<p>In March 2023 I started hacking on an idea to use LLMs to generate interview plans for recruiters. That’s how it started.</p>
<p>Last night I shipped an automated AI interview product named Sprinterview.</p>
<p>This post will break down the steps taken to build Sprinterview, mistakes I made, and what’s next.</p>
<blockquote>
<p>Before we dive in, I want to say a big thank you to the friends and family who patiently gave their time and advice for this project: Adam Shi, Tiffany Little, Elliot O’Connor, Millian Gehrer, Zara Yaqoob, Francesco Agosti, Jack Bridger, Shums Kassam, and my Mom and Dad. You’re all awesome.</p>
</blockquote>
<h2>The problem</h2>
<p>Sourcing and screening is one of the hardest problems in recruitment. The reason it’s hard is because it’s fundamentally limited by human bandwidth. Even when a recruiter follows the most optimized process possible, they will eventually bump against the physical limit of time available in the day.</p>
<p>My last job was with a recruitment startup called Frontier, and we spent a lot of time understanding the raw principles of sourcing and screening. We learned from our customers that high volume screening is often broken into three parts:</p>
<ol>
<li>Attract candidates.</li>
<li>Shortlist candidates.</li>
<li>Screening interview.</li>
</ol>
<p>Each step involves a different process, which are usually:</p>
<ol>
<li><strong>Attract candidates</strong> - advertise the job where candidates will find it.</li>
<li><strong>Shortlist candidates</strong> - convince candidates to fill out a job application form, capture data in an ATS, review applications, and create a shortlist.</li>
<li><strong>Screening interview</strong> - follow up with shortlisted candidates and arrange a screening call.</li>
</ol>
<p>As you can tell, the screening process demands a lot of human input at every stage. Technology doesn’t seem to lower the time cost for recruiters. It just moves it around. The cost still nets out higher for the recruiter.</p>
<p>Recruiters still have to stay on top of inbound applications, pattern match to find good candidates, and then screen them to validate. All of this takes time and effort that scales up with volume.</p>
<p>The secondary cost to the recruiter is candidate drop off. Candidate drop off is a function of time and steps. It happens when candidates are asked to wait for an unknown amount of time and complete an unknown number of steps before they get any signal themselves.</p>
<p>With sourcing and screening being so costly to candidates and recruiters alike, it’s clear there’s ample room to improve the hiring process for both sides. Not only that, but doing so could be incredibly valuable to a business.</p>
<h2>The prototype</h2>
<p>It was sometime in mid March 2023 at 1am. I was reflecting on an area of recruitment that I’ve always found tedious: pre-interview candidate research.</p>
<p>With LLMs in the back of my mind, it occurred to me that candidate research can be reduced to a pattern matching exercise assuming the information is available. I needed to know if this was true so I spent the next day validating the hypothesis.</p>
<h3>ChatGPT to the rescue</h3>
<p>To validate “can an LLM do X” is shockingly trivial with ChatGPT.</p>
<p>I was able to validate my assumption that an LLM can do the heavy lifting of interview preparation by designing a prompt that contained some basic instructions, and pasting different combinations of job description and candidate résumés.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/chatgpt-prototype.png" alt="ChatGPT prototype"></p>
<p>With some tuning, I was able to get ChatGPT to notice gaps and correlations between the résumé and job requirements and then design questions around them.</p>
<p>But can it be automated?</p>
<h3>Automated interview prep</h3>
<p>To create an automated interview preparation application I spent an evening condensing my ChatGPT experiment into a command line tool.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/cli-prototype.png" alt="Command Line Prototype"></p>
<p>Using the GPT-3.5 API, the CLI tool I built took a job description, LinkedIn profile, and interview time limit as inputs. It then returned a JSON formatted set of questions for an interviewer to ask when screening the provided candidate.</p>
<p>I demoed the CLI tool to some friends and asked for real opinions. These were the highlights:</p>
<ul>
<li>It’s cool but how do I know I can trust it?</li>
<li>I’d use it for high volume.</li>
<li>If I could use it from LinkedIn I would pay for it.</li>
<li>I definitely wouldn’t use it for high skilled roles.</li>
<li>Isn’t it unfair to ask different questions to different candidates?</li>
</ul>
<p>The feedback was exactly what I needed. Although the CLI solution was a shitty first draft, people had energy for the value proposition.</p>
<p>At this point it was clear there is value in exploring the combination of LLMs and interview screening. Could this be it? I don’t know, but let’s build an MVP to find out.</p>
<h2>MVP 1.0 - Interview preparation</h2>
<p>In the weeks following the CLI prototype, I designed and built a NextJS app as a simple MVP interview preparation application.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/app_setup.png" alt="Onboarding screen for the first MVP"></p>
<p>The core LLM piece of the app was easy to port over from the CLI tool I made the week before. But I was unprepared for the difficult task of scraping content from LinkedIn.</p>
<p>Scraping slowed the app down significantly because it meant running a headless browser to fully automate the process, which I saw as important. So I continued to hack together a secondary background service that could do all the heavy lifts that were scraping and generating content.</p>
<p>I was able to solve the LinkedIn scraping problem by storing my authentication cookies securely in a database and injecting them into a headless browser session to do the scraping. Then I used GPT to clean up the text, and finally run the preparation operation.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/app_questions.png" alt="Final questions for the first MVP"></p>
<p>Automation took me longer than planned because I was working against anti-scraping measures. But when it worked it was magical! I saw this could have saved me a lot of work in the past. What more could I ask for?</p>
<p>I booked more customer development calls with friends and founders in the recruitment space and demoed the MVP to them. Here’s the highlights from those conversations:</p>
<ul>
<li>Might be useful for people new to recruitment, but not the pros.</li>
<li>I’d like it to ask the questions too.</li>
<li>What we really need is a deterministic way to know what good looks like.</li>
<li>Can it do followup questions in real time?</li>
<li>I’d use this for customer development instead of recruitment.</li>
</ul>
<p>With this round of demos the excitement was higher. People had more ideas and were translating what they saw to their own problems. I saw this as signal. But was this MVP really solving a problem or just playing close to the solution?</p>
<p>As I went over feedback and explored ideas I arrived back at an idea I’d been trying to avoid: fully autonomous AI interviews. Why was I avoiding AI interviews? I didn’t see the value; I saw it as a gimmick (which might still be accurate), and didn’t think there was a problem AI interviews would meaningfully solve.</p>
<p>What finally drove me into the warm embrace of automated interviews was the thought that this would be really cool if it was done well, and it felt naughty to automate a recruiter. So I went to bed excited for what was next.</p>
<h2>MVP 2.0 – Autonomous AI interviews</h2>
<p>April 21st 2023, sometime in the morning. I spun up a NextJS app with LangChain and TailwindCSS. This was going to be a one day build so I turned on GitHub CoPilot and made a promise to myself that I wouldn’t get distracted.</p>
<p>By 4pm I was sitting in a chatroom with “AI Interviewer” answering questions about my experience with Software Development. When I gave vague or bad answers the interviewer just moved on. I adjusted the prompt and then “AI Interviewer” asked me a followup question. There it was, that magical feeling again.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/ai_interviewer.png" alt="AI Interviewer"></p>
<p>By the time evening rolled around I had designed a data model and basic process to add new interviews for “AI Interviewer” to conduct.</p>
<p>I got my girlfriend to sit for the first full interview at 11pm. The UI was terrible, “AI Interviewer” wasn’t always professional, but it worked! That was enough for one day.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/tiffany.png" alt="Tiffany demoing AI Interviewer"></p>
<p>The following days I demoed the chat with “AI Interviewer” to a number of different people. Energy was even higher, and I could see there was light at the end of the tunnel. But there was just one problem…</p>
<h3>Text interviews, really?</h3>
<p>Text sucks. In fact, the reason recruiters need to screen candidates after they’ve already applied for a job is because text is low signal.</p>
<p>What if “AI Interviewer” could understand what a candidate says and respond in real time? What if the candidate’s answers could be analyzed, the audio saved, and a similar outcome as a screening call could be reached in less time and with fewer steps?</p>
<p>Imagine if candidates could apply for jobs by doing a 5 minute AI screening call with “AI Interviewer” while doing housework instead of the alternative of filling out application forms and praying. What would this mean for candidates and recruiters?</p>
<h3>Speak to answer</h3>
<p>The interview application was reworked once again to achieve the spoken screening call process by paying even more money to OpenAI via the Whisper API. I built a new service named “Fluent” that transcribes messages and keeps the audio ready for automated screening.</p>
<p>For the app itself I built a mobile-first experience that gives candidates the freedom to interview on a portable device at their own pace. All a candidate has to do to chat with the “AI Interviewer” is hold down record and then hit send.</p>
<p>As for “AI Interviewer”, let’s change that too.</p>
<h3>Hello Amy.</h3>
<p>Amy (as in A(m)i) is “AI Interviewer” V2. She operates under a set of laws and principles that make her more competent, professional, and compassionate than “AI Interviewer”. She has more information at her disposal, and she has a generated face which I created with Midjourney.</p>
<p><img src="https://dl.nosaj.io/posts/ai-interviews/amy_interview_2.png" alt="Amy interview live example"></p>
<p>Amy is the real value of Sprinterview. She is the face of the product. Eventually, Amy could help employers figure out what a good candidate looks like, and even help them decide which candidates are the best post-interview.</p>
<h2>What now?</h2>
<p>If you’re interested in testing Sprinterview for yourself just email me on <a href="mailto:jason@sprinterview.ai">jason@sprinterview.ai</a> for a personal access link.</p>
<p>I’m excited to show this thing to more people and focus on product market fit. With the product I’m exploring how the analysis of interviews will work so that it saves recruiters time and gives candidates the tools to put their best selves forward.</p>
<hr>
<p>I’d love to chat to anyone interested in using Sprinterview or helping out, so if you’re interested, don’t hesitate to reach out via <a href="mailto:hi@nosaj.io">email</a> or <a href="https://twitter.com/nosajio">Twitter</a>.</p>
<p>Thank you for reading.</p>


