Hacker News Reader

Comment by scrollbar | Hacker News Reader

This is super cool.

In case you don't make it through to the project's webpage howisfelix.today, and read through to the end, here is the creator's conclusion:

Overall, having spent a significant amount of time building this project, scaling it up to the size it’s at now, as well as analysing the data, the main conclusion is that it is not worth building your own solution, and investing this much time. When I first started building this project 3 years ago, I expected to learn way more surprising and interesting facts. There were some, and it’s super interesting to look through those graphs, however retrospectively, it did not justify the hundreds of hours I invested in this project.

I’ll likely continue tracking my mood, as well as a few other key metrics, however will significantly reduce the amount of time I invest in it.

View on HN

> There were some, and it’s super interesting to look through those graphs, however retrospectively, it did not justify the hundreds of hours I invested in this project.

This is my general feeling towards all quantified-self projects. Perhaps it creates interesting insights for some people, but I can imagine the kinds of conclusions I'd actually get from it will be things like "I'm happier on sunny days" and I'll say "yeah..."

You’ve verbalized more about data science than most executives ever will.

It's a bit more nuanced. It's true that you do research looking for insights that would give you a competitive advantage, and you don't find many - and very often the results don't justify the cost. On the other hand, if you don't do that, you risk staying behind, so some form of data analysis is absolutely necessary for any business - you can't just run on your gut feeling especially once you exceed a certain size.

As a side note, people tend to be very unimaginative in the initial stages of data analysis: very often they just take the data they have instead of thinking what data they need to have. And obviously the results correspond to the initial limitations. So much effort is put into data munging and other stages when your inputs are incomplete or plain irrelevant to what your business actually needs.

Data Science manager here. I've worked with companies that didn't know the profitability of their different client contracts because the data was stored in separate systems. Fixing that kind of thing is super profitable. Every additional data project after that adds less incremental value, and it's easy to hit the point where more analysis doesn't make enough money to be worth it.

Maybe the executive is asking the wrong questions and puts it on the data scientist. Hint : 'Can you create a report that shows that my last decision was a success ?' is not a good question.

It’s a great question when the desired outcome is climbing the corporate ladder

"Why am I the best?"

For me "quantified self" has been half meditative. I have tracked daily water intake (and only water to keep it simple) and that has been a huge noticeable quality of life improvement. Waterminder on apple watch is the best.

This year I added in daily mood tracking, and it's mostly focused on taking 5 minutes a day for myself to just jot notes down quickly. I'm using dailo and I have a few categories too (work good/ok/bad - food homecooked/takeout - vices) and just the space I've created to do that has been worth it. I don't expect to get anything too interesting long term besides patterns in takeout/cooking, but I find myself more connected to my day good or bad, and that presence of mind to combat the autopilot is worth it.

What was the quality of life improvement in tracking water intake? I find it interesting, because your body tracks it whether you want to or not, and I never found forcing myself to drink more water beneficial.

It’s surprisingly easy for some to get disconnected from the signs of dehydration. For example: when focused on a project I can go a couple days before realizing that I did not drink any water at all.

All that dehydration talk is fascinating to me. It’s like most people are aliens. I drink about 6L per day just in tap water, not including coffee and tea. I love drinking water and could not imagine limiting myself to what others perceive as their daily goal. And as the question tends to come up: I’m not diabetic, and I have a home-office job. I do vape.

Are you not drinking anything to supplement your loss of water? You don't feel thirsty?

Not usually, and no don’t feel thirsty.

The signs I easily ignore are a little bit of brain fog, some reduction in physical energy, and dry skin. The dry skin is meh, and the other two are solved* with food that provides energy.

* = It does not actually resolve it, but it does “enough” and has a long enough lead time that it’s easy to have an instinct of “Oh I’m feeling run down, time to grab a snack” followed by “that didn’t quite get where I expected, I’ll give it time to digest” before forgetting about the whole thing as the project reabsorbs my attention.

I briefly tracked water intake since it was a feature in the Garmin app I was using for activity tracking. Ironically it made me worse off, because it interrupted my (extremely simple) natural hydration process.

There were 3 customizable preset capacities, but even using those required a couple taps each time. Drinking a non-standard amount and manually adjust it required extra effort and taps. Sometimes I’d forget to record it right away, then have to try to remember if I was double-counting a refill. The mental burden was clearly out of balance with the (practically nonexistent) reward.

I quit tracking it when I found myself adjusting my behavior, like avoiding the natural habit to top off my water bottle when it was less than half full and I was getting up from the desk anyway. Ultimately it would be interesting to have that data, but it’s only worth it to me for zero cost/effort.

I think for me it helped me to see how often when I was thirsty I might actually eat a snack.

From there, it's also valuable because it's clear that in the past I might have a dehydration headache, eat something, and still have a headache because I was thirsty. By just incorporating a routine and simple 1 click tracking, that consistency has helped me to easily manage this. But as you said, for you it works "naturally". For me, I was pretty out of tune with that.

Human life is just not that interesting it seems. Even when it's yours.

It doesn't make a lot of sense to try to correlate things with your mood that you think are correlated. Because giving a mood grade is biased. For example if you feel like you don't sleep enough, the day you will sleep a lot you will want to give a higher grade.

