Stay Ahead, Stay ONMINE

Practical SQL Puzzles That Will Level Up Your Skill

There are some Sql patterns that, once you know them, you start seeing them everywhere. The solutions to the puzzles that I will show you today are actually very simple SQL queries, but understanding the concept behind them will surely unlock new solutions to the queries you write on a day-to-day basis. These challenges are all based on real-world scenarios, as over the past few months I made a point of writing down every puzzle-like query that I had to build. I also encourage you to try them for yourself, so that you can challenge yourself first, which will improve your learning! All queries to generate the datasets will be provided in a PostgreSQL and DuckDB-friendly syntax, so that you can easily copy and play with them. At the end I will also provide you a link to a GitHub repo containing all the code, as well as the answer to the bonus challenge I will leave for you! I organized these puzzles in order of increasing difficulty, so, if you find the first ones too easy, at least take a look at the last one, which uses a technique that I truly believe you won’t have seen before. Okay, let’s get started. I love this puzzle because of how short and simple the final query is, even though it deals with many edge cases. The data for this challenge shows tickets moving in between Kanban stages, and the objective is to find how long, on average, tickets stay in the Doing stage. The data contains the ID of the ticket, the date the ticket was created, the date of the move, and the “from” and “to” stages of the move. The stages present are New, Doing, Review, and Done. Some things you need to know (edge cases): Tickets can move backwards, meaning tickets can go back to the Doing stage. You should not include tickets that are still stuck in the Doing stage, as there is no way to know how long they will stay there for. Tickets are not always created in the New stage. “`SQL CREATE TABLE ticket_moves ( ticket_id INT NOT NULL, create_date DATE NOT NULL, move_date DATE NOT NULL, from_stage TEXT NOT NULL, to_stage TEXT NOT NULL ); “` “`SQL INSERT INTO ticket_moves (ticket_id, create_date, move_date, from_stage, to_stage) VALUES — Ticket 1: Created in “New”, then moves to Doing, Review, Done. (1, ‘2024-09-01’, ‘2024-09-03’, ‘New’, ‘Doing’), (1, ‘2024-09-01’, ‘2024-09-07’, ‘Doing’, ‘Review’), (1, ‘2024-09-01’, ‘2024-09-10’, ‘Review’, ‘Done’), — Ticket 2: Created in “New”, then moves: New → Doing → Review → Doing again → Review. (2, ‘2024-09-05’, ‘2024-09-08’, ‘New’, ‘Doing’), (2, ‘2024-09-05’, ‘2024-09-12’, ‘Doing’, ‘Review’), (2, ‘2024-09-05’, ‘2024-09-15’, ‘Review’, ‘Doing’), (2, ‘2024-09-05’, ‘2024-09-20’, ‘Doing’, ‘Review’), — Ticket 3: Created in “New”, then moves to Doing. (Edge case: no subsequent move from Doing.) (3, ‘2024-09-10’, ‘2024-09-16’, ‘New’, ‘Doing’), — Ticket 4: Created already in “Doing”, then moves to Review. (4, ‘2024-09-15’, ‘2024-09-22’, ‘Doing’, ‘Review’); “` A summary of the data: Ticket 1: Created in the New stage, moves normally to Doing, then Review, and then Done. Ticket 2: Created in New, then moves: New → Doing → Review → Doing again → Review. Ticket 3: Created in New, moves to Doing, but it is still stuck there. Ticket 4: Created in the Doing stage, moves to Review afterward. It might be a good idea to stop for a bit and think how you would deal with this. Can you find out how long a ticket stays on a single stage? Honestly, this sounds intimidating at first, and it looks like it will be a nightmare to deal with all the edge cases. Let me show you the full solution to the problem, and then I will explain what is happening afterward. “`SQL WITH stage_intervals AS (     SELECT         ticket_id,         from_stage,         move_date          – COALESCE(             LAG(move_date) OVER (                 PARTITION BY ticket_id                  ORDER BY move_date             ),              create_date         ) AS days_in_stage     FROM         ticket_moves ) SELECT     SUM(days_in_stage) / COUNT(DISTINCT ticket_id) as avg_days_in_doing FROM     stage_intervals WHERE     from_stage = ‘Doing’; “` The first CTE uses the LAG function to find the previous move of the ticket, which will be the time the ticket entered that stage. Calculating the duration is as simple as subtracting the previous date from the move date. What you should notice is the use of the COALESCE in the previous move date. What that does is that if a ticket doesn’t have a previous move, then it uses the date of creation of the ticket. This takes care of the cases of tickets being created directly into the Doing stage, as it still will properly calculate the time it took to leave the stage. This is the result of the first CTE, showing the time spent in each stage. Notice how the Ticket 2 has two entries, as it visited the Doing stage in two separate occasions. With this done, it’s just a matter of getting the average as the SUM of total days spent in doing, divided by the distinct number of tickets that ever left the stage. Doing it this way, instead of simply using the AVG, makes sure that the two rows for Ticket 2 get properly accounted for as a single ticket. Not so bad, right? The goal of this second challenge is to find the most recent contract sequence of every employee. A break of sequence happens when two contracts have a gap of more than one day between them.  In this dataset, there are no contract overlaps, meaning that a contract for the same employee either has a gap or ends a day before the new one starts. “`SQL CREATE TABLE contracts (     contract_id integer PRIMARY KEY,     employee_id integer NOT NULL,     start_date date NOT NULL,     end_date date NOT NULL ); INSERT INTO contracts (contract_id, employee_id, start_date, end_date) VALUES      — Employee 1: Two continuous contracts     (1, 1, ‘2024-01-01’, ‘2024-03-31’),     (2, 1, ‘2024-04-01’, ‘2024-06-30’),     — Employee 2: One contract, then a gap of three days, then two contracts     (3, 2, ‘2024-01-01’, ‘2024-02-15’),     (4, 2, ‘2024-02-19’, ‘2024-04-30’),     (5, 2, ‘2024-05-01’, ‘2024-07-31’),     — Employee 3: One contract     (6, 3, ‘2024-03-01’, ‘2024-08-31’); “` As a summary of the data: Employee 1: Has two continuous contracts. Employee 2: One contract, then a gap of three days, then two contracts. Employee 3: One contract. The expected result, given the dataset, is that all contracts should be included except for the first contract of Employee 2, which is the only one that has a gap. Before explaining the logic behind the solution, I would like you to think about what operation can be used to join the contracts that belong to the same sequence. Focus only on the second row of data, what information do you need to know if this contract was a break or not? I hope it’s clear that this is the perfect situation for window functions, again. They are incredibly useful for solving problems like this, and understanding when to use them helps a lot in finding clean solutions to problems. First thing to do, then, is to get the end date of the previous contract for the same employee with the LAG function. Doing that, it’s simple to compare both dates and check if it was a break of sequence. “`SQL WITH ordered_contracts AS (     SELECT         *,         LAG(end_date) OVER (PARTITION BY employee_id ORDER BY start_date) AS previous_end_date     FROM         contracts ), gapped_contracts AS (     SELECT         *,         — Deals with the case of the first contract, which won’t have         — a previous end date. In this case, it’s still the start of a new         — sequence.         CASE WHEN previous_end_date IS NULL             OR previous_end_date < start_date – INTERVAL '1 day' THEN             1         ELSE             0         END AS is_new_sequence     FROM         ordered_contracts ) SELECT * FROM gapped_contracts ORDER BY employee_id ASC; “` An intuitive way to continue the query is to number the sequences of each employee. For example, an employee who has no gap, will always be on his first sequence, but an employee who had 5 breaks in contracts will be on his 5th sequence. Funnily enough, this is done by another window function. “`SQL — — Previous CTEs — sequences AS (     SELECT         *,         SUM(is_new_sequence) OVER (PARTITION BY employee_id ORDER BY start_date) AS sequence_id FROM     gapped_contracts ) SELECT * FROM sequences ORDER BY employee_id ASC; “` Notice how, for Employee 2, he starts his sequence #2 after the first gapped value. To finish this query, I grouped the data by employee, got the value of their most recent sequence, and then did an inner join with the sequences to keep only the most recent one. “`SQL — — Previous CTEs — max_sequence AS (     SELECT         employee_id,         MAX(sequence_id) AS max_sequence_id FROM     sequences GROUP BY     employee_id ), latest_contract_sequence AS (     SELECT         c.contract_id,         c.employee_id,         c.start_date,         c.end_date     FROM         sequences c         JOIN max_sequence m ON c.sequence_id = m.max_sequence_id             AND c.employee_id = m.employee_id         ORDER BY             c.employee_id,             c.start_date ) SELECT     * FROM     latest_contract_sequence; “` As expected, our final result is basically our starting query just with the first contract of Employee 2 missing!  