In March 2023 I started hacking on an idea to use LLMs to generate interview plans for recruiters. That's how it started.

Last night I shipped an automated AI interview product named Sprinterview.

This post will break down the steps taken to build Sprinterview, mistakes I made, and what's next.

> Before we dive in, I want to say a big thank you to the friends and family who patiently gave their time and advice for this project: Adam Shi, Tiffany Little, Elliot O'Connor, Millian Gehrer, Zara Yaqoob, Francesco Agosti, Jack Bridger, Shums Kassam, and my Mom and Dad. You're all awesome.

## The problem
Sourcing and screening is one of the hardest problems in recruitment. The reason it's hard is because it's fundamentally limited by human bandwidth. Even when a recruiter follows the most optimized process possible, they will eventually bump against the physical limit of time available in the day.

My last job was with a recruitment startup called Frontier, and we spent a lot of time understanding the raw principles of sourcing and screening. We learned from our customers that high volume screening is often broken into three parts:
1. Attract candidates.
2. Shortlist candidates.
3. Screening interview.

Each step involves a different process, which are usually:
1. **Attract candidates** - advertise the job where candidates will find it.
2. **Shortlist candidates** - convince candidates to fill out a job application form, capture data in an ATS, review applications, and create a shortlist.
3. **Screening interview** - follow up with shortlisted candidates and arrange a screening call.

As you can tell, the screening process demands a lot of human input at every stage. Technology doesn't seem to lower the time cost for recruiters. It just moves it around. The cost still nets out higher for the recruiter. 

Recruiters still have to stay on top of inbound applications, pattern match to find good candidates, and then screen them to validate. All of this takes time and effort that scales up with volume.

The secondary cost to the recruiter is candidate drop off. Candidate drop off is a function of time and steps. It happens when candidates are asked to wait for an unknown amount of time and complete an unknown number of steps before they get any signal themselves.

With sourcing and screening being so costly to candidates and recruiters alike, it's clear there's ample room to improve the hiring process for both sides. Not only that, but doing so could be incredibly valuable to a business.

## The prototype
It was sometime in mid March 2023 at 1am. I was reflecting on an area of recruitment that I've always found tedious: pre-interview candidate research.

With LLMs in the back of my mind, it occurred to me that candidate research can be reduced to a pattern matching exercise assuming the information is available. I needed to know if this was true so I spent the next day validating the hypothesis.

### ChatGPT to the rescue
To validate "can an LLM do X" is shockingly trivial with ChatGPT.

I was able to validate my assumption that an LLM can do the heavy lifting of interview preparation by designing a prompt that contained some basic instructions, and pasting different combinations of job description and candidate résumés.

![ChatGPT prototype](https://dl.nosaj.io/posts/ai-interviews/chatgpt-prototype.png)

With some tuning, I was able to get ChatGPT to notice gaps and correlations between the résumé and job requirements and then design questions around them.

But can it be automated?

### Automated interview prep
To create an automated interview preparation application I spent an evening condensing my ChatGPT experiment into a command line tool. 

![Command Line Prototype](https://dl.nosaj.io/posts/ai-interviews/cli-prototype.png)

Using the GPT-3.5 API, the CLI tool I built took a job description, LinkedIn profile, and interview time limit as inputs. It then returned a JSON formatted set of questions for an interviewer to ask when screening the provided candidate.

I demoed the CLI tool to some friends and asked for real opinions. These were the highlights:
- It's cool but how do I know I can trust it?
- I'd use it for high volume.
- If I could use it from LinkedIn I would pay for it.
- I definitely wouldn't use it for high skilled roles.
- Isn't it unfair to ask different questions to different candidates?

The feedback was exactly what I needed. Although the CLI solution was a shitty first draft, people had energy for the value proposition.

At this point it was clear there is value in exploring the combination of LLMs and interview screening. Could this be it? I don't know, but let's build an MVP to find out.

## MVP 1.0 - Interview preparation
In the weeks following the CLI prototype, I designed and built a NextJS app as a simple MVP interview preparation application.

![Onboarding screen for the first MVP](https://dl.nosaj.io/posts/ai-interviews/app_setup.png)

The core LLM piece of the app was easy to port over from the CLI tool I made the week before. But I was unprepared for the difficult task of scraping content from LinkedIn. 

