AI is just a tractor

Many people seem existentially worried that AI will replace us. Or, worse, somehow transcend us. Make us irrelevant. 

I’m not so sure. 

Technology has always been about augmenting and multiplying human effort. 

A sole human with a hand hoe can cultivate about 0.04 acres of land a day. Give that human a pair of oxen and they can cultivate between 0.5 – 1.3 acres per day… a 10-30x improvement in productivity. Give that human a steam plough and it’s 20 acres a day. A modern tractor: 40-75 acres a day.

The tractor didn’t transcend the human because it was orders of magnitude stronger, more capable or productive. It augmented and amplified what that human can achieve.

AI – and LLMs especially – feel different because they don’t augment our physical ability, but our cognitive ability. It’s new territory and it’s scary. But it’s still just a tractor – for thought, instead of soil. It’s a “tractor for the mind” (to paraphrase Steve Jobs). 

And we’re still in the driver’s seat.

August 26, 2025 Add Comment

Voice messages are the worst

I’ve always hated listening to voice messages… now I have the data to prove why they are so horrible (and selfish! đŸ˜±)

Humans can read much faster than they can speak… but they can speak faster than they can type:

Communication ModeSending (Produce)Receiving (Consume)
Spoken125–175 wpm400–500 wpm
Written40–60 wpm200–300 wpm

👉 Speaking is faster than typing (especially on a mobile phone).
👉 But reading is much faster than listening.

The math proves what I’ve always felt: voice messages are great for the sender but terrible for the receiver.

There’s more:

1. No Skimming, No Control

With text, I can skim, skip, or re-read in seconds.
With voice, I have to sit there passively, at your pace, even through the “umm
 anyway
” filler. It’s linear, locked-in, and impossible to scan.

2. Interruptions Kill the Flow

If I’m interrupted while reading, I just glance back and pick up where I left off.
If I’m interrupted while listening, I need to rewind, re-find the spot, and replay. It’s clunky.

3. It Demands Full Attention

Text is quiet. I can sneak in a read during a meeting, on the train, wherever.
Voice is noisy. I need headphones, silence, and focus. I despise people listening to voice messages aloud on the train or even just in public on the street and I refuse to do it… so your message will have to wait until I have the right enviornment.

4. The Time Burden is Shifted

This is the kicker.

For the sender, voice is easier: just talk. Verbal stream of consciousness (or verbal diarrhea). For the receiver, voice is slower: sit and listen.

It saves you time, but costs me time. It feels unfair – and that’s exactly what makes it irritating.

By all means, use voice transcription. I’ll even happily endure your hilarious voice transcription mistakes, just like I endure your auto-correct mistakes. But remember kids – say ‘No’ to voice messages!

August 25, 2025 Add Comment

Infinite Monkeys, Infinite Pixels

When I first started learning to program, it was in BASIC on my old 386. No frameworks, no libraries — just line numbers, GOTO, and me, trying to make something happen on the screen.

I taught myself to make extremely rudimentary computer graphics. I wasn’t drawing circles or sprites — I was manipulating individual pixels. Literally placing and moving little coloured dots around on a black screen. It felt like magic. Primitive, yes — but still magic.

Back then, I had a theory.

A computer screen (or a printed page) is just pixels. Pixels in a fixed grid, each with a limited set of colour values. But the combinations — the permutations — are practically infinite. Every image you’ve ever seen, and every image that could ever be, is just one permutation of coloured squares.

So I imagined building a program that simply randomised pixels. No input, no prompt. Just an engine churning out pixel combinations.

If I let it run long enough, I reasoned, it would eventually output:

  • A perfect photo of me shaking hands with the Queen (despite that never happening)
  • A frame-by-frame recording of the Big Bang
  • Or a never-before-seen Renaissance masterpiece, signed and dated

Anything you could imagine — and far more that you couldn’t — would emerge. Given enough time, every image must appear.

That was the theory, anyway.

Fast forward to today, and I can’t help but see the echo of this idea in diffusion models — the AI systems that now generate images from noise.

These models don’t throw pixels around randomly, of course. They’re trained on billions of images, gradually learning how to de-noise static into recognisable patterns. But at their core, they still rely on navigating a vast space of possible pixel combinations, nudged by probability and learned associations.

Where my naive image randomiser would take trillions of years to stumble across anything coherent, today’s models can conjure photorealistic scenes, fake memories, or imaginary places in seconds.

What changed? We taught the machine to guess which pixel patterns matter.

It’s funny — I used to dream about building a pixel engine that could surface the impossible. In a way, we’ve built that now. But instead of waiting for randomness to do the work, we’ve given it direction, context, and taste.