Finally, the last puzzle — I’m glad you made it this far.  For me, this is the most mind-blowing one, as when I first encountered this problem I thought of a completely different solution that would be a mess to implement in SQL. For this puzzle, I’ve changed the context from what I had to deal with for my job, as I think it will make it easier to explain.  Imagine you’re a data analyst at an event venue, and you’re analyzing the talks scheduled for an upcoming event. You want to find the time of day where there will be the highest number of talks happening at the same time. This is what you should know about the schedules: Rooms are booked in increments of 30min, e.g. from 9h-10h30. The data is clean, there are no overbookings of meeting rooms. There can be back-to-back meetings in a single meeting room. Meeting schedule visualized (this is the actual data).  “`SQL CREATE TABLE meetings (     room TEXT NOT NULL,     start_time TIMESTAMP NOT NULL,     end_time TIMESTAMP NOT NULL ); INSERT INTO meetings (room, start_time, end_time) VALUES     — Room A meetings     ('Room A', '2024-10-01 09:00', '2024-10-01 10:00'),     ('Room A', '2024-10-01 10:00', '2024-10-01 11:00'),     ('Room A', '2024-10-01 11:00', '2024-10-01 12:00'),     — Room B meetings     ('Room B', '2024-10-01 09:30', '2024-10-01 11:30'),     — Room C meetings     ('Room C', '2024-10-01 09:00', '2024-10-01 10:00'),     ('Room C', '2024-10-01 11:30', '2024-10-01 12:00'); “` The way to solve this is using what is called a Sweep Line Algorithm, or also known as an event-based solution. This last name actually helps to understand what will be done, as the idea is that instead of dealing with intervals, which is what we have in the original data, we deal with events instead. To do this, we need to transform every row into two separate events. The first event will be the Start of the meeting, and the second event will be the End of the meeting. “`SQL WITH events AS (   — Create an event for the start of each meeting (+1)   SELECT      start_time AS event_time,      1 AS delta   FROM meetings   UNION ALL   — Create an event for the end of each meeting (-1)   SELECT     — Small trick to work with the back-to-back meetings (explained later)     end_time – interval '1 minute' as end_time,     -1 AS delta   FROM meetings ) SELECT * FROM events; “` Take the time to understand what is happening here. To create two events from a single row of data, we’re simply unioning the dataset on itself; the first half uses the start time as the timestamp, and the second part uses the end time. You might already notice the delta column created and see where this is going. When an event starts, we count it as +1, when it ends, we count it as -1. You might even be already thinking of another window function to solve this, and you’re actually right! But before that, let me just explain the trick I used in the end dates. As I don’t want back-to-back meetings to count as two concurrent meetings, I’m subtracting a single minute of every end date. This way, if a meeting ends and another starts at 10h30, it won’t be assumed that two meetings are concurrently happening at 10h30. Okay, back to the query and yet another window function. This time, though, the function of choice is a rolling SUM. “`SQL — — Previous CTEs — ordered_events AS (   SELECT     event_time,     delta,     SUM(delta) OVER (ORDER BY event_time, delta DESC) AS concurrent_meetings   FROM events ) SELECT * FROM ordered_events ORDER BY event_time DESC; “` The rolling SUM at the Delta column is essentially walking down every record and finding how many events are active at that time. For example, at 9 am sharp, it sees two events starting, so it marks the number of concurrent meetings as two! When the third meeting starts, the count goes up to three. But when it gets to 9h59 (10 am), then two meetings end, bringing the counter back to one. With this data, the only thing missing is to find when the highest value of concurrent meetings happens. “`SQL — — Previous CTEs — max_events AS (   — Find the maximum concurrent meetings value   SELECT      event_time,      concurrent_meetings,     RANK() OVER (ORDER BY concurrent_meetings DESC) AS rnk   FROM ordered_events ) SELECT event_time, concurrent_meetings FROM max_events WHERE rnk = 1; “` That’s it! The interval of 9h30–10h is the one with the largest number of concurrent meetings, which checks out with the schedule visualization above! This solution looks incredibly simple in my opinion, and it works for so many situations. Every time you are dealing with intervals now, you should think if the query wouldn’t be easier if you thought about it in the perspective of events. But before you move on, and to really nail down this concept, I want to leave you with a bonus challenge, which is also a common application of the Sweep Line Algorithm. I hope you give it a try! Bonus challenge The context for this one is still the same as the last puzzle, but now, instead of trying to find the period when there are most concurrent meetings, the objective is to find bad scheduling. It seems that there are overlaps in the meeting rooms, which need to be listed so it can be fixed ASAP. How would you find out if the same meeting room has two or more meetings booked at the same time? Here are some tips on how to solve it: It’s still the same algorithm. This means you will still do the UNION, but it will look slightly different. You should think in the perspective of each meeting room. You can use this data for the challenge: “`SQL CREATE TABLE meetings_overlap (     room TEXT NOT NULL,     start_time TIMESTAMP NOT NULL,     end_time TIMESTAMP NOT NULL ); INSERT INTO meetings_overlap (room, start_time, end_time) VALUES     — Room A meetings     ('Room A', '2024-10-01 09:00', '2024-10-01 10:00'),     ('Room A', '2024-10-01 10:00', '2024-10-01 11:00'),     ('Room A', '2024-10-01 11:00', '2024-10-01 12:00'),     — Room B meetings     ('Room B', '2024-10-01 09:30', '2024-10-01 11:30'),     — Room C meetings     ('Room C', '2024-10-01 09:00', '2024-10-01 10:00'),     — Overlaps with previous meeting.     ('Room C', '2024-10-01 09:30', '2024-10-01 12:00'); “` If you’re interested in the solution to this puzzle, as well as the rest of the queries, check this GitHub repo. The first takeaway from this blog post is that window functions are overpowered. Ever since I got more comfortable with using them, I feel that my queries have gotten so much simpler and easier to read, and I hope the same happens to you. If you’re interested in learning more about them, you would probably enjoy reading this other blog post I’ve written, where I go over how you can understand and use them effectively. The second takeaway is that these patterns used in the challenges really do happen in many other places. You might need to find sequences of subscriptions, customer retention, or you might need to find overlap of tasks. There are many situations when you will need to use window functions in a very similar fashion to what was done in the puzzles. The third thing I want you to remember is about this solution to using events besides dealing with intervals. I’ve looked at some problems I solved a long time ago that I could’ve used this pattern on to make my life easier, and unfortunately, I didn’t know about it at the time. I really do hope you enjoyed this post and gave a shot to the puzzles yourself. And I’m sure that if you made it this far, you either learned something new about SQL or strengthened your knowledge of window functions!  Thank you so much for reading. If you have questions or just want to get in touch with me, don’t hesitate to contact me at mtrentz.com. All images by the author unless stated otherwise.