Scraping slowed the app down significantly because it meant running a headless browser to fully automate the process, which I saw as important. So I continued to hack together a secondary background service that could do all the heavy lifts that were scraping and generating content.

I was able to solve the LinkedIn scraping problem by storing my authentication cookies securely in a database and injecting them into a headless browser session to do the scraping. Then I used GPT to clean up the text, and finally run the preparation operation.

![Final questions for the first MVP](https://dl.nosaj.io/posts/ai-interviews/app_questions.png)

Automation took me longer than planned because I was working against anti-scraping measures. But when it worked it was magical! I saw this could have saved me a lot of work in the past. What more could I ask for?

I booked more customer development calls with friends and founders in the recruitment space and demoed the MVP to them. Here's the highlights from those conversations:

- Might be useful for people new to recruitment, but not the pros.
- I'd like it to ask the questions too.
- What we really need is a deterministic way to know what good looks like. 
- Can it do followup questions in real time?
- I'd use this for customer development instead of recruitment.

With this round of demos the excitement was higher. People had more ideas and were translating what they saw to their own problems. I saw this as signal. But was this MVP really solving a problem or just playing close to the solution?

As I went over feedback and explored ideas I arrived back at an idea I'd been trying to avoid: fully autonomous AI interviews. Why was I avoiding AI interviews? I didn't see the value; I saw it as a gimmick (which might still be accurate), and didn't think there was a problem AI interviews would meaningfully solve.

What finally drove me into the warm embrace of automated interviews was the thought that this would be really cool if it was done well, and it felt naughty to automate a recruiter. So I went to bed excited for what was next.

## MVP 2.0 – Autonomous AI interviews
April 21st 2023, sometime in the morning. I spun up a NextJS app with LangChain and TailwindCSS. This was going to be a one day build so I turned on GitHub CoPilot and made a promise to myself that I wouldn't get distracted.

By 4pm I was sitting in a chatroom with "AI Interviewer" answering questions about my experience with Software Development. When I gave vague or bad answers the interviewer just moved on. I adjusted the prompt and then "AI Interviewer" asked me a followup question. There it was, that magical feeling again.

![AI Interviewer](https://dl.nosaj.io/posts/ai-interviews/ai_interviewer.png)

By the time evening rolled around I had designed a data model and basic process to add new interviews for "AI Interviewer" to conduct. 

I got my girlfriend to sit for the first full interview at 11pm. The UI was terrible, "AI Interviewer" wasn't always professional, but it worked! That was enough for one day.

![Tiffany demoing AI Interviewer](https://dl.nosaj.io/posts/ai-interviews/tiffany.png)

The following days I demoed the chat with "AI Interviewer" to a number of different people. Energy was even higher, and I could see there was light at the end of the tunnel. But there was just one problem...

### Text interviews, really?
Text sucks. In fact, the reason recruiters need to screen candidates after they've already applied for a job is because text is low signal.

What if "AI Interviewer" could understand what a candidate says and respond in real time? What if the candidate's answers could be analyzed, the audio saved, and a similar outcome as a screening call could be reached in less time and with fewer steps?

Imagine if candidates could apply for jobs by doing a 5 minute AI screening call with "AI Interviewer" while doing housework instead of the alternative of filling out application forms and praying. What would this mean for candidates and recruiters?

### Speak to answer
The interview application was reworked once again to achieve the spoken screening call process by paying even more money to OpenAI via the Whisper API. I built a new service named "Fluent" that transcribes messages and keeps the audio ready for automated screening.

For the app itself I built a mobile-first experience that gives candidates the freedom to interview on a portable device at their own pace. All a candidate has to do to chat with the "AI Interviewer" is hold down record and then hit send.

As for "AI Interviewer", let's change that too.

### Hello Amy.
Amy (as in A(m)i) is "AI Interviewer" V2. She operates under a set of laws and principles that make her more competent, professional, and compassionate than "AI Interviewer". She has more information at her disposal, and she has a generated face which I created with Midjourney.

![Amy interview live example](https://dl.nosaj.io/posts/ai-interviews/amy_interview_2.png)

Amy is the real value of Sprinterview. She is the face of the product. Eventually, Amy could help employers figure out what a good candidate looks like, and even help them decide which candidates are the best post-interview.

## What now?
If you're interested in testing Sprinterview for yourself just email me on [jason@sprinterview.ai](mailto:jason@sprinterview.ai) for a personal access link.

I'm excited to show this thing to more people and focus on product market fit. With the product I'm exploring how the analysis of interviews will work so that it saves recruiters time and gives candidates the tools to put their best selves forward.

---

I'd love to chat to anyone interested in using Sprinterview or helping out, so if you're interested, don't hesitate to reach out via [email](mailto:hi@nosaj.io) or [Twitter](https://twitter.com/nosajio).

Thank you for reading. 

Can AI Interviews solve one of the most costly problems in recruitment?

AI Interviews as a Service

<p>I love to think about the potential of generative AI. But the realist in me can’t stop thinking about the brave new world we might be about to enter into. Specifically, the implications of generative AI tools like GPT-3/4 and DALL-E on human mastery.</p>
<p>Generative AI can already take a prompt similar to a search engine query and return content that rivals experts in many different fields. This alone completely subverts our current understanding of mastery.</p>
<p>Right now, instead of working for thousands of hours to master a new field, we can use a tool that works as a shortcut to the goal. Skills that were once valuable – such as acedemic pursuits, writing, teaching, and even law – are easily simulated with software.</p>
<p>Why spend years honing a craft or unique style when software like DALL-E, Midjourney, and others can be prompted to create any style of art, music, or text?</p>
<p>In a world where mastery is cheap, people will be disincentivised to practice and acquire skills. Future generations might be less motivated to understand the underlying principles of their field. Instead they will rely on software as the shortest path to creation.</p>
<p>And if mastery is no longer worthwhile, where will the visionaries come from to move a field forward? Innovation requires a deep intuitive understanding of a domain, and if humans aren’t motivated to acquire this understanding, progress will stall.</p>
<p>Furthermore, in a world without progress, where will the new materials used to train the next generation of AI systems come from? It seems that the answer might be more AI-generated content. Leading to a feedback loop of AI training AI with no extrinsic evolution.</p>
<p>In closing, let’s ask GPT-4 for its opinion on the problems raised in this article:</p>
<blockquote>
<p>While the rapid advancement of generative AI tools like GPT-3/4 and DALL-E raises valid concerns, it also presents an opportunity for humanity to redefine mastery and creativity. By embracing these tools, we can enhance our problem-solving abilities, foster collaboration between humans and AI, and unlock new avenues of innovation. By leveraging AI as an extension of human ingenuity, we can combine our unique insights with AI-generated content, allowing us to tap into a broader range of ideas and experiences. Ultimately, this synergy could lead to a more diverse and enriched understanding of the world, empowering future generations to push the boundaries of human knowledge and drive progress forward in unprecedented ways.</p>
<p>– OpenAi GPT-4</p>
</blockquote>


I love to think about the potential of generative AI. But the realist in me can't stop thinking about the brave new world we might be about to enter into. Specifically, the implications of generative AI tools like GPT-3/4 and DALL-E on human mastery.

