A lot of classic software essentially worked more like a database. In the last 10-15 years it's all moving to an algorithm.
Here is what I mean. Photos apps used to let you search through your photos using filters.
The same kinds of things are happening on the web which already happened to apps (desktop and mobile).
In the modern world, some marketing company wants to tell YOU which of YOUR photos you wanted, so they can sell you some prints, harvest your data, or something.
I would like any apps that have to do with collections of files, photos, music, etc to be more of a deterministic DATABASE and less of a nondeterministic algorithm.
> A lot of classic software essentially worked more like a database. In the last 10-15 years it's all moving to an algorithm.
You just described what I missed about the older software. Older software gives users control over sorting and show data in a tabular format. Modern software sorts data with an algorithm, with ads mixed in, and shows data in a card format, making it a lot less usable.
Exactly. My related observation: half of the SaaS products I see would be more useful and ergonomic for the user if they were implemented as an Excel sheet.
(I actually worked for one of such "better off as an .xls file" startup in the past, and its main competitor was an incumbent that sold the same stuff as an Excel extension. Trying to replace that with a React app is not a worthwhile use of life.)
Algorithms are fine. I'll happily apply the most advanced ones I can get. The problem is with who applies them to what - as you and GP said, it's about user control - or, currently, lack of.
I wonder if being a more powerful backing data store for Excel is one of the remaining reasonable uses for Microsoft Access, at least for users of the Windows desktop version of Office? Access is still included in many editions of that (although not on Mac or web or mobile). Officially it’s even still supported and not deprecated, although of course it’s very much not emphasized by MS any more.
Other options might be SQL Server Express or SQL Server Express LocalDB, the latter of which seems conceptually very much like SQLite within the MS ecosystem, and both of which are usable for production purposes at no cost within the technical limitations that differentiate them from paid editions.
The SaaS I am now working for is a react app + some fancy intelligence.
We are not front-end people, so the app is built with the expectation that people will be doing their filtering, searching and using the intelligence we provide, but in their home turf (excel).
Our app also lets users "track" certain events, and we do not use push notifications, rather we respect our user's attention and email them a short summary, and link to a csv that they can use!
That immediately made me think of the digital TV switchover. The elderly father of a friend of mine would spend much of his time in front of the TV, and could operate it without assistance, thanks to the simple 1:1 mapping between buttons and functions.
After the digital switchover, there was now a set-top box, and electronic program guide and three-figure channel numbers thrown into the mix, as well as stateful aspect such as whether the TV was set to AV or still trying to use its now-obsolete tuner.
For someone with poor eyesight, limited feeling in his fingers and limited ability (and admittedly willingness, too) to build a mental model of how the menus worked and how they can be navigated, it spelled the end of his unsupervised access to TV.
The big difference for me between database-query-driven and algorithmically-driven is that the latter makes it very hard to know when you've completed an exhaustive search. Indeed for the likes of meta and tiktok that's a feature, not a bug, since their goal is to keep you engaged and plugged into "their" content forever.
I have a "Roku TV". It has many pointless problems. But the biggest is that it takes a very long time to turn on. If you turn it on by remote control, there is no indication that anything has happened until the TV finally starts showing something. You just have to hope that the signal got through. It's fairly frequent that we fail to turn the TV on because we assume that it just isn't done getting ready to turn on yet.
You can avoid this by using the physical power button, which is conveniently located behind the TV. It will still take forever to turn on, but there's no ambiguity over whether you've started the process.
I still have trouble believing the device was allowed to reach customers in this state.
Not every change is for the better. You gotta admit that TVs used to be able to switch channels much faster then they do now. And analogue controls in cars are safer and better then touch screens for everything.
A lot of change is for the better, but quite a lot is a regression.
I have access to around 1000 "channels", if you include live broadcasts and network-like apps. How exactly were old TVs better at helping navigate that?
For one thing, the clicker seems to work instantly for analog TV because it will sync up on the next frame, which is 0.03 sec which is less than the 0.2 sec it takes the human mind to perceive two events as two different events.
With digital broadcast TV and cable whenever you switch to a different carrier there is a long delay (at least 0.5 sec) for the radio and the rest of the processing train to sync up. With streaming you have to do multiple network round trips to establish a stream. Either way you don't have the immediacy that old TVs had.
The question of UI in modern TV is interesting. 15 years ago the 500 channel problem looked difficult, my impression was that Comcast Xfinity (2010) was the first really good STB interface for the digital age.
I have a NVIDIA Shield which has an Android TV interface that convincingly makes FAST services like Pluto, Plex and Tubi look like linear TV on an STB. What you find though is that going "back" from one of those channels can put you, disorientingly, in the app for those channels, and also that you can usually navigate better if you start out with the FAST app (and have a more consistent experience watching FAST on a computer, tablet or XBOX) Except for those things which, for some reason, are easy to find in Android TV but hard in the app.
