Blog

It’s an opportunity to tinker.

To try something fresh.

To bet a little, but not the whole farm.

And to see if a new approach can beat the old way.

Everything is an experiment.

Always be experimenting.

When I’m sharing a piece of work, or an idea, with other people, I always ask three questions:

1. What’s wrong?
2. What’s missing?
3. What could be made simpler or clearer?

I find that with those three questions, more 90% of errors, issues, or bits of faulty thinking can be uncovered.

#1 is relatively easy. It’s not too difficult react to what is on the page, and then to inconsistent logic or numerical errors.

#3 is also straightforward. With training and practise, one can learn how to communicate with clarity.

But #2 is unquestionably the hardest task out of the three. Seeing what’s not on the page is hard. It’s difficult to imagine, consider, and articulate what is missing. And to do it well requires great skill and judgement.

When you find someone who can see what isn’t there, and help you spot issues before they arise, they’re a keeper.

I’ve likely written more than a dozen times over the past 4 years about momentum.

A lesson I continually learn and re-learn is ‘momentum begets momentum’.

Little actions, performed repeatedly, add up over time. And the small successes and habits eventually snowball into bigger ones.

It’s something I try and remind myself of both when times are easy and good, and when times are tough and trying.

It’s good to sit with a problem and try to think your way out of it. You learn a lot about yourself, and a problem, by following this method.

However, there comes a point of diminishing returns, where each additional unit of time invested has a lower and lower probability of bringing us to an answer.

At this point, it’s better to ask for help.

In my teams at work, we ask each person to set a threshold in their own mind for how long they’re willing to stay stuck on a problem.

Asking for help after 15 minutes sub-optimal. But so is asking for help after being stuck for 3 days.

We don’t set a hard ‘stuck threshold’, but we generally encourage people to try solve a problem for themselves, and if it isn’t resolved in a reasonable period of time—within which they’re tried and exhausted most reasonable options—then it’s time to seek the help and advice of someone else.

Set a stuck threshold, and don’t be afraid to pull someone else in for help.

I loved this video by Peter Whidden. It’s a primer on Reinforcement Learning—a branch of Machine Learning—but explained in a simple way using the classic Pokemon Red Gameboy game.

Reinforcement learning uses an AI-driven system to “learn through trial and error using feedback from its actions”.

In this video, Whidden has the AI play 20,000 games with over 5-years of simulated game time. The feedback that Whidden provides the AI is a reward for certain actions in the game, such as exploration, catching a Pokemon, or defeating an opponent. Because the inputs are simple: Up, Down, Left, Right, Button-A, and Button-B, the AI learns over time how to adjust it’s inputs to get to the reward faster.

What makes this video so excellent is Whidden’s editing. He visualises the journey of the 20,000 simulations in a way that makes it really easy to understand how reinforcement learning works.

Pokemon Red was my first ever video game, and it’s one that has a lot of personal nostalgia. It was awesome to see it used to explain reinforcement learning in a clear and simple way.

“Play long-term games with long-term people.”
—Naval Ravikant

This Naval quote has stayed with me since hearing it 4-5 years ago, and I believe it will stay with me for life.

Implicit in this quote is a belief in playing in spaces that aren’t zero sum games, in investing in things that compound, and working with people who have a strong moral compass and possess a bold vision for what the world should look like.

Spotify’s recommendation algorithm continues to get better.

I started listening to a song this morning from my teens, and the songs that followed were a perfectly picked mix of nostalgia.

Other other evening, I picked a playlist for a relaxing afternoon vibe, and it mixed some of my favourites with classics and new songs alike.

If machine learning and AI is built on data and distribution, the data is my listening history for the past 7 years, as well as the listening history from other users similar to me. The distribution is (almost) all the songs in the world.

Spotify has a powerful value proposition.

When a company’s website uses language like, “your results will be ready in a while,” without giving any signal on what “a while” means, or how long you might be waiting, the imprecision reveals the level of care put into the product.

It says, “we didn’t really think about how you, the user, would experience this, we just put down the first thing that came to mind and shipped it”.

It’s lazy, and it shows.

Note: this was a loading screen I came across whilst on the website of a European low-cost airline.

During the past month, I learnt how to solve a Rubik’s cube, and have reduced my time-to-solve from 5 minutes to a little above 3 minutes.

This is slow compared to the pros, who can regularly solve a cube in under 5 seconds.

I’ve merely learnt the basic 7 algorithms that, if implemented in that exact order, will get you to a solved cube every time. And I’ve only practised getting faster at these 7 steps. However, this approach is slow and inefficient.

