I do find vibe coding fulfilling. Very much so. But not in the way someone who gets into a flow state with coding typically does.
Why? Because I can finally make rapid progress on ideas I've had that would have taken me years to develop by myself. I enjoy the outputs. Only the outputs.
There is genuinely no flow state for me. It is a grinding list of frustrations I handle whenever the agent falls short of my expectations.
It is assembly work in that sense. I am an order of magnitude more productive, but the work itself is not intrinsically enjoyable the way coding used to be.
I am a factory worker churning out features as I monitor the LLM agent and its outputs. There is no craft. Only productivity.
I do not mind because I was never a very competent programmer to begin with, so there is no major loss on my part. This black box produces what I truly want, but I will not pretend it has any meaning for me. It provides no more meaning to me than cranking a lever for hours on end.
But I love seeing my ideas come to life. These distant dreams I had only in my head are suddenly walking and breathing after years of only thinking about them.
I used to have far more, but I learned the habit of bookmarking tabs that I've kept open for a long period of time.
Actually, now that I think about it, this would make for a useful browser extension: have an LLM that automatically asks to bookmark tabs that have been inactive but present for X number of days.
If you could do calculus in high school, you likely have more than enough neural horsepower to build a wide array of commercial products with AI, since it can breakdown, suggest, recommend, and educate you on the conceptual gaps that remain.
The hard part is that you'll have to inefficiently learn about these concepts as you encounter them, as opposed to a more skilled engineer who can already anticipate and use them ahead of time.
In general, increasing GDP per capita has all sorts of positive second and third order effects on net, likely improving the collective welfare of society.
It's a bit far removed from what you might envision as "doing good" though. It's a bit too abstract to feel on an emotional level.
And true, not all acts are equal. The emotional valence of improving ad rates for Meta can't really be compared to saving a child from cancer on a moral scale that most humans intuitively understand. Nor would I demand that you try to do so.
I often find that "making the world a better place" is synonymous with "looking for meaning in my own life".
Having children is usually a good start, if you don't have any already. Making the world a better place for them does wonders for the soul.
Or so I've been told.
Alternatively, if neither children nor economics appeals to you, you can always just donate money to St. Jude's Hospital or another charity. The money you earn would then make the world a better place in a way you can directly feel.
In a tweet, the account for Similar Web said they don't actually measure iOS app traffic. They just measure traffic to the specific URL (https://chat.openai.com/).
If you notice the app came out in May 2023 and traffic decreases for the first time in the following month in June 2023.
Additionally, you may notice the tweets refer to a specific chart, which might give the impression if you go on the site, they have the actual data, but this is not so.
As of July 7, 2023, Similar Web can't measure iOS app usage. Only their relative ranking on the App Store which doesn't tell you much about daily traffic usage, although we might be able to infer some patterns.
"Usage data is currently available only for apps in Google Play Store. We're working hard to support iOS apps soon"
Additionally, the article has this to say, which might also give a false impression:
> Downloads of the bot’s iPhone app, which launched in May, have also steadily fallen since peaking in early June, according to data from Sensor Tower.
Downloads isn't the same as traffic. Once you've downloaded an app, you don't generally delete it, then download it over and over again. You just use it and that isn't being measured. Although I would expect traffic usage to slow down or decrease somewhat as downloads decrease as well. I just think there are other factors playing a role here as well that aren't being documented.
Growth can't happen forever and it may be that it's slowed down considerably, but it's also possible the decrease is due to users switching the platform they use to access ChatGPT. As it stands, it's not clear from the data from what I've seen. It may be that users switched en masse to the iOS app or something else entirely.
Would need more data to clarify and confirm.
Also, what the heck is up with this tagline for their newsletter? It sounds so adversarial.
"Tech is not your friend. We are. Sign up for The Tech Friend newsletter."
Why? Because I can finally make rapid progress on ideas I've had that would have taken me years to develop by myself. I enjoy the outputs. Only the outputs.
There is genuinely no flow state for me. It is a grinding list of frustrations I handle whenever the agent falls short of my expectations.
It is assembly work in that sense. I am an order of magnitude more productive, but the work itself is not intrinsically enjoyable the way coding used to be.
I am a factory worker churning out features as I monitor the LLM agent and its outputs. There is no craft. Only productivity.
I do not mind because I was never a very competent programmer to begin with, so there is no major loss on my part. This black box produces what I truly want, but I will not pretend it has any meaning for me. It provides no more meaning to me than cranking a lever for hours on end.
But I love seeing my ideas come to life. These distant dreams I had only in my head are suddenly walking and breathing after years of only thinking about them.