Generative AI can already take a prompt similar to a search engine query and return content that rivals experts in many different fields. This alone completely subverts our current understanding of mastery. 

Right now, instead of working for thousands of hours to master a new field, we can use a tool that works as a shortcut to the goal. Skills that were once valuable – such as acedemic pursuits, writing, teaching, and even law – are easily simulated with software.

Why spend years honing a craft or unique style when software like DALL-E, Midjourney, and others can be prompted to create any style of art, music, or text?

In a world where mastery is cheap, people will be disincentivised to practice and acquire skills. Future generations might be less motivated to understand the underlying principles of their field. Instead they will rely on software as the shortest path to creation.

And if mastery is no longer worthwhile, where will the visionaries come from to move a field forward? Innovation requires a deep intuitive understanding of a domain, and if humans aren't motivated to acquire this understanding, progress will stall.

Furthermore, in a world without progress, where will the new materials used to train the next generation of AI systems come from? It seems that the answer might be more AI-generated content. Leading to a feedback loop of AI training AI with no extrinsic evolution.

In closing, let's ask GPT-4 for its opinion on the problems raised in this article:

> While the rapid advancement of generative AI tools like GPT-3/4 and DALL-E raises valid concerns, it also presents an opportunity for humanity to redefine mastery and creativity. By embracing these tools, we can enhance our problem-solving abilities, foster collaboration between humans and AI, and unlock new avenues of innovation. By leveraging AI as an extension of human ingenuity, we can combine our unique insights with AI-generated content, allowing us to tap into a broader range of ideas and experiences. Ultimately, this synergy could lead to a more diverse and enriched understanding of the world, empowering future generations to push the boundaries of human knowledge and drive progress forward in unprecedented ways.
> 
> – OpenAi GPT-4


Future humans won't need mastery. They will rely on the masters who came before them

The End of Mastery

<p>The rise of social media was possibly the most exciting time in internet history.</p>
<p>It marked the moment when the internet <em>woke up</em> by enabling people to communicate at an unprecedented scale. It was the most disruptive moment to the social order since the invention of the printing press and the ensuing democratisation of information.</p>
<p>Startups that kicked off the social media revolution were aggressively lean and purpose driven. Think of Instagram’s boastful 270 million users with 13 employees before the <a href="https://www.dailymail.co.uk/news/article-2127343/Facebook-buys-Instagram-13-employees-share-100m-CEO-Kevin-Systrom-set-make-400m.html">Facebook acquisition</a>, and <a href="https://www.newyorker.com/magazine/2010/09/20/the-face-of-facebook">Facebook’s obsession with openness</a>, or Twitter’s goal of being the place where things happen.</p>
<p>Each social media innovator was staged to be a unique beacon of progress for the future of the internet. Things looked bright.</p>
<p>It’s now 2022 and lot has changed since the early days. Social media companies have undergone massive transformations, from human driven information centres, to growth machines.</p>
<p>By changing their purpose, platforms made decisions that divorced them from serving users. All of them traded trust for status, and they grew into the low trust Big Tech oligarchies they are today.</p>
<p>All except for Twitter.</p>
<p>Twitter is substantially different from the likes of Instagram and Facebook because it never fully divorced itself from its founding purpose. Twitter was, and still is, the place where things happen, and this purpose is worth defending.</p>
<p>When Twitter, Facebook, Instagram, and the rest of social media banned high profile accounts like President Trump, it caused an exodus of users from those platforms.</p>
<p>Users had to go somewhere. But parallel platforms didn’t appear to replace Facebook and Instagram. Instead there was <a href="https://parler.com/">Parler</a>, <a href="https://gettr.com/">Gettr</a>, and <a href="https://truthsocial.com/">Truth Social</a>. Networks that work like Twitter.</p>
<p>This signals that Twitter’s product solves a deep social problem that other products, like Meta’s portfolio, do not address.</p>
<p>Smart people have long <a href="https://techcrunch.com/2010/10/16/why-twitter-is-massively-undervalued-compared-to-facebook/">argued that Twitter was undervalued</a> because it harmonises with technology and human nature so perfectly. There’s truth in this; Twitter’s UX <em>feels</em> more like a core technology – a protocol like SMS or HTTP – and deeper than a product.</p>
<p><a href="https://twitter.com/jack">Jack</a> was smart to keep Twitter true to its roots. He never followed Zuckerberg by radically evolving the product from its original purpose to unlock more revenue. Dorsey knew that Twitter taps into something deeper in culture and humanity that simply shouldn’t be fucked with.</p>
<p>Under <a href="https://twitter.com/elonmusk">Elon’s</a> control, Twitter is experiencing more changes than it’s seen in a long long time. But Elon understands the function of Twitter better than someone like <a href="https://en.wikipedia.org/wiki/Parag_Agrawal">Parag</a>.</p>
<p>Twitter is a system, and like any system, the parts can change without meaningfully changing the system itself. If the function and purpose is intact, the system will live on.</p>
<p>Godspeed team Twitter and Elon 🚀</p>


The rise of social media was possibly the most exciting time in internet history.

It marked the moment when the internet _woke up_ by enabling people to communicate at an unprecedented scale. It was the most disruptive moment to the social order since the invention of the printing press and the ensuing democratisation of information.

Startups that kicked off the social media revolution were aggressively lean and purpose driven. Think of Instagram's boastful 270 million users with 13 employees before the [Facebook acquisition](https://www.dailymail.co.uk/news/article-2127343/Facebook-buys-Instagram-13-employees-share-100m-CEO-Kevin-Systrom-set-make-400m.html), and [Facebook's obsession with openness](https://www.newyorker.com/magazine/2010/09/20/the-face-of-facebook), or Twitter's goal of being the place where things happen.

Each social media innovator was staged to be a unique beacon of progress for the future of the internet. Things looked bright.

It’s now 2022 and lot has changed since the early days. Social media companies have undergone massive transformations, from human driven information centres, to growth machines.

By changing their purpose, platforms made decisions that divorced them from serving users. All of them traded trust for status, and they grew into the low trust Big Tech oligarchies they are today.

All except for Twitter.

Twitter is substantially different from the likes of Instagram and Facebook because it never fully divorced itself from its founding purpose. Twitter was, and still is, the place where things happen, and this purpose is worth defending.

When Twitter, Facebook, Instagram, and the rest of social media banned high profile accounts like President Trump, it caused an exodus of users from those platforms.

Users had to go somewhere. But parallel platforms didn’t appear to replace Facebook and Instagram. Instead there was [Parler](https://parler.com/), [Gettr](https://gettr.com/), and [Truth Social](https://truthsocial.com/). Networks that work like Twitter.

This signals that Twitter's product solves a deep social problem that other products, like Meta’s portfolio, do not address.

Smart people have long [argued that Twitter was undervalued](https://techcrunch.com/2010/10/16/why-twitter-is-massively-undervalued-compared-to-facebook/) because it harmonises with technology and human nature so perfectly. There’s truth in this; Twitter’s UX _feels_ more like a core technology – a protocol like SMS or HTTP – and deeper than a product.