There are some Sql patterns that, once you know them, you start seeing them everywhere. The solutions to the puzzles that I will show you today are actually very simple SQL queries, but understanding the concept behind them will surely unlock new solutions to the queries you write on a day-to-day basis.

These challenges are all based on real-world scenarios, as over the past few months I made a point of writing down every puzzle-like query that I had to build. I also encourage you to try them for yourself, so that you can challenge yourself first, which will improve your learning!

All queries to generate the datasets will be provided in a PostgreSQL and DuckDB-friendly syntax, so that you can easily copy and play with them. At the end I will also provide you a link to a GitHub repo containing all the code, as well as the answer to the bonus challenge I will leave for you!

I organized these puzzles in order of increasing difficulty, so, if you find the first ones too easy, at least take a look at the last one, which uses a technique that I truly believe you won’t have seen before.

Okay, let’s get started.

I love this puzzle because of how short and simple the final query is, even though it deals with many edge cases. The data for this challenge shows tickets moving in between Kanban stages, and the objective is to find how long, on average, tickets stay in the Doing stage.

The data contains the ID of the ticket, the date the ticket was created, the date of the move, and the “from” and “to” stages of the move. The stages present are New, Doing, Review, and Done.

Some things you need to know (edge cases):

  • Tickets can move backwards, meaning tickets can go back to the Doing stage.
  • You should not include tickets that are still stuck in the Doing stage, as there is no way to know how long they will stay there for.
  • Tickets are not always created in the New stage.
```SQL

CREATE TABLE ticket_moves (
    ticket_id INT NOT NULL,
    create_date DATE NOT NULL,
    move_date DATE NOT NULL,
    from_stage TEXT NOT NULL,
    to_stage TEXT NOT NULL
);

```
```SQL

INSERT INTO ticket_moves (ticket_id, create_date, move_date, from_stage, to_stage)
    VALUES
        -- Ticket 1: Created in "New", then moves to Doing, Review, Done.
        (1, '2024-09-01', '2024-09-03', 'New', 'Doing'),
        (1, '2024-09-01', '2024-09-07', 'Doing', 'Review'),
        (1, '2024-09-01', '2024-09-10', 'Review', 'Done'),
        -- Ticket 2: Created in "New", then moves: New → Doing → Review → Doing again → Review.
        (2, '2024-09-05', '2024-09-08', 'New', 'Doing'),
        (2, '2024-09-05', '2024-09-12', 'Doing', 'Review'),
        (2, '2024-09-05', '2024-09-15', 'Review', 'Doing'),
        (2, '2024-09-05', '2024-09-20', 'Doing', 'Review'),
        -- Ticket 3: Created in "New", then moves to Doing. (Edge case: no subsequent move from Doing.)
        (3, '2024-09-10', '2024-09-16', 'New', 'Doing'),
        -- Ticket 4: Created already in "Doing", then moves to Review.
        (4, '2024-09-15', '2024-09-22', 'Doing', 'Review');
```

A summary of the data:

  • Ticket 1: Created in the New stage, moves normally to Doing, then Review, and then Done.
  • Ticket 2: Created in New, then moves: New → Doing → Review → Doing again → Review.
  • Ticket 3: Created in New, moves to Doing, but it is still stuck there.
  • Ticket 4: Created in the Doing stage, moves to Review afterward.

