An antidote to burnout

On adjusting the scope, so that it can be finished.

Dec 28, 2024

Every day I was slipping deeper into that status quo. It was soul-crushing being under the ambivalence of a role that neither rewarded brilliance, nor punished mediocrity. I was turning into a person that only reacts to immediate pressures. If it wasn’t in my inbox, I wasn’t going to care. The longer I let this go on, the duller I would become. I knew I needed to get out, and build my way to freedom.

I know enough as a product manager, that the road to good business is a good understanding of people and their problems. And that good business is a road to freedom. Making premature solutions is an insidious trap most of us fall for. But talking to people is hard. Finding them, teasing out their actual needs and desires, and often doing this without a clear goal in mind — it’s draining.

I was pushing through several months at this point, finding people to talk to, learning what I could and thinking really hard (lol) before I woke up one morning, and decided I couldn’t do anything useful that day. Or for several days after, which turned into a couple of weeks.

The burn out I experienced felt like the whisper you get when you’re running too fast. Flying down the road, moving my muscles, feeling my heart pump, it’s great, but eventually the whisper starts. The thought of stopping, of slowing down. It grows, as I feel more uncomfortable, as the oxygen struggles to fuel me in pace with consumption. My mental strength resists for a while, as the whisper grows into a plea, and eventually a command. Some switch inside me flips, and I slow down to a crawl, or stop entirely. The crash feels proportional to the push. The more pressure I put on myself to achieve, knowing that it was possible for me, if uncomfortable, the more indifferent I feel after stopping.

Ari, Viktor and I are in an accountability group together. We meet weekly and on one session we got to talking about AI. This was during my quest to talk to people, before I crashed. Ari asked if AI could make phone calls, as he could see avenues in the medical space where this could be useful. Being a compulsive teacher and a touch of a show-off, I plunged immediately into a “yes, but no” answer. You would need to convert your caller’s voice into text, run it through an LLM, before converting that output into a generated voice. And people on a phone call have an expectation of a quick response. After my monologue, Viktor calmly shared a link to an online demo1.

Not only did that prove me wrong. It was a working example of an extendable voice-to-voice tech with latency of < 500ms. Faster than my grandfather could speak, maybe not as fast as someone with ADHD. More than good enough.

This was inspiring. Where voice could be an interface into any application out there.
Pausing my main efforts at a startup, and freedom, I took a brief side quest to bang out a prototype of this that I could use to help me take notes, something I love doing.

Standing there in my office, with a timer running I started to talk to it. And it worked, but there were flaws. As I rambled out loud, the prototype would always respond after a pause. “Ok”, “Understood, Josh”. Which broke my frail line of thought. And some of the phrases were a little different to what I had said, it had taken some liberties in processing my words. It was annoying. So I parked it.

After my burnout/crash, I knew I needed to get back into the saddle.
Althought the thought of restarting my startup (restartup?) caused my stomach to churn, but freedom wasn’t going to land on my lap. I decided that finishing the little side project would be fun, and if I could get it complete, would be worth while.

Finishing things is so important – my greatest accomplishments were things I finished. Sounds silly right? It’s true though, unfinished things don’t count for much.

To borrow from the running analogy, I wanted to adjust my pace to be something much slower. So that I could complete the distance without stopping. For the sake of a win and for the sake of playing the game.

I was going to finish making this thing, then start with a clear mind on the next project, whatever that may be.

There were a lot of directions, features and challenges I could see with building some note taking app. Firstly, who was it going to help?
The biggest challenge I had with my startup with finding the right people to serve, finding out their needs. So to make it easy on myself, and the biggest change of pace, I was going to make it fit my needs alone. I would fantasize about it being useful for others, but build it for me.

The second, was that this shouldn’t remain a project only . My goals of freedom, financial and creative, were still the overarching driver. And to satisfy those needs, this needed to be a complete product. Meaning, it should be publicly available, should accept payment and should do something useful. I consider it a personal product. To the point where I need to pay for it myself, no freebies.

So what am I building now?

An infinite journal. Going to take notes on my A5 spiral bound notepad and photos of it. I’ll type out entries when it’s quicker. And I’ll record voice memos. From these sources, I want to gather similar chunks together and create lists, discrepancies and anything AI can help me on the data with.

The requirements to be accountable to my goals of freedom are:

It needs to be public
It needs to accept payment
It needs to work without any hacks or workarounds

The design involves four components,

One. Capturing notes — voice, text and photos of notepads.
Two. Digesting those notes into topics. Summed up notes in one place, like my shopping list or my goal of one day doing a 5km in 20 minutes. Every note adds a bit to different topics, some which are known (because I told the journal they’re topics) and some that emerge from AI.
Three. Reflection. Notes will often have little to no context. AI can be made to seek clarity when I’m too lazy to add it or too blind to see it as missing.
Four. Self-improvement. AI is going to mess stuff up initially, and going to be amazingly better in the near future. To fix the bad and keep up to date with the new, requires building a library of examples and expectations. That can be used for guiding the AI.

With this design, here is how I imagine the day-to-day of using this app.

I take a walk. Airpods in, ramble out loud and recording it into a voice note. The note is transcribed and as soon as possible, there is a moment of reflection where AI prompts me for missing context and clarification. Any corrections on mis-categorized topics are added into the example library.

Later, I decide to write an essay about a topic, to clarify my thinking on it. I ask my Infinite Journal about the topic and I’m given a map of past thoughts and explorations on it.

I don’t want it to write the essay for me, I’m not that interested in producing AI generated content. Instead I want to see if it can remove the incidental complexity of writing, ease the friction a little by providing ample material to work from. Part of why I write is to clarify my thinking, which requires effort. Like, I wouldn’t want a robot to go to the gym and lift weights for me.

The first component is underway. Capturing voice memos, using Google/Microsoft to capture hand written documents. It’d be nice to build an app for this, but the tools already exist and I need to do as much as I can manually and quickly to get to something that is useful for me. All good things are simpler than we think. And all software is really just automation.

My startup failed to gain traction because I couldn’t keep up the effort of talking to people. A step that I still believe is necessary and that I will get back into. While my brain feels like I strained a muscle somewhere, I’m going to complete this personal product. If you can’t finish a marathon, run a 5km.

Time to go through 71+ notes.