[Jack](https://twitter.com/jack) was smart to keep Twitter true to its roots. He never followed Zuckerberg by radically evolving the product from its original purpose to unlock more revenue. Dorsey knew that Twitter taps into something deeper in culture and humanity that simply shouldn't be fucked with.

Under [Elon’s](https://twitter.com/elonmusk) control, Twitter is experiencing more changes than it’s seen in a long long time. But Elon understands the function of Twitter better than someone like [Parag](https://en.wikipedia.org/wiki/Parag_Agrawal).

Twitter is a system, and like any system, the parts can change without meaningfully changing the system itself. If the function and purpose is intact, the system will live on.

Godspeed team Twitter and Elon 🚀

What makes Twitter uniquely equipped to win.

Why Twitter Deserves to Live

<p>If you sail the seas of tech Twitter or LinkedIn you'll have likely noticed a trend lately: People can't seem to stop
  telling each other that consistency is the secret to whatever success is on the menu.</p>
<p>Despite this being an obvious idea, I still find myself feeling motivated by some of these posts. And it&#8217;s not
  news to me that doing something every day will result in progress. So what about it is motivating?</p>
<p>My best guess is that there's comfort in the simplicity of what consistency represents. It makes something
  potentially difficult and rife with unknowns into a prescriptive and manageable process. Just do the thing every day.
  Make the sacrifice and rewards will come.</p>
<p>Problem is, it doesn't matter how many times you read "consistency will get you everything", it won't make you more
  consistent. What makes you more consistent is two things: <strong>Conviction</strong> and <strong>discipline</strong>.
</p>
<h2>&#127775; Conviction</h2>
<p>Conviction is just a strong &#8211; bordering on delusional &#8211; belief that something is worth the sacrifices
  necessary to make it happen. A good example is building a company.</p>
<p>Over the last few months I've been exploring the idea of building a company myself. Actually exploring a lot of
  different ideas hoping to find one that I can't not pursue. </p>
<p>I've worked with early stage startups and founders for years to get product companies off the ground, so I thought I
  was prepared for all the difficulties of starting one. But I never knew what the period was like pre-conviction. It
  takes serious energy to find what matters.</p>
<p>Conviction is the source of energy that drives the engine of consistency. Without deep belief that something is worth
  spending time on, it will be an uphill battle to do it with any form of consistency.</p>
<p>If you want to be consistent with something, discover what gives you strong conviction about it.</p>
<h2>&#128293; Discipline</h2>
<p>Discipline is the brute brother of conviction. While conviction wants to influence a behaviour, discipline forces it.
</p>
<p>Conviction and discipline are both needed for consistency because humans are imperfect creatures. We might feel
  conviction to build a thing, but then sleep in late and miss an opportunity to get work done.</p>
<p>A personal anecdote comes to mind about cold showers. After learning about the positive effects for the mind and body
  from 11+ minutes of cold exposure each week, I developed a strong conviction for it. That means I have to take a cold
  shower for at least 1.5 minutes per day, and it's not something I want to do ever. I must rely on the brutal hand of
  discipline to force the behaviour.</p>
<p>The easiest way to use discipline is to force yourself to start something. Starting is always the hardest part. Once
  you've started, stopping will get harder and harder as you invest more effort.</p>
<div>
  <hr>
</div>
<p>Consistency is easy when you have conviction and are willing to force yourself to make a sacrifice. Then consistency
  will get you everything.</p>
<p><em>Thanks as always for reading &#8211; I appreciate all of you &lt;3</em> </p>

If you sail the seas of tech Twitter or LinkedIn you'll have likely noticed a trend lately: People can't seem to stop telling each other that consistency is the secret to whatever success is on the menu.

Despite this being an obvious idea, I still find myself feeling motivated by some of these posts. And it’s not news to me that doing something every day will result in progress. So what about it is motivating?

My best guess is that there's comfort in the simplicity of what consistency represents. It makes something potentially difficult and rife with unknowns into a prescriptive and manageable process. Just do the thing every day. Make the sacrifice and rewards will come.

Problem is, it doesn't matter how many times you read "consistency will get you everything", it won't make you more consistent. What makes you more consistent is two things: **Conviction** and **discipline**.

🌟 Conviction
-------------

Conviction is just a strong – bordering on delusional – belief that something is worth the sacrifices necessary to make it happen. A good example is building a company.

Over the last few months I've been exploring the idea of building a company myself. Actually exploring a lot of different ideas hoping to find one that I can't not pursue.

I've worked with early stage startups and founders for years to get product companies off the ground, so I thought I was prepared for all the difficulties of starting one. But I never knew what the period was like pre-conviction. It takes serious energy to find what matters.

Conviction is the source of energy that drives the engine of consistency. Without deep belief that something is worth spending time on, it will be an uphill battle to do it with any form of consistency.

If you want to be consistent with something, discover what gives you strong conviction about it.

🔥 Discipline
-------------

Discipline is the brute brother of conviction. While conviction wants to influence a behaviour, discipline forces it.

Conviction and discipline are both needed for consistency because humans are imperfect creatures. We might feel conviction to build a thing, but then sleep in late and miss an opportunity to get work done.

A personal anecdote comes to mind about cold showers. After learning about the positive effects for the mind and body from 11+ minutes of cold exposure each week, I developed a strong conviction for it. That means I have to take a cold shower for at least 1.5 minutes per day, and it's not something I want to do ever. I must rely on the brutal hand of discipline to force the behaviour.

The easiest way to use discipline is to force yourself to start something. Starting is always the hardest part. Once you've started, stopping will get harder and harder as you invest more effort.

* * *

Consistency is easy when you have conviction and are willing to force yourself to make a sacrifice. Then consistency will get you everything.

_Thanks as always for reading – I appreciate all of you <3_

The parts they don't tell you.