It might be a good idea to stop for a bit and think how you would deal with this. Can you find out how long a ticket stays on a single stage?

Honestly, this sounds intimidating at first, and it looks like it will be a nightmare to deal with all the edge cases. Let me show you the full solution to the problem, and then I will explain what is happening afterward.

```SQL

WITH stage_intervals AS (
    SELECT
        ticket_id,
        from_stage,
        move_date 
        - COALESCE(
            LAG(move_date) OVER (
                PARTITION BY ticket_id 
                ORDER BY move_date
            ), 
            create_date
        ) AS days_in_stage
    FROM
        ticket_moves
)
SELECT
    SUM(days_in_stage) / COUNT(DISTINCT ticket_id) as avg_days_in_doing
FROM
    stage_intervals
WHERE
    from_stage = 'Doing';
```

The first CTE uses the LAG function to find the previous move of the ticket, which will be the time the ticket entered that stage. Calculating the duration is as simple as subtracting the previous date from the move date.

What you should notice is the use of the COALESCE in the previous move date. What that does is that if a ticket doesn’t have a previous move, then it uses the date of creation of the ticket. This takes care of the cases of tickets being created directly into the Doing stage, as it still will properly calculate the time it took to leave the stage.

This is the result of the first CTE, showing the time spent in each stage. Notice how the Ticket 2 has two entries, as it visited the Doing stage in two separate occasions.

With this done, it’s just a matter of getting the average as the SUM of total days spent in doing, divided by the distinct number of tickets that ever left the stage. Doing it this way, instead of simply using the AVG, makes sure that the two rows for Ticket 2 get properly accounted for as a single ticket.

Not so bad, right?

The goal of this second challenge is to find the most recent contract sequence of every employee. A break of sequence happens when two contracts have a gap of more than one day between them. 

In this dataset, there are no contract overlaps, meaning that a contract for the same employee either has a gap or ends a day before the new one starts.

```SQL
CREATE TABLE contracts (
    contract_id integer PRIMARY KEY,
    employee_id integer NOT NULL,
    start_date date NOT NULL,
    end_date date NOT NULL
);

INSERT INTO contracts (contract_id, employee_id, start_date, end_date)
VALUES 
    -- Employee 1: Two continuous contracts
    (1, 1, '2024-01-01', '2024-03-31'),
    (2, 1, '2024-04-01', '2024-06-30'),
    -- Employee 2: One contract, then a gap of three days, then two contracts
    (3, 2, '2024-01-01', '2024-02-15'),
    (4, 2, '2024-02-19', '2024-04-30'),
    (5, 2, '2024-05-01', '2024-07-31'),
    -- Employee 3: One contract
    (6, 3, '2024-03-01', '2024-08-31');
```

As a summary of the data:

  • Employee 1: Has two continuous contracts.
  • Employee 2: One contract, then a gap of three days, then two contracts.
  • Employee 3: One contract.

The expected result, given the dataset, is that all contracts should be included except for the first contract of Employee 2, which is the only one that has a gap.

Before explaining the logic behind the solution, I would like you to think about what operation can be used to join the contracts that belong to the same sequence. Focus only on the second row of data, what information do you need to know if this contract was a break or not?

I hope it’s clear that this is the perfect situation for window functions, again. They are incredibly useful for solving problems like this, and understanding when to use them helps a lot in finding clean solutions to problems.

First thing to do, then, is to get the end date of the previous contract for the same employee with the LAG function. Doing that, it’s simple to compare both dates and check if it was a break of sequence.

```SQL
WITH ordered_contracts AS (
    SELECT
        *,
        LAG(end_date) OVER (PARTITION BY employee_id ORDER BY start_date) AS previous_end_date
    FROM
        contracts
),
gapped_contracts AS (
    SELECT
        *,
        -- Deals with the case of the first contract, which won't have
        -- a previous end date. In this case, it's still the start of a new
        -- sequence.
        CASE WHEN previous_end_date IS NULL
            OR previous_end_date < start_date - INTERVAL '1 day' THEN
            1
        ELSE
            0
        END AS is_new_sequence
    FROM
        ordered_contracts
)
SELECT * FROM gapped_contracts ORDER BY employee_id ASC;
```

An intuitive way to continue the query is to number the sequences of each employee. For example, an employee who has no gap, will always be on his first sequence, but an employee who had 5 breaks in contracts will be on his 5th sequence. Funnily enough, this is done by another window function.

```SQL
--
-- Previous CTEs
--
sequences AS (
    SELECT
        *,
        SUM(is_new_sequence) OVER (PARTITION BY employee_id ORDER BY start_date) AS sequence_id
FROM
    gapped_contracts
)
SELECT * FROM sequences ORDER BY employee_id ASC;
```

Notice how, for Employee 2, he starts his sequence #2 after the first gapped value. To finish this query, I grouped the data by employee, got the value of their most recent sequence, and then did an inner join with the sequences to keep only the most recent one.

```SQL
--
-- Previous CTEs
--
max_sequence AS (
    SELECT
        employee_id,
        MAX(sequence_id) AS max_sequence_id
FROM
    sequences
GROUP BY
    employee_id
),
latest_contract_sequence AS (
    SELECT
        c.contract_id,
        c.employee_id,
        c.start_date,
        c.end_date
    FROM
        sequences c
        JOIN max_sequence m ON c.sequence_id = m.max_sequence_id
            AND c.employee_id = m.employee_id
        ORDER BY
            c.employee_id,
            c.start_date
)
SELECT
    *
FROM
    latest_contract_sequence;
```

As expected, our final result is basically our starting query just with the first contract of Employee 2 missing! 

Finally, the last puzzle — I’m glad you made it this far. 

For me, this is the most mind-blowing one, as when I first encountered this problem I thought of a completely different solution that would be a mess to implement in SQL.

For this puzzle, I’ve changed the context from what I had to deal with for my job, as I think it will make it easier to explain. 

Imagine you’re a data analyst at an event venue, and you’re analyzing the talks scheduled for an upcoming event. You want to find the time of day where there will be the highest number of talks happening at the same time.