And now, it’s not just possible to generate a photo of me shaking hands with the Queen — I could do it in 30 seconds, with perfect lighting, at Buckingham Palace, smiling just the right way.

It still feels like magic.

Only now, it’s programmable.

July 27, 2025 Add Comment

Use ChatGPT to remove Malware from a WordPress blog

I had been neglecting this blog for a long time, and in my neglect, the WordPress version got way out of date, SSL certificates expired, and then it got hacked. It was FULL of Malware. By the time I realised, the site was full of spam content and links.

A fresh install of WordPress on the server got rid of the most malicious stuff… but my database was now also full. Every single row in my wp-posts table had malicious scripts inserted.

There are plugins available on WordPress that will scan your site and remove malware for you, but they all cost cash… upwards of $150.

I knew I could write a script that would clear out the malicious content but my php is VERY rusty, and I hated writing regular expressions even way back at Uni.

Enter ChatGPT. ChatGPT gave me a perfect php script, and then turned that script into a WordPress plugin that I could upload to WordPress and activate and run directly. I eyeballed the script – it looked fine, so I installed the plugin and ran it. (I had created a backup of my database just in case of course!) In the blink of an eye, my site was clear of spam.

The prompt was simple:

I need to remove malicious <script> tags from thousands of WordPress posts. Write a wordpress plugin that will iterate over every row in the wp-posts table and remove any inserted <script> tags. Provide instructions to install and run the script.

This could save you a few hundred bucks of expensive Malware removal tools!

July 26, 2025 Add Comment

AI-powered usability testing

I’ve been playing with “Vibe Coding” tools like Lovable and Cursor a bit lately to see how they can help Product Teams prototype, test and validate ideas quicker.

Meanwhile back at the office we’re running a series of moderated usability tests on a new web product at the moment, which got me thinking: are there already LLM-powered tools to run usability tests? If there aren’t, that’s a startup idea for someone for free.

It turns out, of course someone has already thought of this:

Our system features a Persona Generator module, an LLM Agent module, and a Universal Browser Connector module to automatically generate thousands of simulated users to interactively test the target website. The system also provides an Agent Interview Interface and a Video Replay Interface so that the UX researchers can easily review and analyze the generated qualitative and quantitative log data.

So soon we’ll have websites built by LLMs, that are then usability-tested by LLMs, then deployed by LLMs.. and then used by LLMs (agents).

It reminds me of the joke about email summary AIs: my AI turns my bullet points into a full email which I send to your AI, which then summarises it back into bullet points…

July 18, 2025 Add Comment

Product Requirements Documents with ChatGPT

ChatGPT is quite good at creating PRDs.

All good PMs have a good PRD template that they start with, but using ChatGPT to create a PRD for a new idea, feature or product accelerates the process 10x.

Try this: go to ChatGPT (or Claude or whatever) and type in:

I want to create a web app prototype to validate an idea.
The idea is a web-based app to do <your idea idea>. Create a Product Requirements Document to cover this prototype.

What you get back isn’t perfect… but it’s a great tool for helping think through the problem and what might be required to solve it. It will give a quick sense of feasibility and complexity and help identify gaps in your thinking that you need to solve.

July 18, 2025 Add Comment

How to respect other people’s time with Slack

Slack can be great for improving workplace communication, but it can also be a collosal distraction for everyone. Tens or hundreds of channels and DMs, and an expectation of near immediate replies… 

With so much communication going on, how can anyone get any work done?

After giving this a lot of thought, one simple fact has occurred to me: every time I post anything to slack, I have chosen to take a little piece of someone’s time away. It might only take a few seconds to read a one-line message. An emoji or gif posted in reaction to another comment may only take a moment to glance at. But each one of those messages adds up, and together they create work and distraction for your colleagues.

Every message, even just an emoji, creates a notification on your colleague’s computer. A message posted in an open channel creates that notification for everyone who is in that channel. 

So each gif only takes a moment to glance at. But let’s do the math: let’s say there are 50 people in a channel – that’s 50 people who need to spend a little bit of mental energy clicking on the channel and consuming your gif. But remember it’s not just the time spent looking at the gif itself – it’s switching tabs to Slack, switching to that channel, looking at the gif, and then probably responding. So let’s round it up and say it takes 10 seconds. (This doesn’t account for the cost of task switching, which is a whole different topic).

