🧑‍💻 Over-optimization is hurting you

PLUS office-to-apartment conversion case studies, AI in HR, and why remote employees should not take a pay cut

Hey Insider, happy Monday! Hope you had an incredible weekend 😊 Respond to this email and tell me what the last highlight of the last week was. I will select a few and share them in the next email so we can see what other Insiders are up to!

Now, getting to business…

Today in 5min 29sec you’ll learn:

  • 🚀 Why over-optimization is bad and how to fix it

  • 🏢 New interesting stats & case studies on office-to-apt conversions

  • 🤖 How AI could change the HR field

  • ➕ PLUS the CEOs bullish on remote work, remote manager’s survival guide, why remote employees should definitely NOT have to take a pay cut, and a little dose of internet nostalgia

Let’s dive in…


Here’s why you should stop trying to optimize everything

As knowledge workers, we are inundated with content that glorifies hustle and shows us how to optimize every corner of our lives from health to relationships and work.

But is there such a thing as being over-optimized?

As it turns out there is, and it can be VERY damaging…

In a new Every piece, Dan Shipper explains how over-optimizing actually creates a phenomenon in machine learning called “overfitting”, where the AI basically becomes really good at taking the test but does not function well at achieving its core objective.

We can see this in the real world with schools and standardized testing.

The idea behind a standardized test is to see how every kid is progressing in learning a core set of skills compared to the masses.

However, in some cases, teachers have focused on this too much and now teach the kids how to take the tests instead of actually learning the material.

These kids are overfitted.

They are optimized at taking the tests but are missing the real goal of learning. This can show up in many different ways in our lives.

Personally, I originally decided I wanted to start my own business because I wanted location independence so I could travel and experience other cultures.

At some point, I overoptimized on my goal of building a business and spent the majority of my travels becoming what a friend calls a “laptop zombie with different backgrounds”.

My goal was to build a business to gain location independence in order to travel, but when I started traveling I overoptimized on business building, spent the majority of my time working, and missed the original objective of immersive travel.

I was just changing locations!

So how do we make sure we don’t overoptimize?

We can again learn about this from machine learning. There are 3 ways to solve “overfitting” AKA overoptimization in ML:

  1. Early stopping - regularly stopping and checking if the model is training toward its true objective

  2. Introduce random noise - add random data into the model. In the standardized testing example, this would be like testing at random times of the year so teachers can’t plan & cram students for tests

  3. Regularization - in ML this means penalizing models for becoming too complex (the more complex a model is, the more likely they are to become overfitted).

If I were to apply these ML solutions to my goal in order to reduce the chances of overoptimizing I might do something like…

  1. Hold quarterly reflection sessions to check if my actions are aligned with my vision of using business for long-term travel (Early stopping)

  2. Schedule a travel-focused event every weekend at least a month in advance so I don’t get too focused on working and “forget” to actually do the fun travel stuff (Random noise)

  3. Create a rule of taking a purely "travel-focused” trip one quarter if I did not do that the previous quarter (Regularization)

Have you ever experienced over-optimization? If yes, feel free to share the situation 😊

These answers may be shared in a following email

Login or Subscribe to participate in polls.


How empty offices become apartments in the US

Office-to-apartment conversions have been a popular topic of discussion over the last 2 years and on the surface it sounds like a great idea:

We don’t have enough supply when it comes to affordable housing, and we have an oversupply of offices in a remote work world.

However, there are a few problems with these conversions:

  1. Not all office buildings can be turned into apartments due to a lack of lighting or infrastructure (plumbing).

  2. In order for a building to be converted, it must be totally empty, and even though we have a ton of open offices, it takes a single lease in a building to stop an office conversion from happening.

  3. Office-to-apartment conversions can be very expensive, which means that the newly built apartments won’t exactly be affordable.

  4. Many cities do not have favorable policies to incentivize office conversions.

Despite these problems, several cities in the US are making strides towards turning unused office buildings into apartments with Washington DC leading the way.

Office conversion alone however will not be enough to fulfill the housing demand expected in the US over the next decade, but it can certainly help in revitalizing downtowns and turning them into livable areas as opposed to 5 to 9 ghost towns.

Want to learn more and check out some interesting case studies? Watch this video


Will HR departments of the future be AI-powered?

Darren Murph, VP of Workplace Design & Remote Experience at Andela and ex-Head of Remote at GitLab, spoke at the recent Fortune Brainstorm Tech event and had a very interesting takeaway.

Successful remote-first companies have always placed importance on a well-developed handbook, or Single Source of Truth, that is routinely updated and contains all aspects of how the company functions.

But what happens when these handbooks become AI-powered?

Perhaps there will be a day, not so far in the future, when team members can simply explain to an AI what they want to achieve, and that AI can scan the contents of that handbook and then suggest the best action.

The use cases here are numerous and an AI system like this could create massive leverage for HR professionals and team leads. But as Darren mentions in this post, these AI systems will only be as good as the data in the handbook.


  • 🧑‍💻 The CEOs still bullish on remote work. There are many execs that are still bullish on remote work including from companies like HubSpot and Yelp and with good reason… it’s working!

  • 🐶 Remote Manager Survival Guide. Our friends at Kona published a 60 page report with case studies, expert interviews, and key stats to help leaders of remote teams thrive

  • 💸 No, remote employees should not take a pay cut. An Australian politician has suggested that if an employee wants to work remotely, they shoud take a pay cut since not all employees can WFH.

  • 🔍 Wiby is a search engine with lovely 2000s flavor. These days it seems like Google SERPs are filled with massive websites. Wiby on the other hand prefers smaller sites, and its “surprise me…” button allows you to get a taste of the internet’s past by bringing up randon Web1 pages & serving you a daily dose of nostalgia.


July 15-29: Nomad City Festival 2023 [📍Gran Canaria] - the original nomadic unconference presented by Repeople. Get inspired, learn, & make connections at the biggest nomad event in the Canary Islands.

Dec 7-13: Nomad Island Fest [📍 Madeira, Portugal] - Learn from world-class industry experts on topics like mindset, marketing, tech, wealth & health

More events coming soon!


Looking for a job that’s not just remote, but that you can work from anywhere in the world? We’ve done the heavy lifting for you. Check out these awesome WFA positions 👇

Thanks for spending a few minutes with us today! I’d love to hear what you think about this edition of the newsletter…

What do you think of today's edition?

Let us know what you thought of today's email so we can make our content as awesome as possible

Login or Subscribe to participate in polls.

If you are enjoying this newsletter consider sharing it with someone you think may like it 👇

Hope you have an awesome rest of your day! Cheers 🍻

Today’s email was written by Mitko Karshovski