For those willing to go deep down Rubik’s cube rabbit hole, there are a dizzying array of algorithms to learn. For example, just to solve the last layer of the cube there are 57 algorithms to “orient” and an additional 21 to “permute”—with orient meaning the last layer has the right colours, and permute meaning all cubes in the last layer are in their correct position. Committing these algorithms to brain and muscle memory can help you shave minutes off your solve time.

I don’t have time, nor the interest, in learning all the Rubik’s cube algorithms, but they are a great insight into what’s possible. When I look at an unsolved cube, there is the potential for it to be solved much faster if I’m willing to put in the work.

I’ll take that energy and apply it to an algorithm elsewhere.

The owe our best books a re-reading. They deserve to be studied, pored over, dissected, debated, and digested.

As we move slowly and thoughtfully through the text, and revisit the same ideas year after year, their lessons unfold into new shapes and colours.

This is how we should treat our best books.

I have an odd interest in organisational communication. How companies do it, what tools they use, and how their choices help or hinder them.

My interest stems from disappoint in the current set of tools.

Email is the default (and is here to stay for inter-company communication), but it’s woefully inadequate for intra-company use cases. Slack and Teams have helped chat-based communication grow, but this medium is inefficient, scattered, and non-persistent.

Over time, I’ve built a personal thesis that the ideal form of organisational communication is (1) asynchronous, (2) open, (3) searchable, and (4) persistent.

(1) Asynchronous communication means we don’t require immediate responses. Chat—which is the most common form of synchronous communication—is the biggest culprit here. The immediacy of chat gives people the illusion that they’re productive. It also creates anxiety to respond quickly to look good in front of others in the thread. Asynchronous communication on the other hand gives people the time and space to respond thoughtfully. I’d much rather receive an answer a few hours later that’s well considered than get a half-baked answer now.

(2) Open communication means information is accessible to all. Email hides things between people, and often gets lost between who is Cc’d and Bcc’d. For small organisations, I think everything except HR should be accessible to all employees.

(3) Searchable content only works when a company is doing (2) well (i.e., when a company is open with its information). Again, email and chat fall short here. They’re closed ecosystems of communication that lose their value as people quit and move jobs. A searchable content ecosystem on the other hand allows employees to find and see the history of company’s decisions and mistakes.

(4) Persistence is the preservation of documents, decisions, and information. The argument should be familiar by now: email and chat are impermanent and gated to a few individuals. Just like stories and lessons are passed down in families from generation to generation, we need to do the same in our organisations.

There is software that tries to do all of the above: Workplace by Meta creates a historical hive mind, but many companies feel it looks and works too similarly to the Facebook consumer app. Slack and Teams have implemented good search functionality, and the use of channels provides some persistence, but in my experience information still gets lost. There are also smaller products like P2 from Automattic, which use posts and pages to create internal blogs centred around common projects and themes. I like the concept behind P2, but it still feels a bit janky.

I don’t have answers. Only observations. But I do think organisational communication can still be much better.

The sound check allows the performer to set up their gear, make sure everything works, and to check the sound balance and mix is good.

The team can prepare and tweak and test endlessly, but once hundreds or thousands of human bodies cram into the venue, it changes how the sound travels, absorbs, and reverberates around the space.

No matter the amount of preparation, show time changes everything.

I enjoyed this NY Times article on making time for oddball rituals.

Unlike traditions—which are customs or beliefs handed down from one generation to the next—rituals are “acts regularly repeated in a specific manner”.

Maru and I have a handshake that we use to congratulate each other when the other has done something well.

My family had a code word when I was a kid to let the others know when they were tired and wanted to leave a gathering or event.

Every kids sports team back home in New Zealand brings a container of cut up oranges for the team to eat at the half-time break.

There is evidence that rituals can help reduce performance anxiety and help us feel more in control”. And although we shouldn’t force them, being open to adopting and creating them can inject fun and familiarity into our day-to-day.

We’re an Argentine / Kiwi household. Argentina has one of the best football teams in the world, and New Zealand one of the best rugby teams.

Tonight our countries played each other in the Rugby World Cup semi final. Maru and I had a fun week bantering about which team was likely to win. Despite the joking, we already had a “Football / Rugby agreement” in place.

We each support our respective countries in both sports, but the moment one of our teams gets knocked out in a competition, we become a 100% fan of the other country’s team.

I loved cheering for Argentina in last year’s scintillating Football World Cup Final. And after the Kiwi win this evening, Maru and I will be cheering for the boys in black in next week’s Rugby final!

They say you’re a true Londoner when it’s raining and you don’t change your plans.

Despite the rain, I’ve pressed through with plans multiple times this week, and returned home either damp or wet.

After 7 years in this city, I think that either makes me a true Londoner, or just someone who chronically forgets their umbrella.