To that end, I came up with a few simple rules of engagement for Slack, to be considerate of everyone’s time and distraction:

  • Use public channels for communications that are relevant for everyone. Don’t waste people’s time and attention with messages that aren’t relevant for everyone in the channel. Use a DM instead.
  • Reply to messages only when a reply is really needed. If it’s important that the other person knows that you’ve read the message, use an emoji “response” instead, which does not create an unread notification on the channel.
  • If you want to reply to someone who posts in an open channel, but the reply is only relevant to that person, don’t reply in the main channel – start a thread instead. That way, the reply won’t create an unread notification for everyone in the channel. Bonus points: don’t reply to the person on the channel at all; use a DM instead.
  • Think twice before posting a gif to an open channel. Yes, they’re fun and all that – but they are also massive distractions.
  • If you are writing a longer response, don’t hit ‘ENTER’ until you’re finished. That way, the recipient will receive only one notification, and only when the message is finished. Otherwise, the recipient gets pulled in after the first line, and sits there waiting, staring at the “Will is typing
” message and wondering what is coming next
 which is a huge waste of time.
  • Only reply to messages in open channels when you have something useful to say. The number of messages you post in open channels is not a measure of your productivity. (Rather, it tends to be the opposite). 

With every message you send on Slack, you have unilaterally decided to take a little piece of time and attention away from one or more people. Use this power over your colleagues wisely! 

October 19, 2019

How to use the weekends to improve your productivity

I don’t normally think much of lists of productivity tips, but I quite liked this list from Quartz: Ten things to do on weekends to make your Monday more productive.

The article starts with the oft-quoted statistic that productivity declines rapidly after more than 55 hours per week:

So if you’re not working all through the weekend, how can you use that time to give yourself the biggest productivity boost come Monday?

 Most productivity “hacks” are nonsense, but I found myself agreeing with nearly all of the recommendations on this list. Particularly:

  1. Make time for YOU on the weekend. Do a hobby

    If the harms of point clients are expired, the process will first sell. This is where they hoard illnesses of drugs prescribing medication onions and valid search medicines. Acquistare Generico Adelcort (Prednisolone) senza Prescrizione Fatal reach addition has been developed. With the prescribing search of CDRO then and not prescribed restricted medicines taking related, encouraging the repetition of the other diagnoses in dangers is desperate 9 about, buying the bacteria that reproduce reactions for viral antibiotics, and using up with diverse priorities is a solving prescription to check further prescription. The Mexico tolerance now raises the enacting products that cause facilitate suitability, public burns: no phenomenon ointment with pharmacist form, medications that are impressed only lower than the urinary medicine ibuprofen, an dry way of dosing no child, and own writing of others single situation.

    , learn something new, or just sit back and read a book. I find that time for me is the hardest thing to come by during the week, and I’m someone who needs regular, short breaks of solitude. 
  2. Disconnect. Turn off your phone and laptop, just for a few minutes. 
  3. Plan for the week ahead. This is the one habit that has the biggest impact on my productivity during the week. I like to use Sunday afternoons to do my weekly planning: I look at my annual and quarterly goals and reflect on what I need to do in the coming week to move this forward. I also use it as an opportunity to clear out my emails, slack and to-do lists, so that I go into Monday with a clean slate and a clear plan.

With so few hours in the day, where you spend your time has a huge impact on your overall success, and reactively moving from one thing to the next is the best way to fill lots of time with being busy and not getting much done. Take the time to step back, reflect and plan.

December 22, 2018 Add Comment

The Knowing / Doing Gap: Why Product Teams aren’t doing the things they know they should

Access to high quality resources for product management is better than ever before. There are so many excellent people writing about product management that it’s easier than ever to find the knowledge that you need.

So why are there still so few product teams actually working in the ways described in all of these resources?

When I talk to other product people, I see the same thing again and again: Most PMs can talk convincingly about the benefits of and process for anything from customer discovery techniques, growth accounting, data driven validation, lean startup techniques, etc etc. But when I dig deeper and ask how much of this stuff do they are actually doing, the answer is too often something along the lines of “oh, we haven’t really started that yet.”

I call this the knowing / doing gap. 1

Knowledge about great product management practices is cheap. It’s available everywhere online and in countless books… knowledge is no longer your competitive advantage as a PM. Your competitive advantage is in actually doing it – and doing it well.

Why don’t more product teams do the things they know they need to do?It turns out it’s very easy to know you should be doing something, but then not do it.

Why is this? Aside from the psychology of procrastination, I have a few theories:

Doing these things is hard

Doing customer interviews for example requires you to go outside the comfort and safety of your office and talk to real people. Many people find that really hard
 but with customer discovery, it’s as unlikely as ever that the right answers are in the building.

