Last week’s challenge report:

I picked thex tea challenge because it barely felt like a challenge. I already drink too much tea, and I’m apparently still learning how to sit still.

I spent about 30 minutes doing nothing and wandering through my thoughts. Both of them.

For the “false” fact, I picked the obvious one!

My wife was surprised to learn that wombats poop little cubes. Her exact response: “Why are you telling me this?!”

I was more surprised by the reason: they use them to mark their territory, and the cubes can sit on top of rocks without rolling away. It is still a working hypothesis, but Nature apparently decided this was easier than letting wombats flatten the droppings themselves.

Now you know.

Ok, my picks for next week:

HUGE UPDATE: Jakub (the second person behind this newsletter. I would call him the better half, but we’re not married and my wife reads this) has decided to join the challenge!

He’ll be recording proof and posting Reels somewhere on social media, which means I’m now not entirely sure how long this challenge will last.

But let’s gooooo.

Report in a week. With a link to his socials, I guess.

I noticed something while preparing for a QBR.

I did not open a dashboard.

I did not ask the data team for a new report.

I opened a chat box and typed: “What changed since the last review?”

A few follow-up questions later, I had the trends, project updates, and risks.

If you use AI (which is becoming a fairly safe assumption) you may have noticed the same shift.

The interface matters less now.

I ask a question.

AI checks the information it can access.

Then I ask the next question.

Before, the data team built dashboards with fixed filters and charts. My job was to study them and decide what the numbers meant.

Now I ask:

No new dashboard.

And by data, I do not only mean revenue, conversion rates, or device volumes.

Data is also meeting notes, emails, project documents, and decisions.

This changed what I consider valuable.

Giving AI access to the right information matters most.

We use Microsoft for almost everything, so Copilot is officially our friend now.

I activate it in around 90 percent of the Teams meetings I attend.

It transcribes the conversation, creates notes, and identifies action items.

I still keep a notebook and pen in front of me.

Partly to think. Partly to look like I am thinking.

Someone says something important.

I write down a sentence or two. Quietly.

Writing makes the silence look intentional.

The other person finishes speaking.

I am still writing.

They wait.

The silence becomes uncomfortable, so they keep talking.

That is often when I might get a useful part.

After the meeting, my paranoia kicks in, and I copy the AI-generated notes into what I call the vault.

“Vault” is a vague term because it does not have one specific destination.

The key part is OneDrive.

Why?

Copilot, of course.

When I feel like it is time to create some value for our shareholders, I ask: “Give me a summary of my commitments”.

Copilot checks the notes and documents in OneDrive, and even relevant email conversations in Outlook.

Then it gives me a bullet-point list of what I promised and by when.

Now I have my work plan ready. I would love to argue with some of it, but the meeting notes are quite specific.

Preparation for the next meeting has also become much easier: “Give me a summary of the previous meetings so we can check the status and see what has changed.”

Done.

No searching through old notes and trying to remember what did we agree on.

Pro-tip: Copilot in Excel is a game changer.

Now I do not need to remember every formula (did I say remember? More like search for a tutorial) and apply it before I can start working with the data.

I can simply ask: “Analyze the trend and build a forecast through 2045”.

Or: “Redesign this table, group the months into quarters, swap the axes, and put the timeline on the horizontal axis.”

Then I check the result. Because Copilot is still a Minion, not Gru.

Is there life outside of Copilot?

Apparently, yes.

Corporate rule: our Engineering team document everything in Confluence.

Everything.

Projects, milestones, roadmaps, issues, decisions, updates, ownership...

And luckily we have an IT-approved AI tool that can access all of it.

In my case, that is Claude Sonnet or Opus.

So I ask: “I am preparing for a QBR with ChatGPT leadership (a random company name - who would name anything like that?!). Pull a comprehensive status update of the projects, progress, roadblocks, and risks”.

A few minutes later, I follow up with: “Analyze the content, score it, and decide whether it needs improvement”.

The second round nails it.

Magic!

For slides with graphs (because executives apparently prefer graphs to words), I tap into Amazon QuickSight - our data warehouse.

There are a bunch of prebuilt dashboards.

I mostly ignore them.

Because QuickSight has an AI agent too!

So I ask: “Analyze the data from the past two years and identify any anomalies in the latest quarter”.

Now, instead of reporting what already happened, we can spend the QBR discussing why it happened.

And then comes the creative part of the job.

PowerPoint slides.

Once the content is collected, I place it randomly across several slides and hope nobody sees the original version.

Guess what!

PowerPoint has Copilot too.