How to be Consistent

<p>I’m not a betting man, but if I was, I’d bet that I’ve read more leadership material than most people in my position. And nearly all of it ignores a key characteristic that I’ve seen time and time again in the most effective leaders I’ve studied and worked with.</p>
<p>This characteristic isn’t limited to leaders either. I’ve seen it in all kinds of effective individuals. It’s just more important that leaders do this because it makes or breaks them.</p>
<p>It’s not that leaders who don’t behave this way are ineffective, it’s that they literally don’t exist. At least not for very long. Because leaders who don’t do this go extinct, while leaders who do tend to ascend higher and higher.</p>
<h3>Roles</h3>
<p>Why do some people find it monotonous being managed (because they’re bored, not to be confused with people with an authority complex), while others are totally comfortable with a manager telling them what to do?</p>
<p>I’m sure most people reading this can think of a few people in their own circles that fit both types. But I’d hazard to guess — unless you’re part of an extremely ambitious circle — that the people who don’t like being managed are fewer in number.</p>
<p>This is how things should be. For society to work the majority of people need to be comfortable being led. An orchestra doesn’t work when everyone is a conductor.</p>
<p>But the world also needs the difficult, disagreeable, bored by management type of people. In particular it needs them to capitalise on the discomfort they feel from being told what to do all the time. The world needs them to branch off and lead for themselves.</p>
<h3>The fire is hot</h3>
<p>What happens when a person who’s bored by management decides to lead? More often than not they fall flat. Then what do they do? Well, that brings us back to the point of this essay. Because what they do next defines how effective they will be as a leader.</p>
<p>The vast majority of people, even the difficult, disagreeable ones, don’t want to make life any more painful than it needs to be. Life is already full of problems, pain, and discomfort. Why add to it?</p>
<p>Falling flat, or failing at something you care about, is one of the most painful events you can go through. Not because it’s actually bad, but because you’re so invested in the outcome that it feels like a life and death situation.</p>
<p>The human neural circuitry that feels pain from failure evolved when failure probably was a life or death situation. It probably meant you and your family weren’t going to eat for the week, or worse. For most people today, that’s far from the reality of the failures we experience.</p>
<p>I like to think of this avoidance mechanism as an illusion. Our minds see fire, so we act like it’s hot. But looking at the situation objectively, it’s only the response that gives it weight. Seeing the fire doesn’t mean you’re burning…</p>
<p>You didn’t meet your quarterly targets? Ok, well now you can see how much they really matter.</p>
<p>You can’t sell your product? Ok, well there are lessons to be learned from trying to sell it.</p>
<p>You’re nervous about a big investor call? Ok, well of course you are. Don’t overcompensate, and if it goes badly, you’ll learn something for the next one.</p>
<p>It’s the ability to see fire and move towards it that defines the most effective leaders. They refuse to play the mental game of being burned. It closes what would otherwise be an open loop; when most people would give up, the most effective people reframe and push on through.</p>
<p>So what does the effective leader do after falling flat for the first time? She notices, reframes the situation, and keeps going. She realises the fire is not so hot. Meanwhile the less effective leader feels the heat, avoids the pain, and does something less painful instead. Perhaps that’s the smarter move, I don’t know.</p>


I'm not a betting man, but if I was, I'd bet that I've read more leadership material than most people in my position. And nearly all of it ignores a key characteristic that I've seen time and time again in the most effective leaders I've studied and worked with.

This characteristic isn't limited to leaders either. I've seen it in all kinds of effective individuals. It's just more important that leaders do this because it makes or breaks them.

It's not that leaders who don't behave this way are ineffective, it's that they literally don't exist. At least not for very long. Because leaders who don't do this go extinct, while leaders who do tend to ascend higher and higher.

### Roles

Why do some people find it monotonous being managed (because they're bored, not to be confused with people with an authority complex), while others are totally comfortable with a manager telling them what to do?

I'm sure most people reading this can think of a few people in their own circles that fit both types. But I'd hazard to guess — unless you're part of an extremely ambitious circle — that the people who don't like being managed are fewer in number.

This is how things should be. For society to work the majority of people need to be comfortable being led. An orchestra doesn't work when everyone is a conductor.

But the world also needs the difficult, disagreeable, bored by management type of people. In particular it needs them to capitalise on the discomfort they feel from being told what to do all the time. The world needs them to branch off and lead for themselves.

### The fire is hot

What happens when a person who's bored by management decides to lead? More often than not they fall flat. Then what do they do? Well, that brings us back to the point of this essay. Because what they do next defines how effective they will be as a leader.

The vast majority of people, even the difficult, disagreeable ones, don't want to make life any more painful than it needs to be. Life is already full of problems, pain, and discomfort. Why add to it?

Falling flat, or failing at something you care about, is one of the most painful events you can go through. Not because it's actually bad, but because you’re so invested in the outcome that it feels like a life and death situation.

The human neural circuitry that feels pain from failure evolved when failure probably was a life or death situation. It probably meant you and your family weren't going to eat for the week, or worse. For most people today, that's far from the reality of the failures we experience.

I like to think of this avoidance mechanism as an illusion. Our minds see fire, so we act like it's hot. But looking at the situation objectively, it's only the response that gives it weight. Seeing the fire doesn’t mean you’re burning...

You didn't meet your quarterly targets? Ok, well now you can see how much they really matter.

You can't sell your product? Ok, well there are lessons to be learned from trying to sell it.

You're nervous about a big investor call? Ok, well of course you are. Don't overcompensate, and if it goes badly, you'll learn something for the next one.

It's the ability to see fire and move towards it that defines the most effective leaders. They refuse to play the mental game of being burned. It closes what would otherwise be an open loop; when most people would give up, the most effective people reframe and push on through.

So what does the effective leader do after falling flat for the first time? She notices, reframes the situation, and keeps going. She realises the fire is not so hot. Meanwhile the less effective leader feels the heat, avoids the pain, and does something less painful instead. Perhaps that's the smarter move, I don't know.

(Short) The misunderstood characteristic that makes effective leaders.

Effective Leaders

<p><em>This post is influenced the book <a href="https://bulletproofproblemsolving.com/">Bulletproof Problem
      Solving</a>, written by two McKinseyites, Charles Conn, and Robert McLean. I recommend picking it up if you want a
    deep dive into problem solving frameworks.</em></p>
<h2>Problem disaggregation</h2>
<p>&#8220;Problem disaggregation&#8221; can be summarised as: breaking down a problem into logical components so that it
  can be understood. </p>

