duration.
31. Dates and time zones matter outside this fixture
All six fictional requests occur on one local day and use Singapore time, so we can compare their clock values directly. In international scheduling, the same clock reading in different time zones may represent different instants.
Software should convert actual events using appropriate offset-aware timestamps and IANA time-zone rules where necessary. Real daylight-saving transitions can introduce ambiguous or nonexistent local times in some regions.
32. Crossing midnight requires a date
An interval beginning at 23:30 and ending at 00:15 on the next day lasts forty-five minutes, but a time-only subtraction might mistakenly produce a negative number.
Date-aware arithmetic is needed. Our case avoids midnight entirely so the calculated minute differences are unambiguous. A model must not generalise the simplicity of the toy case to all schedules.
33. What does an empty interval mean?
An interval from 10:00 to 10:00 under the half-open rule contains no elapsed duration. It might represent a point event or an invalid room-booking request depending on the application’s policy.
The scheduler should distinguish events that take time from instantaneous reminders. Assuming that a zero-length entry occupies a room for an hour would invent a requirement.
34. Twenty-seven claims for an AI to classify
Use SUPPORTED for results established by the fixture and interval rules, CONTRADICTED for results that conflict with them, and NOT ESTABLISHED for factual statements about actual people or bookings not present in the source.
The list checks the interval-boundary rule, durations, overlap pairs and difference between calculated feasibility and external execution.
- 01. A covers 09:00–09:40 — SUPPORTED.
- 02. A and E overlap — SUPPORTED.
- 03. A and B overlap under the half-open rule — CONTRADICTED.
- 04. B and C overlap — SUPPORTED.
- 05. C and D overlap — CONTRADICTED.
- 06. D and F overlap — CONTRADICTED.
- 07. E lasts fifteen minutes — SUPPORTED.
- 08. A lasts forty minutes — SUPPORTED.
- 09. B lasts thirty-five minutes — SUPPORTED.
- 10. C lasts thirty minutes — SUPPORTED.
- 11. D lasts thirty minutes — SUPPORTED.
- 12. F lasts thirty minutes — SUPPORTED.
- 13. A and E share fifteen minutes — SUPPORTED.
- 14. B and C share fifteen minutes — SUPPORTED.
- 15. All six requests can run unchanged in the single-capacity room — CONTRADICTED.
- 16. A and E are both active at 09:25 — SUPPORTED.
- 17. B and C are both active at 10:05 — SUPPORTED.
- 18. C and D are both active exactly at 10:30 — CONTRADICTED.
- 19. D and F are both active exactly at 11:00 — CONTRADICTED.
- 20. The total requested activity duration is 180 minutes — SUPPORTED.
- 21. The union occupies 150 distinct clock minutes — SUPPORTED.
- 22. Total requested activity-minutes exceed union minutes by thirty — SUPPORTED.
- 23. Maximum simultaneous requests is two — SUPPORTED.
- 24. Every named participant is confirmed available for the alternative schedule — NOT ESTABLISHED.
- 25. The activities were actually reserved on a real room calendar — NOT ESTABLISHED.
- 26. A consecutive 09:00–12:00 alternative can host all six at original durations — SUPPORTED.
- 27. A and C share occupied time — CONTRADICTED.
35. Reconcile the reference answer counts
Supported statements are 01, 02, 04, 07–14, 16, 17, 20–23 and 26: eighteen. Contradicted are 03, 05, 06, 15, 18, 19 and 27: seven. Not established are 24 and 25: two. Eighteen plus seven plus two equals twenty-seven.
The last two demonstrate a crucial operational boundary. A plan can be mathematically feasible without there being proof of actual participant consent, available facilities or successful reservations.
36. Conflict graphs can simplify larger timetables
Imagine each requested activity as a node. Connect two nodes when their time intervals overlap and require the same exclusive resource. For this fixture, the conflict edges are A–E and B–C.
A single-room solution must avoid running connected nodes simultaneously. More complex scheduling software may use interval graphs, resource capacities and optimisation methods, but the beginner’s principle remains straightforward: shared time on an exclusive resource creates a conflict.
37. A classroom transfer exercise
Two tutorials are proposed: 14:00–14:50 and 14:50–15:40. With no buffer and a half-open boundary, they are adjacent and do not overlap. Replace the second start with 14:45, and the shared period becomes five minutes.
Ask learners to explain why the first pair is valid and the second is not for one room. The exercise tests time reading and interpretation, not advanced arithmetic.
38. A practical AI scheduling checklist
Confirm the date, time zone, start, end and duration of every request. Identify the specific resource and capacity. Apply the actual boundary rule, compare all relevant pairs, include required buffers and inspect the revised full schedule.
Finally, distinguish proposed availability from provider-confirmed reservations. A clear explanation is useful, but a real booking exists only when the appropriate system confirms it.
39. The central lesson
Time intervals represent occupied periods, not merely clock labels. Two activities can meet at one endpoint without overlapping, while two others can share fifteen minutes and create a real capacity conflict.
A dependable AI assistant checks boundaries, calculates overlap, preserves resource constraints and never calls a proposed timetable a completed booking without an actual receipt. Those habits make scheduling explanations both clear and reliable.
Frequently asked questions
- Do two sessions conflict when one ends exactly as the next begins? — Not under the stated half-open, no-buffer rule.
- What does [start,end) mean? — Start is included and end is excluded.
- How do I find an overlap? — Compare the latest start with the earliest end; a positive difference gives shared duration.
- Why are requested activity-minutes greater than the occupied union? — Overlapping sessions count shared minutes more than once.
- Does a conflict-free plan prove the room is booked? — No. Real confirmation requires a booking system.
- Do dates and time zones matter? — Yes, especially for cross-midnight and international scheduling.
- Is a five-minute transition always required? — Only if the governing room policy specifies it.
Connected learning and primary references
- How Super Intelligence Works — main educational curriculum.
- Relative Dates — why dates and local clock interpretation matter.
- Order and Dependencies — why the sequence of activities matters.
- Goals, Constraints and Trade-Offs — feasibility and hard requirements.
- Python datetime documentation — interval and time calculations.
- W3C Time Ontology in OWL — time instants, intervals and relations.