It can apply the corporate template, organize the content, and add a little aesthetics: “Apply the corporate template and improve the layout without changing the meaning”.

PowerPoint Copilot cleans up whatever I put on the slides.

Sometimes, it even exceeds expectations.

My key takeaway?

AI cuts a week of hard work and preparation down to two hours.

Most of those two hours are spent waiting for AI to finish.

Till next time.

9 Trends Shaping Work in 2026 and Beyond - 10 mins read.

Does your company have an AI strategy that nobody can explain? This will feel familiar. HBR looks at premature layoffs, mental strain, and the growing gap between executive expectations and actual results. Now you just need to figure out how to get past the paywall.

Stigmergy: the most important concept you’ve never heard of - 10 min read.

I had never heard the word “stigmergy,” which is odd because I rely on it all the time. It’s how ants, Wikipedia editors, and good teams coordinate without a weekly sync.

What chess’s “intermezzo” moves can teach us about making better life decisions - 6 min read.

I like any idea that gives me permission not to make a huge decision immediately. This article borrows the chess move “intermezzo” to explain why an in-between step can buy you time and better information.

Julia Minson is a Harvard behavioral scientist who has spent years studying why disagreements go wrong. She explains how to make people feel heard, consider opposing views, and have a useful conversation without pretending everyone is right.

I have not read it. I am not even planning to.

But I strongly recommend it to nearly everyone I know. Please read it. This is personal.

Simple. Minimal. Just tea and 78 keys (I didn’t count. Google did, so blame google). The screen is a pleasant bonus.

Work-With-Me Manual: Kaz Nejatian keeps a ten-point guide to working with him. It is roughly a decade old and, by his count, has changed by only about twelve words. Brilliant!

Seventy-Five Percent Trust Start: Nejatian says new colleagues begin at roughly seventy-five percent trust, compared with about fifty percent for Tobi Lütke (co-founder of Shopify), but his assessment changes quickly when their work provides new evidence. Give meaningful responsibility early, then judge what happens.

Missing-Skills Pairing: Nejatian worked with Giang Hoang at Shopify and Opendoor because Hoang was strong where he was weak; he used a similar pairing at Facebook. Instead of trying to fix every weakness himself, he delegates work to someone already good at it.

One-Week, Ten-Year Planning: Nejatian focuses on what can be validated or shipped this week and what the company should become in ten years; he finds quarters less useful for his own planning. Ask what can be finished now and whether it will still matter later.

Working Product Review: At Shopify, Nejatian expected teams to show the product through a demonstration or code instead of presenting polished slides. If the product cannot be shown, the presentation is not evidence that it works.

Complete First-Version Standard: Nejatian accepts limited scope in a first version, but it must still be complete, usable, and connected to a clear product direction. Cut features if needed, but not the parts that make the product usable.

Meeting Reapproval Test: On Shopify’s first working day of 2023, Nejatian canceled recurring meetings with three or more people and made managers wait two weeks before restoring any. About 12,000 events disappeared. An existing calendar invitation is not proof that a meeting is still needed.

Meeting Cost Check: Shopify showed the estimated employee cost on meeting invitations, let people decline or leave early, and kept one-on-ones. Nejatian preferred writing and code when possible. Before scheduling a meeting, show its cost and ask whether a written record would work instead.

Public Error Record: Nejatian periodically shared a company-wide list of his recent mistakes; during one thirty-day Slack review, his most frequent phrase to Tobi Lütke was “I disagree.” A mistake does not become private because a senior person made it.

Learning-Ability Hiring Test: Under Nejatian’s Shopify approach, marketers used GitHub and wrote some code, many salespeople learned SQL, and specialists could advance without becoming managers. Hire people willing to learn unfamiliar tools, and do not make management a condition for promotion.

Current-Facts Decision Rule: When Shopify reconsidered its logistics strategy, Nejatian says the decision to leave took roughly three to four weeks. Judge the choice using current facts; do not continue solely because the original decision once made sense.

Training Before Requirement: Before AI use became part of Opendoor job descriptions and reviews, Nejatian provided about ten weeks of practice with help from Cursor, Gumloop, and OpenAI, followed by a company hackathon. The requirement came after instruction and practice. A surprisingly sensible sequence.

We track steps, sleep, and calories, but rarely track whether life is actually going well. Chris Musser spent 90 seconds a day rating nine areas of his life and started noticing problems he had been ignoring.

Enjoying the newsletter? Please forward this issue to a friend who might enjoy it too 😊🙏🏻

It only takes 10 seconds. Making this one took us 9 hours…