<p>Problem solving is effectively understanding a situation well enough so it can be resolved. It&#8217;s often said
  that a well understood problem is already half solved.</p>
<p>To demonstrate the importance of understanding a problem before trying to solve it, Abraham Lincoln famously said
  &#8220;Give me six hours to chop down a tree and I will spend the first four sharpening the axe.&#8221; </p>
<p>In this sense, problem disaggregation is a tool that exposes the different parts of a problem, making it easier to
  understand, prioritise, and work with.</p>
<h3>How to disaggregate any problem</h3>
<p>The best way to learn the disaggregation process is by example, so let's look at a simple example followed by a more
  complex one.</p>
<h3>Example 1: Simple problem</h3>
<h4><strong>Problem</strong></h4>
<p>You need to get from your hotel to the airport with all your luggage in time for a flight home.</p>
<h4><strong>Context</strong></h4>
<p>You're in a country you've never visited before, and you don't speak the language. You know there are several
  transport options available, each with tradeoffs. </p>
<h4><strong>Components</strong></h4>
<ul>
  <li>
    <p><strong>Transport options:</strong> Uber, Taxi, Coach.</p>
  </li>
  <li>
    <p><strong>Time:</strong> you've got 3hrs until boarding.</p>
  </li>
  <li>
    <p><strong>Luggage:</strong> one large suitcase and a small bag.</p>
  </li>
  <li>
    <p><strong>Assistance:</strong> one helpful member of hotel staff speaks English.</p>
  </li>
</ul>
<h4>Confidence values</h4>
<p>With the components identified, they can be compared to give a confidence rating for each transport option. </p>
<div class="captioned-image-container">
  <figure><a class="image-link is-viewable-img image2" target="_blank"
      href="https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png">
      <div class="image2-inset">
        <picture>
          <source type="image/webp"
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 1456w"
            sizes="100vw"><img
            src="https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png"
            width="1200" height="675.8241758241758"
            data-attrs="{&quot;src&quot;:&quot;https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/e65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png&quot;,&quot;fullscreen&quot;:false,&quot;imageSize&quot;:&quot;large&quot;,&quot;height&quot;:820,&quot;width&quot;:1456,&quot;resizeWidth&quot;:1200,&quot;bytes&quot;:71517,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;internalRedirect&quot;:null}"
            class="sizing-large" alt=""
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png 1456w"
            sizes="100vw" loading="lazy">
        </picture>
      </div>
    </a>
    <figcaption class="image-caption">The formula to find the confidence value for each transport option</figcaption>
  </figure>
</div>
<p>Confidence is subjective but it's still measurable. By weighting each variable based on its subjective importance,
  the different components can be independently measured.</p>
<div class="captioned-image-container">
  <figure><a class="image-link is-viewable-img image2" target="_blank"
      href="https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png">
      <div class="image2-inset">
        <picture>
          <source type="image/webp"
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 1456w"
            sizes="100vw"><img
            src="https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png"
            width="1200" height="715.3846153846154"
            data-attrs="{&quot;src&quot;:&quot;https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/afbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png&quot;,&quot;fullscreen&quot;:false,&quot;imageSize&quot;:&quot;large&quot;,&quot;height&quot;:868,&quot;width&quot;:1456,&quot;resizeWidth&quot;:1200,&quot;bytes&quot;:120203,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;internalRedirect&quot;:null}"
            class="sizing-large" alt=""
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png 1456w"
            sizes="100vw" loading="lazy">
        </picture>
      </div>
    </a>
    <figcaption class="image-caption">Exploring which transport option is safest based on weighted variables (<a
        href="https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing">raw
        data</a>)</figcaption>
  </figure>
</div>
<h3>Example 2: Complex problem</h3>
<p><em>This example is borrowed from Bulletproof Problem Solving.</em></p>
<h4>Problem</h4>
<p>A family is moving home. Where should they live?</p>
<h4>Components (weighted)</h4>
<ul>
  <li>
    <p>Proximity to good schools [0.35]</p>
    <ul>
      <li>
        <p>&#8230;2x sub variables</p>
      </li>
    </ul>
  </li>
  <li>
    <p>Nice environment [0.25]</p>
    <ul>
      <li>
        <p>&#8230;3x sub variables</p>
      </li>
    </ul>
  </li>
  <li>
    <p>A cool, friendly town [0.25]</p>
    <ul>
      <li>
        <p>&#8230;6x sub variables</p>
      </li>
    </ul>
  </li>
  <li>
    <p>Not too far from everything [0.15]</p>
    <ul>
      <li>
        <p>&#8230;1 x sub variable</p>
      </li>
    </ul>
  </li>
</ul>
<h4>Weighted score</h4>
<div class="captioned-image-container">
  <figure><a class="image-link is-viewable-img image2" target="_blank"
      href="https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png">
      <div class="image2-inset">
        <picture>
          <source type="image/webp"
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 1456w"
            sizes="100vw"><img
            src="https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png"
            width="1200" height="1034.3406593406594"
            data-attrs="{&quot;src&quot;:&quot;https://bucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com/public/images/f3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png&quot;,&quot;fullscreen&quot;:false,&quot;imageSize&quot;:&quot;large&quot;,&quot;height&quot;:1255,&quot;width&quot;:1456,&quot;resizeWidth&quot;:1200,&quot;bytes&quot;:305027,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;internalRedirect&quot;:null}"
            class="sizing-large" alt=""
            srcset="https://substackcdn.com/image/fetch/w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 424w, https://substackcdn.com/image/fetch/w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 848w, https://substackcdn.com/image/fetch/w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 1272w, https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png 1456w"
            sizes="100vw" loading="lazy">
        </picture>
      </div>
    </a>
    <figcaption class="image-caption">Exploring which location is best based on weighted variables (<a
        href="https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing">raw
        data</a>)</figcaption>
  </figure>
</div>
<div>
  <hr>
</div>
<h2>Further use cases</h2>
<p>I hope you've found learning about disaggregation theory as revealing as I did when I first stumbled into it. </p>
<p>The disaggregation framework is a real Swiss Army knife. It's useful across domains, and with problems of different
  complexities. It can be applied to business questions like "should we market to segment X or Y," or "which product
  idea higher risk".</p>
<p>If you want to learn more about framing the problem well, <a
    href="https://nosaj.substack.com/p/software-that-makes-money">read my previous post</a> which explains how to define
  problem statements.</p>
<p>Catch you soon,<br>&#8211; Jason</p>
<p><em>PS &#8211; You can explore the data used in this post in <a
      href="https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing">this
      Google sheet</a>!</em></p>

_This post is influenced the book [Bulletproof Problem Solving](https://bulletproofproblemsolving.com/), written by two McKinseyites, Charles Conn, and Robert McLean. I recommend picking it up if you want a deep dive into problem solving frameworks._

Problem disaggregation
----------------------

“Problem disaggregation” can be summarised as: breaking down a problem into logical components so that it can be understood.

Problem solving is effectively understanding a situation well enough so it can be resolved. It’s often said that a well understood problem is already half solved.

To demonstrate the importance of understanding a problem before trying to solve it, Abraham Lincoln famously said “Give me six hours to chop down a tree and I will spend the first four sharpening the axe.”

In this sense, problem disaggregation is a tool that exposes the different parts of a problem, making it easier to understand, prioritise, and work with.

### How to disaggregate any problem

The best way to learn the disaggregation process is by example, so let's look at a simple example followed by a more complex one.

### Example 1: Simple problem

#### **Problem**

You need to get from your hotel to the airport with all your luggage in time for a flight home.

#### **Context**

You're in a country you've never visited before, and you don't speak the language. You know there are several transport options available, each with tradeoffs.

#### **Components**

* **Transport options:** Uber, Taxi, Coach.
    
* **Time:** you've got 3hrs until boarding.
    
* **Luggage:** one large suitcase and a small bag.
    
* **Assistance:** one helpful member of hotel staff speaks English.
    

#### Confidence values

With the components identified, they can be compared to give a confidence rating for each transport option.

[![](https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fe65933b9-7f6b-4125-9bdc-df4d65f1cd2d_3240x1824.png)

The formula to find the confidence value for each transport option

Confidence is subjective but it's still measurable. By weighting each variable based on its subjective importance, the different components can be independently measured.

[![](https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fafbfdd2d-be61-4a07-a13e-10a3a8e7700d_3240x1932.png)

Exploring which transport option is safest based on weighted variables ([raw data](https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing))

### Example 2: Complex problem

_This example is borrowed from Bulletproof Problem Solving._

#### Problem

A family is moving home. Where should they live?

#### Components (weighted)

* Proximity to good schools \[0.35\]
    
    * …2x sub variables
        
* Nice environment \[0.25\]
    
    * …3x sub variables
        
* A cool, friendly town \[0.25\]
    
    * …6x sub variables
        
* Not too far from everything \[0.15\]
    
    * …1 x sub variable
        

#### Weighted score

[![](https://substackcdn.com/image/fetch/w_2400,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Ff3ff931b-3805-4fc6-b0c9-b9e5b63e1580_3240x2793.png)

Exploring which location is best based on weighted variables ([raw data](https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing))

* * *

Further use cases
-----------------

I hope you've found learning about disaggregation theory as revealing as I did when I first stumbled into it.