This is what you should know about the schedules:

  • Rooms are booked in increments of 30min, e.g. from 9h-10h30.
  • The data is clean, there are no overbookings of meeting rooms.
  • There can be back-to-back meetings in a single meeting room.

Meeting schedule visualized (this is the actual data). 

```SQL
CREATE TABLE meetings (
    room TEXT NOT NULL,
    start_time TIMESTAMP NOT NULL,
    end_time TIMESTAMP NOT NULL
);

INSERT INTO meetings (room, start_time, end_time) VALUES
    -- Room A meetings
    ('Room A', '2024-10-01 09:00', '2024-10-01 10:00'),
    ('Room A', '2024-10-01 10:00', '2024-10-01 11:00'),
    ('Room A', '2024-10-01 11:00', '2024-10-01 12:00'),
    -- Room B meetings
    ('Room B', '2024-10-01 09:30', '2024-10-01 11:30'),
    -- Room C meetings
    ('Room C', '2024-10-01 09:00', '2024-10-01 10:00'),
    ('Room C', '2024-10-01 11:30', '2024-10-01 12:00');
```

The way to solve this is using what is called a Sweep Line Algorithm, or also known as an event-based solution. This last name actually helps to understand what will be done, as the idea is that instead of dealing with intervals, which is what we have in the original data, we deal with events instead.

To do this, we need to transform every row into two separate events. The first event will be the Start of the meeting, and the second event will be the End of the meeting.

```SQL
WITH events AS (
  -- Create an event for the start of each meeting (+1)
  SELECT 
    start_time AS event_time, 
    1 AS delta
  FROM meetings
  UNION ALL
  -- Create an event for the end of each meeting (-1)
  SELECT 
   -- Small trick to work with the back-to-back meetings (explained later)
    end_time - interval '1 minute' as end_time,
    -1 AS delta
  FROM meetings
)
SELECT * FROM events;
```

Take the time to understand what is happening here. To create two events from a single row of data, we’re simply unioning the dataset on itself; the first half uses the start time as the timestamp, and the second part uses the end time.

You might already notice the delta column created and see where this is going. When an event starts, we count it as +1, when it ends, we count it as -1. You might even be already thinking of another window function to solve this, and you’re actually right!

But before that, let me just explain the trick I used in the end dates. As I don’t want back-to-back meetings to count as two concurrent meetings, I’m subtracting a single minute of every end date. This way, if a meeting ends and another starts at 10h30, it won’t be assumed that two meetings are concurrently happening at 10h30.

Okay, back to the query and yet another window function. This time, though, the function of choice is a rolling SUM.

```SQL
--
-- Previous CTEs
--
ordered_events AS (
  SELECT
    event_time,
    delta,
    SUM(delta) OVER (ORDER BY event_time, delta DESC) AS concurrent_meetings
  FROM events
)
SELECT * FROM ordered_events ORDER BY event_time DESC;
```

The rolling SUM at the Delta column is essentially walking down every record and finding how many events are active at that time. For example, at 9 am sharp, it sees two events starting, so it marks the number of concurrent meetings as two!

When the third meeting starts, the count goes up to three. But when it gets to 9h59 (10 am), then two meetings end, bringing the counter back to one. With this data, the only thing missing is to find when the highest value of concurrent meetings happens.

```SQL
--
-- Previous CTEs
--
max_events AS (
  -- Find the maximum concurrent meetings value
  SELECT 
    event_time, 
    concurrent_meetings,
    RANK() OVER (ORDER BY concurrent_meetings DESC) AS rnk
  FROM ordered_events
)
SELECT event_time, concurrent_meetings
FROM max_events
WHERE rnk = 1;
```

That’s it! The interval of 9h30–10h is the one with the largest number of concurrent meetings, which checks out with the schedule visualization above!

This solution looks incredibly simple in my opinion, and it works for so many situations. Every time you are dealing with intervals now, you should think if the query wouldn’t be easier if you thought about it in the perspective of events.

But before you move on, and to really nail down this concept, I want to leave you with a bonus challenge, which is also a common application of the Sweep Line Algorithm. I hope you give it a try!

Bonus challenge

The context for this one is still the same as the last puzzle, but now, instead of trying to find the period when there are most concurrent meetings, the objective is to find bad scheduling. It seems that there are overlaps in the meeting rooms, which need to be listed so it can be fixed ASAP.

How would you find out if the same meeting room has two or more meetings booked at the same time? Here are some tips on how to solve it:

  • It’s still the same algorithm.
  • This means you will still do the UNION, but it will look slightly different.
  • You should think in the perspective of each meeting room.

You can use this data for the challenge:

```SQL
CREATE TABLE meetings_overlap (
    room TEXT NOT NULL,
    start_time TIMESTAMP NOT NULL,
    end_time TIMESTAMP NOT NULL
);

INSERT INTO meetings_overlap (room, start_time, end_time) VALUES
    -- Room A meetings
    ('Room A', '2024-10-01 09:00', '2024-10-01 10:00'),
    ('Room A', '2024-10-01 10:00', '2024-10-01 11:00'),
    ('Room A', '2024-10-01 11:00', '2024-10-01 12:00'),
    -- Room B meetings
    ('Room B', '2024-10-01 09:30', '2024-10-01 11:30'),
    -- Room C meetings
    ('Room C', '2024-10-01 09:00', '2024-10-01 10:00'),
    -- Overlaps with previous meeting.
    ('Room C', '2024-10-01 09:30', '2024-10-01 12:00');
```

If you’re interested in the solution to this puzzle, as well as the rest of the queries, check this GitHub repo.

The first takeaway from this blog post is that window functions are overpowered. Ever since I got more comfortable with using them, I feel that my queries have gotten so much simpler and easier to read, and I hope the same happens to you.

If you’re interested in learning more about them, you would probably enjoy reading this other blog post I’ve written, where I go over how you can understand and use them effectively.

The second takeaway is that these patterns used in the challenges really do happen in many other places. You might need to find sequences of subscriptions, customer retention, or you might need to find overlap of tasks. There are many situations when you will need to use window functions in a very similar fashion to what was done in the puzzles.

The third thing I want you to remember is about this solution to using events besides dealing with intervals. I’ve looked at some problems I solved a long time ago that I could’ve used this pattern on to make my life easier, and unfortunately, I didn’t know about it at the time.