> 15 years ago the 500 channel problem looked difficult
Knowing what was on 500 channels may have been difficult, but that's equally difficult now. The problem of navigating 500 channels was solved more than 25 (not 15) years ago by remote controls that had numpad buttons on them. You navigate to channel 351 by pressing buttons 3, 5, and 1 in sequence.
Now you have to log into Netflix, check Netflix. Then log into Hulu, check Hulu. log into prime, then check prime. Then log into Disney plus, check Disney plus. Then check live tv, then find out it’s not in live tv but the tv stations app. So you log into peacock, navigate their insane UI. Finally get to watch what you want
Modern TV interfaces are way better at navigating 1000 channels, but whether or not having 1000 channels available through one TV interface is highly debatable. It's also not a given that the current way most TVs use is anywhere close to the optimal one (especially when you have multiple devices that all have to work together because the cable provider insists on their own box, made even worse when a helpful family member installs a sound system with its own dedicated remote).
It takes ages till the channel switches. I had already more channels then I needed, so more channels add nothing to it. There is no difference between having 800 vs 1000 channels. And when the TV is slow to change them, the discovery process becomes painful enough that I just go to do something else.
The problem is with cataloging and discoverability. After a certain number of photos (or movies, songs, whatever), finding what's relevant is non-trivial.
Or the problem is what's relevant, such as Spotify's 'what the user likes and is in our bargain bin for plays'. If the photo algo wants you to get multiple prints which mostly happen when you are sending them to others it's going to push group photos and baby pictures not that epic moonrise.
Deterministic software puts the user in control of the product. Nondeterministic algos put the products in control of the user. Naturally companies want the latter and under guises of the ‘now better’ give the user worse and charge them more. A new generation isn’t even aware they’re being fleeced because they don’t have anything to compare with. And the frog boils slowly…
This entire thread reads like multiple people circling around product opportunities. There are users out there who want control and will pay for something no-frills. You might not get big VC money to build this, but you could build this.
I think of your suggested idea of objective reviews and classifications quite often, but the problem ends up being discoverability. There have been a handful of sites that were highly useful by aggregating price and consumer sentiment/recommendations from Reddit. They get initial visibility and search traffic, then get penalized by Google. Froogle/Google shopping gets priority and is the only aggregator that Google really wants to include.
Look at defunct sites like Nextag that were moderately useful in the space-- they had free and paid placements. They were steadily growing search visibility until Google started pushing their own product (Froogle, free product listings in 2014ish) and Nextag suddenly "violated Google's policies" and lost 90%+ of their traffic rather quickly (probably 1MM daily visits to under 10k basically overnight). Google shopping technically offers "free product listing placement", hidden well below the ads-- likely as a defense to anti-trust on monopolization of that specific space.
Brickseek and CamelCamelCamel are the two most successful/long lived tools in the space-- and they grew their visibility from the now defunct but once huge deal site FatWallet and SlickDeals. Walled garden subreddits frequently disallow posting of specific tools, so it makes organic growth super challenging-- given that pretty much all commercial queries start with Google or Amazon.
I was speaking more generally than reviews or classification sites. I really meant virtually any application. It's just that it's a long, slow grind to build the reputation and differentiation from the big tech products you're competing with.
Consumers can't pay. Part of the grift has been the concentration of capital in the hands of the highest wealth-holders, who all seem to have the same idea as to where to put their money (hence, uncomfortably, the bidding up of tech compensation to where it is now). So, no, you're not going to be able to Adobe/Autodesk-style price gouge. However, if you show customers the value proposition (generally some aspect of longevity or portability that prevents expensive upgrades or lock-in), they'll pay when they can. And until sanity returns to the economy, partnering with a credit facility of some sort might be wise (for you, not the customer).
Okay, no, we don't have to go down this nobody-can-afford-anything rabbit hole.
There are people out there that will pay. The issue is not that there isn't willing consumers to pay for it, it's that you're undiscoverable until word-of-mouth builds around your product among people(like those in this very thread) know and talk about your product going against the grain.
Reminds of Windows and search. I am not anymore even sure if there is some nice dialog where I could put some pattern and folder or list of folder and have it run through it for me...
It's just enshitification. It's pretty funny when I search for putty and get the help doc, or my email client and get some windows setting option. I switched to powertoys search to get around it but wonder if anyone responsible for maintaining this feature actually uses it.
Here is what I mean. Photos apps used to let you search through your photos using filters.
The same kinds of things are happening on the web which already happened to apps (desktop and mobile).
In the modern world, some marketing company wants to tell YOU which of YOUR photos you wanted, so they can sell you some prints, harvest your data, or something.
I would like any apps that have to do with collections of files, photos, music, etc to be more of a deterministic DATABASE and less of a nondeterministic algorithm.