The disaggregation framework is a real Swiss Army knife. It's useful across domains, and with problems of different complexities. It can be applied to business questions like "should we market to segment X or Y," or "which product idea higher risk".

If you want to learn more about framing the problem well, [read my previous post](https://nosaj.substack.com/p/software-that-makes-money) which explains how to define problem statements.

Catch you soon,  
– Jason

_PS – You can explore the data used in this post in [this Google sheet](https://docs.google.com/spreadsheets/d/1XPuDdN1AzAgfpFgksabOOBcqDKuZBKue37VxF4wGkQE/edit?usp=sharing)!_

Today I'm excited to share a high leverage framework that makes complex problems easy to navigate with little effort.

Solve Problems Really Fast

The art of computer programming has produced more entrepreneurs than any other creative discipline in human history.

A programmer is rarely short of motivation to build new things, but nobody, no matter how experienced, can predict how the market will react to something new. Knowing **what** to build in the first place requires a process informed by robust principles.

So how can a software engineer learn to produce not just great code, but inherently valuable software that people will love? Keep reading and I'll share the process I've used to achieve this many times in my career. A process called "customer development," or "prototype-driven development".

#### Prerequisites

Reading this guide should shift your perspective about value creation from product framing to problem framing. Get used to seeing problems as the North Star, and not the product.

There's three things that will make this process easier for you:

1.  The willingness to work hard for little potential reward for an unknown amount of time.
    
2.  Have things in your life you're actively passionate about.
    
3.  The ability to build basic software and ship it. It can be done without writing code, but you will have less freedom.
    

Finally, engineers: remove writing code from your mind for now. Coding is the comfort zone, and it's not going to be productive to ship anything until you understand the problem you want to solve. So let's start with the problem.

* * *

Step 1: Problem discovery & value proposition
---------------------------------------------

Outcome for this step: A list of problem statements that are specific, measurable, and action-orientated, and a few value propositions for each one. See end of section for an example.

**Problem discovery** is about building a deep understanding of _specific_ problems that _specific_ people face day-to-day. The word specific is emphasized because problems must be precise, measurable, have clear boundaries, and be grounded in reality.

**Value propositions** capture how you intend to solve the problem, framed from the point of view of the customer. One sign of a good problem statement is that the value props will often reveal themselves. A good problem statement has many potential value propositions.

Focus on things you're already attracted to working on when going down the problem rabbit hole. Imagine you're going to work every day on this for the next three to five years. Ask yourself "will I lose interest in this?"

Here are some tried and tested ways to think through finding problems:

> **🔁 Write down your routines.** Be detailed about this. Write down each discrete step. Then challenge each step: what could be easier about it? What happens if the step is removed?
> 
> **💸 Count what you spend money on.** You might learn more than you expect by going through your transaction history. Note down the things you purchase regularly. Ask yourself 5 why's for each one to discover your underlying motives.
> 
> **😓 Look for wasted energy.** This one can be done for yourself or other people. Problems that are very painful will make solutions way more attractive. Focus on the painful over the inconvenient and the boring.

#### Example: Problem statement and value proposition

**Problem Statement**: Independent video creators don't have much time to re-edit videos for short-form platforms like TikTok, Instagram, and YouTube Shorts, but doing so would be advantageous.

**VP1**: A plugin for video editing software that exports clips along with the final edit, saving time.

**VP2**: Autonomous software that prepares and uploads clips with no human input needed.

**VP3**: An editing team that can be outsourced affordably, specializing in short-form viral content.

#### What if the problem was already solved?

[![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F7fb316be-48dd-45a2-b366-1e91ad8c72fa_2160x1520.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F7fb316be-48dd-45a2-b366-1e91ad8c72fa_2160x1520.png)

Focus on an underserved (niche) segment of the market to deliver move value.

It's hard to think of a perennial product or service. Everything is open to change at any time, and if there's a solution on the market better than the status-quo, it will be valuable — even if it's valuable to a subset of customers.

The specific "who" in your problem statement is useful here. It directs you — the value producer — to serve a group of customers, also known as a niche. A large ratio of the fastest growing startups got going by capitalizing on an under-served niche.

Don't be afraid of small niches. If you can secure 1,000 customers globally paying $10/mo, that's $120,000/yr revenue.

[![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fb367affb-1419-4269-b2b8-64783bc9deb1_2160x1520.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fb367affb-1419-4269-b2b8-64783bc9deb1_2160x1520.png)

For a solo maker or small team, a small slice of the market is enough.

Step 2: The first prototype
---------------------------

Your first prototype will test your value prop with your chosen market segment. You can, and should, run the test with minimal effort. You can do it with a press release, a tweet, a reddit post, or a webpage.

A successful outcome will get potential customers to show interest. Interest can be measured with a simple "email me when it's ready" form, an "I'm interested" button, or by counting the number of readers or page views on the webpage.

Get comfortable with the customer development loop. You can use the same methodology to drive more value creation, test business models, and prevent mistakes as you grow.

We prototype and test for the same reason a rock climber leading the climb anchors into the rock face as they climb higher. The climber could probably reach the top without falling, but if they overcommit and fall from a great height without an anchor, they are more at risk of disaster.

[![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fd351aad9-a028-4478-a5e7-dd2d3fb63bc0_2160x1520.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fd351aad9-a028-4478-a5e7-dd2d3fb63bc0_2160x1520.png)

Each release delivers new information which leads to a better product.

**Steps to test different value propositions or solution statements:**

1.  Pick 3-5 solutions to test. Make sure solutions are anchored to the problem statement.
    
2.  Go to where the customer is. Social media is a good start, but get as close as possible, use forums, Facebook groups , etc.
    
3.  Observe the reaction to each post, and ensure you collect any feedback you might have received to work back into the process.
    

When you have your data and feedback, you should start to see a clear picture of which solution could win out.

Step 3: Prototype loop
----------------------

[![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F525e4513-6f1d-4f23-b699-dfe81021f573_2160x1520.png)](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F525e4513-6f1d-4f23-b699-dfe81021f573_2160x1520.png)

The prototyping process leads to ever-improving products.

After the results are in from the first prototype loop, the process can start again on the next one. The next loop could be building a demo, or testing further value propositions until you have the direction dialled in.

Functionally speaking, prototyping (or customer development) is the process of going from low resolution to high resolution, using lessons from the current step to improve subsequent steps.

A better way to build?
----------------------

There's an art to customer development. Like [TDD](https://en.wikipedia.org/wiki/Test-driven_development), it can over-complicate or distract focus from customer experience.

When it makes sense to use it, customer development feels like you have a treasure map that nobody else has seen. It invites the beneficiaries of a product to help shape it, and I personally find that to be a more meaningful way to work.

Good luck building!

Zero to one startup building & engineering

Connect

Posts

AI Interviews as a Service

The End of Mastery

Why Twitter Deserves to Live

How to be Consistent

Effective Leaders

Solve Problems Really Fast

Make Software That Makes Money

Zero to one startup building & engineering

Connect

Subscribe to the blog

Posts

AI Interviews as a Service

The End of Mastery

Why Twitter Deserves to Live

How to be Consistent

Effective Leaders

Solve Problems Really Fast

Make Software That Makes Money