I really do hope you enjoyed this post and gave a shot to the puzzles yourself. And I’m sure that if you made it this far, you either learned something new about SQL or strengthened your knowledge of window functions! 

Thank you so much for reading. If you have questions or just want to get in touch with me, don’t hesitate to contact me at mtrentz.com.

All images by the author unless stated otherwise.

Shape
Shape
Stay Ahead

Explore More Insights

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy,  bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

Shape

Cloud overspend is an accountability problem not a tooling gap

He added that neither self-service nor automation fixes the issues. “Without redesign, you get responsibility without leverage,” he said. “So, governance, cost management, and security practices, if they remain rooted in sort of pre-cloud thinking, that’s going to create tension between the control mechanisms that we have in place and the

Read More »

How an outdated oil benchmark erodes U.S. pricing power

The United States is the world’s largest oil producer backed by the deepest capital markets. Yet the financial system still voluntarily prices American crude off a landlocked benchmark that doesn’t fully account for where the oil is produced or who buys it. This carries structural consequences far beyond the energy sector that threaten U.S. borrowing costs, weaken sanctions power, and chip away at the dollar’s status as the world’s reserve currency. The greenback’s share of global reserves has already slid from over 70% in 2000 to around 58%. While this has been driven by many macro factors, the trend accelerates as more oil producers bypass the dollar system and non-dollar payment channels develop. The NYMEX WTI futures contract reflects an era when the U.S. was the world’s largest crude importer. The shale boom reversed that dynamic, routing Permian Basin production straight to Gulf Coast export terminals. Today, the interplay between global crude demand and Permian Basin production is captured by the ICE Midland WTI futures contract (HOU), a price marker guaranteeing on-spec Midland WTI crude delivered straight to Houston. Meanwhile, NYMEX WTI’s market share of global crude oil futures benchmark volume has dropped from 62% in 2017 to 40% today as Cushing is supported by institutional inertia via long-term contract pricing, commodity index inclusion, and price-reporting agencies being resistant to change. When NYMEX WTI diverges from Midland WTI on the Gulf Coast (priced via HOU) it distorts the price signal for U.S. exports and forces international buyers to absorb hedging inefficiencies. The physical market has been moving away from Cushing for years, pricing Gulf Coast exports against Brent or using multi-instrument trades. Using these workarounds introduces unnecessary basis risk and execution costs, risking America’s reputation as standard for transparent energy price discovery. This inertia is arguably more damaging than external

Read More »

Diamondback to acquire remaining Double Eagle Midland basin interest in multibillion cash, stock deal

The deal continues Diamondback’s long-running consolidation strategy in the basin. Andrew Dittmar, principal analyst at Enverus Intelligence Research (EIR), said the company has emerged as “the most consistent consolidator in the Midland basin” through a series of deals that have expanded its inventory position, noting Diamondback has spent more than $40 billion on acquisitions since 2020 and now trails only ExxonMobil in ownership of Midland basin inventory and owns more than two times the locations of the third-place exploration and production company.  Dittmar said the latest deal strengthens Diamondback’s exposure to the Barnett-Woodford, which he described as “the largest undeveloped oil expansion opportunity in US shale.” Along with news of the acquisition, Van’t Hof reported early results from the operator’s first full-section Barnett development in the Spanish Trail area of Midland County. The project consists of four 10,000-ft lateral wells that at peak produced about 7,500 b/d of oil and 37 MMcfd of natural gas, exceeding internal expectations and supporting plans for broader Barnett development beginning in 2027, he said.  Diamondback expects the transaction to close Dec. 21, 2026, subject to customary closing conditions.

Read More »

ExxonMobil Guyana starts production at Uaru

@import url(‘https://fonts.googleapis.com/css2?family=Inter:wght@100..900&display=swap’); .ebm-page__main h1, .ebm-page__main h2, .ebm-page__main h3, .ebm-page__main h4, .ebm-page__main h5, .ebm-page__main h6 { font-family: Inter; } body { line-height: 150%; letter-spacing: 0.025em; } button, .ebm-button-wrapper { font-family: Inter; } .label-style { text-transform: uppercase; color: var(–color-grey); font-weight: 600; font-size: 0.75rem; } .caption-style { font-size: 0.75rem; color: color-mix(in srgb, currentColor 60%, transparent); } ExxonMobil Guyana Ltd. started production at Uaru, raising Guyana’s installed crude oil production capacity to 1.1 million b/d. With production now underway, the Errea Wittu FPSO is slated to achieve its initial production capacity of 250,000 b/d in first-quarter 2027. Equipped with a storage capacity of 2 million bbl, Errea Wittu becomes the fifth FPSO operating in Stabroek block. ExxonMobil Guyana operates Stabroek block (45%) on behalf of partners Hess Guyana Exploration Ltd. (30%), and CNOOC Petroleum Guyana Ltd. (25%).

Read More »

China’s oil demand falls 7.1% as economic weakness deepens

In September, China’s oil demand fell sharply by 1.3 million b/d year-on-year (y-o-y), a 7.1% decline, according to latest data from the International Energy Agency (IEA). This drop is nearly double the decline seen in August and only slightly less than the 1.4-million b/d slump recorded in June following the Iran-related conflict. According to IEA data, the decline in demand has affected virtually all categories of petroleum products. Demand for liquefied petroleum gas (LPG)/ethane, naphtha, gasoline, diesel, and fuel oil each fell by about 200,000-250,000 b/d. In percentage terms, naphtha demand dropped by 11%—nearly twice the 6% decline seen in LPG/ethane, another petrochemical feedstock. This reflects an ongoing shift in the feedstock mix for steam crackers, which are increasingly favoring LPG and ethane over naphtha. Demand for gasoline and diesel, the primary fuels for transportation, both fell by about 5% y-o-y. China’s transition toward the electrification of passenger vehicles and trucks is far exceeding that of any other nations, a trend further accelerated by surging fuel prices, IEA said. Retail prices for gasoline and diesel in China have risen by about 25% y-o-y, according to GlobalPetrolPrices, hovering just below the all-time highs reached in mid-2022. Meanwhile, sluggish economic growth continues to weigh on oil consumption and overall demand. The downturn in China’s real estate market has entered its sixth year without yet hitting bottom, persistently eroding household wealth and undermining economic confidence. In the first 9 months of 2026, real estate investment fell by 20.6% y-o-y—an acceleration from the 19.9% ​​decline recorded between January-August. The latest data for October show that economic momentum in industrial output and retail sales continues to weaken. This prolonged slump has severely dampened public sentiment; since the beginning of 2026, the consumer confidence index has remained near post-pandemic lows. Exports have played a growing role