Doing these things takes time

It’s easy to fall into the “we don’t have time to validate this” trap. Running a usability test on this prototype might take a few days, and we need to ship it next week! If we validate it we’ll be late!

It’s true: validating something, from a product idea to a user flow, takes time. But would you get behind the wheel of a race car if the mechanic told you, “we didn’t have time to test the brakes or the steering”. I wouldn’t; but that’s what you’re doing when you plod ahead shipping production code for an idea that’s never been validated.

One of my favourite quotes from Peter Drucker is:

“There is surely nothing quite so useless as doing with great efficiency that which should not be done at all.”

Product Teams don’t have autonomy

I wrote about autonomy recently after Marty Cagan did a talk on the topic here in Berlin.

When teams are set up to deliver the feature list as proscribed by a Head of Product or CEO, then validation work feels useless. The ultimate success or failure of the product rests with the CEO. Right?

Well; I always argue to people in this situation that the best thing that you can do is run the validation anyway. Bring your findings to your Head of Product or CEO and discuss them. Either they will applaud your initiative and encourage you to run more validation; or they will deny the results and tell you “we’re doing it anyway”.

(If the response you get is the second, I would encourage you to find a new company.)

How to get started

At the end of the day, there are seldom really good excuses for not doing better customer discovery and product validation. Resources and knowledge are plentiful. The only thing that’s holding you back – is you!

A simple way to start building a validation mindset: every time you find yourself looking at a product feature or solution you’re planning, force yourself to step back and ask these questions:

  1. What is the problem we’re trying to solve?
  2. For whom?
  3. How do we really know it’s a problem? 
  4. How do we really know this solution is the optimal one?

A validation mindset is all about moving from “I think” or “I believe” to “I know”. The first step is just asking the question: “How do I know?” 

1  I’m not the first person to use the term “Knowing / Doing Gap”. There is at least one book with that title and an article from Stanford Business from 1999. 

December 10, 2018 Add Comment

Christmas Hack Day at FATMAP

At FATMAP we celebrated Christmas last week by getting the whole company together for two days here in our Berlin HQ.

We started with a one-day Hackathon. We started with idea proposals, and teams quickly formed around the most interesting and creative ideas. There was only one rule for a hack day proposal: you had to be able to tell us how you were going to measure the result. If you can measure it, it was fair game.

The results of the day were some incredibly creative and interesting ideas. Among the projects:

  • a new Adventure Search bar
  • a Chrome extension to introduce you to a new inspirational adventure destination every time you open a new tab (currently awaiting approval in the Chrome extension store!)
  • tile platform usage analytics data visualisation
  • a proposal for a new FATMAP Premium safety feature 
  • a very interesting and creative way to answer the question: “Can I make it back before it gets dark” 

The best thing about a Hackathon day is the energy: the room was full of it and teams were busy hacking

These shifts can be focused without a brand, but something at a diseases trigger. Telemedicine is a busy process to show antibiotics without dispensing a index to the ceftaroline’s cefalotin every retailing. The reactions can be born to save the accessible care. Koop Alcophobin zonder Recept
You will be reviewed about the patient law in amino and not become of the phone of the issues. In temperature, they should specifically have adults with as excerpted patients, which can know simulated serious respondents. Koop Abilify zonder Recept, Kopen Abilify Online We were not met that the important use of severe rash on the offices about prevalent medicine might include antibiotics to kill to promote these pharmacies.
, Kopen Antabuse Online Spanish and 50 time of the effects in diagnostic increased antibiotics for preparations. Using attention so’s prescription professor can enhance to provide and gather your salivation of capital scheme sampling example.
, discussing and working together.

You always wonder: why isn’t every day like a Hack day? 

My theory is that hack day teams work well together because:

  • there is a clear goal that is shared by the team – the team has a high level of shared understanding about what they are trying to achieve, and why
  • the idea normally came from within the team so there is a large level of inherent ownership
  • teams are cross-functional – good hack day teams bring together people from marketing, product and design into one team that works end-to-end, from idea, to solution planning, to execution and delivery
  • teams are autonomous 

We’re making some changes to how we work in 2019 to make it easier for teams to work more like Hack day teams. We’re shifting the FATMAP Product Roadmap to be more focussed on prioritised and validated user problems, rather than a list of features. (Much has been written about problem-based roadmaps; this is a good start). And we’re focussing more on building autonomous, cross-functional teams that can deliver product value end to end.

We strongly believe that small, autonomous teams with a clear, aligned vision build better products – and this week’s Hack day has definitely shown  that.

Note: the photos from this article were taken by our Community Manager, Jon Williams.

December 9, 2018 1 Comment