It's impossible to do a proper blinded experiment here, but in this case, even if you're measuring a placebo effect it's still meaningful.

If you think you should be feeling good today because you're well rested, you probably will feel better.

"Am I really happier, or do I just think I am?"

It can be more useful in some cases. E.g. my Fitbit does get me to move more when I notice I've done too few steps and tracking mood, productivity etc. is much more useful when you actually try things like nootropics to actually figure out if they seem to help you.

It's still useful for uploading in a couple of decades. The more digital crumbs we leave, the more will be uploaded. Has anyone tried to fine-tune GPT-3 on their own comments?

As a joke/experimental project, I trained gpt-2 (don't have access to 3) on Joe Biden speeches/interviews/press conferences, cloned his voice (mumbling and all), created a 3d model of him, and connected Twitch chat to make it interactive for the audience.

Didn't do it for politics, just for lolz and prototyping a concept. Got to Twitch's advertising unlock, but stopped running it because it was a headache to maintain with my home setups/remotely.

But yeah, not trained on my comments, but someone else's, and it didn't take a ton of content to get it going for gpt-2.

Edit: Here's a clip for reference

https://clips.twitch.tv/LovelyObedientTaroNerfBlueBlaster

I don't feel you will be able to detect big correlation by using that kind of system. If there is a big correlation to be found, it is likely to have already been found by researchers.

I'm tracking my life by a similar system than OP, but the error with tracking is to start tracking useless things. For example OP track his calorie intake, it is a lot of work, and it is not really usefull at the end. It is unlikely to correlate with any other measures he is doing (apart from weight).

I personally only track mood, activities and position. Each tracking has a specific goal, and only take me 1 minute per day. I have a lot of headaches and could be tempted to track headaches to understand what triggers them. But now that I know what triggers them (when I'm hungry, or I spent too much time exposed to the sun) I would never have been able to detect that correlation from data.

So my advice to people wanting to track their life, identify first what you need the data for, and increase the number of data you track if you have a lack of data to achieve a specific goal, but don't track for the sake of tracking.

For example I track my position, to know where I have been, but also to warn my parents in case I go missing and nobody know where is my last position

I track my mood, to know if I am doing better over time, or if I'm able to improve my sleep not only by feeling it, but by having numbers on

I track my activitites, because I want to know If I work too much, if I do enough sport, or if I spend too much time on my computer in a week

>> For example OP track his calorie intake, it is a lot of work, and it is not really usefull at the end. It is unlikely to correlate with any other measures he is doing (apart from weight).

Tracking calories is very very simple if you use an app like MyFitnessPal. For most foods you can scan the barcode or you can just enter the calorie data from the label.

Secondly, isn't weight a pretty important measure to track? It can impact a lot of things and therefore if you see you start to have too many days in a week that are above your calorie goal you can catch any gain before noticing it on the scales.

Author here - Tracking calories has impact on way more things than just the weight. e.g. choosing a low-fat diet vs a low-carb diet makes you feel very differently day to day, and every person might prefer something else. Also, there is a huge difference in energy levels and quality of sleep while I'm cutting vs while I'm gaining weight.

I don't even get why that's useful.

Just reflect on yourself at the end of the day.

I think it's fascinating - and horrifying - that he tracks everything except his environmental impact on the planet.

He has to be in the top 1% of polluters, given his obsessive flying and consumption of technology.

Dozens if not hundreds of charts, depicting every single thing except the thing that is destroying our world.

I filed an issue on his github.

Fun fact: British Petroleum coined the term “carbon footprint”. It had gotten to the point where it couldn’t deny climate change. So it was time to get well-meaning middle class folks to blame each other instead of the petroleum industry. (See also in recent years the obsession with plastic straws and plastic bags.)

Agreed, the metrics he uses for CO2 consumption via flying are misleading: he compares the flying against total average CO2 production per US Citizen per year. Clearly, he's doing more than just flying in his life, so the comparison is meaningless.

I also find it horrifying that HN folks downvoted you for your well-founded opinion.

I would be interested in detection and correlation abilities here. I think simply collecting the data and graphing it doesn't help.

However, if enough metrics were automatically collected (when did I fall asleep and heart rate lower?) and there was some system that could sort through them (eating a 4th meal increases odds of ruby team scoring additional points by a factor of 1.3x) then you could possibly learn some neat things.

Some of the most interesting things have been pointed out to me by others. Things that are related but I didn't notice. Foods that seem to increase my chances of sickness and things like that.

Men will literally devote hundreds of hours to building a bespoke database tracking every moment of their lives instead of going to therapy.

Instead of living, I would say.

When reading above the experience I have is that I rapidly correlated with apple watch heart data that my heart rate stays high when I drink alcohol in the evening, resulting in poor sleep.

Conversely I can see when I have been likely to have been drinking from my heart rate.

I think the key here is the passive data collection with intuitive data presentation to drive low barrier to entry analytics.

The problem is waiting for three years to gain insight. This non-lean approach fails because in retrospective analyses context is lost, there's hindsight and survivorship biases. The opportunity to discover more relevant data sources is missed.

Reply on news.ycombinator.com