Read More »

People and Vertical AI: The responsible and trusted framework for energy and water

I have always held a simple conviction — energy and water utilities should not be spectators to the AI transformation of their own industry. They should shape it, together with the partners and technologists building alongside them. That conviction matters because few industries carry the weight that energy and water do. They are the services every other part of modern life is built on — be it homes, hospitals, schools, businesses, data centers and increasingly, the AI economy itself. When they work, they are invisible. When they fail, everything else stops. Their success is measured in three outcomes the public feels directly: affordability, reliability and trust. Rising energy costs, extreme weather, electrification and the growing demand from large loads are putting these outcomes under increasing pressure. At the same time, utilities are being asked to modernize operations, make better use of data and respond faster to changing customer and grid demands. This is changing the conversation around AI. The question is no longer simply what AI can do, but where it can deliver value, how much autonomy it should have and what safeguards must govern its use. At this year’s WE3 Global AI Summit, conversations among energy and utility leaders reflected this shift. As AI moves from individual applications into customer, workforce, business and grid operations, the industry must think beyond capability to the conditions under which AI can be trusted to act. My belief is that the answer lies in People + Vertical AI: an approach that combines industry-specific intelligence with human experience, judgment and accountability. Vertical AI brings the data, domain knowledge and operational context required to work within the realities of energy and water. People establish its purpose, define its boundaries and remain accountable for its outcomes. Four principles should guide this approach. 1. Convergence: Bringing customers

Read More »

People and Vertical AI: The new operating model for energy and utilities

For generations, the energy and utility industry has depended on the judgment of its people.  Across the industry, people are responsible for keeping energy and water running and serving the communities that depend on them. A field crew is there when a community needs power restored. A grid operator carries the responsibility of keeping the network stable and reliable. A customer service representative is often the human connection between a customer and the utility service. An engineer makes decisions that can shape the safety, resilience and reliability of infrastructure for decades.  That expertise remains essential. What has changed is the complexity surrounding it.  Electrification, distributed energy resources, renewable generation, electric vehicles, extreme weather, aging infrastructure, affordability pressures, rising customer expectations and data centers becoming new energy consumers are all connecting decisions that were once managed independently.  The industry has spent decades digitizing those decisions, generating more data along the way. The challenge now is different: how do you make that intelligence work together and make the expertise of the people using it scalable?  This is where AI has to evolve beyond generic intelligence to a Vertical AI approach built around the specific context, complexity and interconnected realities of the energy and utility industry.  The opportunity is not to automate the people who run the industry. It is to give them better context, better intelligence and the ability to extend their judgment across more decisions, more customers and more of the ecosystem than was previously possible.  That is the premise behind People + Vertical AI.  From Human Expertise to Connected Intelligence  People + Vertical AI is not a technology layer placed on top of existing operations. It is an operating model in which Vertical AI works alongside people, connecting information across the enterprise while people provide judgment, accountability and context.  The

Read More »

Data center boom faces challenges beyond water and power

According to a report published by BBC Science Focus, the United States had more than 4,000 data centers by the end of 2025, representing approximately 50 gigawatts of capacity. Meanwhile, U.S. data centers consumed approximately 66 billion liters of water in 2023, highlighting the substantial environmental footprint of the industry. These resource requirements have driven considerable opposition to new developments. A recent NBC News poll cited by the publication found that 69% of respondents would oppose construction of an AI data center in their neighborhoods. This is reflected in the recent elections, where pro-data center candidates lost badly while those who publicly opposed for the construction of data centers did well at the ballot box. Data centers can provide economic benefits to the communities where they operate, including increased property tax revenues, permanent employment and infrastructure improvements. Developers sometimes finance upgrades to electrical substations, roads and emergency services to accommodate their facilities.

Read More »

How TI’s ADC34DSO75 Catches the Nanosecond Power Glitches Behind AI Data Center Downtime

Today’s advanced AI data center can be viewed as a vast, complex and interdependent digital organism in which a virtually infinite number of components must be tracked all the time, in real time. It’s a monitoring job of staggering proportions. At stake is the operation of highly critical workloads and the maintenance of return on investment for high-cost compute equipment. At risk is sub-optimal performance or, in worst-case scenarios, the nightmare of data center downtime. Sponsored Resources: Why traditional telemetry misses nanosecond power-rail events Telemetry has long been relied upon to monitor data center servers, test and measurement equipment, wireless infrastructure and other systems for real-time information about voltage, current, power, temperature and overall system health. But most systems operate at millisecond-level intervals or slower, which is six orders of magnitude longer than the transient events that now occur at the processor power rails and the tools meant to observe them. The requirements and high stakes of today’s AI data centers and workloads have called for a new generation of advanced, nanosecond-level telemetry gear. Texas Instruments has established a leadership position in this sector. The company has developed the first analog-to-digital converter (ADC) to enable 24/7, low-power, event-based remote monitoring at nanosecond resolution. Let’s unpack this statement and examine why TI’s new monitoring technology is so important for the data center industry. What is the TI ADC34DSO75?    We’re talking specifically about TI’s ADC34DSO75, the industry’s highest-performance, 3 gigasamples per second (GSPS) ADC, delivering thousands of times faster performance than current offerings. Launched at the Open Compute Project Global Summit conference in October, the ADC34DSO75 addresses the shortcomings of existing monitoring solutions that can’t detect the nanosecond-level power-rail glitches that cause costly data center downtime. The ADC34DSO75 combines ultra-high-speed sampling with advanced on-chip digital signal processing in a small form factor,

Read More »

Schneider Electric’s AI Factory Strategy Moves Beyond the Modular Data Hall

The race to industrialize AI data center construction is moving beyond prefabricated electrical rooms and modular building components. Increasingly, major infrastructure suppliers are attempting to standardize the power, cooling, controls, and engineering systems surrounding the computing equipment itself, with the goal of delivering highly integrated infrastructure through repeatable manufacturing processes rather than site-specific construction. Schneider Electric’s recent activities provide a revealing example of that transition. Its RapidScale Modular Data Center Reference Design 115, introduced October 21 at NVIDIA GTC Berlin, outlines a modular AI factory architecture engineered for the NVIDIA Vera Rubin platform. Separately, a study conducted with Compass Datacenters documents substantial reductions in selected onsite installation activities through prefabricated white-space modules. Related developments involving power-system certification, onsite generation, and digital-twin technology show how the company is extending its reach across the data center development process. As DCF previously reported in its examination of NVIDIA’s DSX POD initiative, the broader industry is working toward factory-integrated computing environments that can be assembled, tested, and delivered with less dependence on conventional field construction. Schneider’s contribution is particularly interesting because its latest engineering documentation provides a detailed view of how electrical and thermal infrastructure might accommodate a new generation of exceptionally power-dense AI systems. A 227-kW Building Block for Vera Rubin Schneider describes Reference Design 115 as a 10–20 MW, Tier III-targeted modular AI factory architecture developed to ANSI standards. The published example concentrates on one deployment unit containing four scalable units, designed around four NVIDIA Vera Rubin clusters with 1,008 GPUs each. That translates to 4,032 GPUs distributed across 56 AI computing racks, supported by another 32 networking racks. The maximum AI rack power reaches 227 kW, while the documented configurations support approximately 10.4 MW to 11.4 MW of active IT load. The reference design assigns approximately 33,518 square feet to the

Read More »

Cisco forecasts $9 billion in AI infrastructure revenue in FY27

Cisco continues to benefit from a “networking supercycle,” reporting record revenue—driven by demand for AI infrastructure, security, and campus refreshes from hyperscalers and enterprise customers—for its fiscal first quarter ended October 24. “In Q1, we delivered record quarterly revenue of $18.5 billion, up 24% year over year, and record non-GAAP earnings per share of $1.40, up 40% year over year. These are our strongest top andbottom-line growth rates in over 15 years,” Cisco CEO Chuck Robbins said on the vendor’s quarterly financial call. “Networking product orders grew 28% in Q1, delivering our ninth consecutive quarter of double-digit growth. In fact, we saw double-digit growth across every product family: service provider routing, compute, enterprise routing, wireless, data center switching and campus switching,” Robbins said. “These broad-based results support our view that we are in the early stages of a multiyear, multibillion-dollar networking supercycle.”

Read More »

How TI’s ADC34DSO75 Catches the Nanosecond Power Glitches Behind AI Data Center Downtime

Today’s advanced AI data center can be viewed as a vast, complex and interdependent digital organism in which a virtually infinite number of components must be tracked all the time, in real time. It’s a monitoring job of staggering proportions. At stake is the operation of highly critical workloads and the maintenance of return on investment for high-cost compute equipment. At risk is sub-optimal performance or, in worst-case scenarios, the nightmare of data center downtime. Sponsored Resources: Why traditional telemetry misses nanosecond power-rail events Telemetry has long been relied upon to monitor data center servers, test and measurement equipment, wireless infrastructure and other systems for real-time information about voltage, current, power, temperature and overall system health. But most systems operate at millisecond-level intervals or slower, which is six orders of magnitude longer than the transient events that now occur at the processor power rails and the tools meant to observe them. The requirements and high stakes of today’s AI data centers and workloads have called for a new generation of advanced, nanosecond-level telemetry gear. Texas Instruments has established a leadership position in this sector. The company has developed the first analog-to-digital converter (ADC) to enable 24/7, low-power, event-based remote monitoring at nanosecond resolution. Let’s unpack this statement and examine why TI’s new monitoring technology is so important for the data center industry. What is the TI ADC34DSO75?    We’re talking specifically about TI’s ADC34DSO75, the industry’s highest-performance, 3 gigasamples per second (GSPS) ADC, delivering thousands of times faster performance than current offerings. Launched at the Open Compute Project Global Summit conference in October, the ADC34DSO75 addresses the shortcomings of existing monitoring solutions that can’t detect the nanosecond-level power-rail glitches that cause costly data center downtime. The ADC34DSO75 combines ultra-high-speed sampling with advanced on-chip digital signal processing in a small form factor,

Read More »

SC26 Preview: AI-HPC Convergence Takes Center Stage in Chicago

As the high-performance computing community prepares to gather in Chicago for SC26, November 15–20, the latest episode of the Data Center Frontier Show offers a timely preview of the technologies, market developments, and infrastructure challenges expected to shape this year’s conference. Hosted by DCF Contributing Editor Doug Black, the conversation brings together Lois McKinnis, Argonne Distinguished Fellow and senior computational scientist at Argonne National Laboratory; Addison Snell, founder and CEO of Intersect 360 Research; and Earl Joseph, CEO of Hyperion Research. The episode also inaugurates Data Center Frontier’s new Frontier Computing channel, expanding DCF’s coverage of the technologies and infrastructure driving the convergence of high-performance computing, AI, QC and next-generation scientific discovery. The central theme is unmistakable. AI is no longer simply an adjunct to traditional supercomputing. It is changing how scientific computing itself is performed. “We’re really moving beyond focusing just on bigger machines, faster machines,” McKinnis observes, pointing to the growing integration of AI into scientific software development, simulations, and research workflows. Joseph underscores the acceleration, citing new Hyperion research finding that 70% of respondents are already using AI for code generation. The panel also examines the U.S. Department of Energy’s Genesis Mission, a sweeping effort to link supercomputers, large-scale AI infrastructure, and cloud resources in support of scientific discovery. Joseph calls Genesis “the first time I’ve ever seen something that’s truly a broad national resource,” while Snell emphasizes that investments in AI infrastructure must be matched by investments in science itself. For data center professionals, the discussion turns to increasingly consequential questions about the economics and sustainability of the AI buildout. Snell warns that the industry’s focus on rapid construction has outpaced serious discussion of infrastructure lifecycles and upgrades. Joseph highlights the mounting difficulty of financing supercomputers costing $100 million or more when new GPU generations arrive

Read More »

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).  In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

Read More »

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

Read More »

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

Read More »

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle

Read More »