Listening to this conversation with the creator of OpenClaw has been a wakeup call regarding the “future” of apps or how to think about building on and making use of what Autonomi offers in that regard.
I think it will take some time to digest but it is clear to me that a shift in thinking is required today, not tomorrow.
Prior to listening to this I already felt that many one size fits all apps will largely become obsolete and replaced by user specific apps easily created by a few prompts.
Why would I use an app that mostly does what I want when exactly what I need is a prompt away, we may not even need to know what app your agent creates to complete a task.
Soon enough that will the norm.
Begs the question as to what is worth spending time on building and what is not.
There are so many problems with this but I will just stand by and watch.
It’s like Musk’s latest nonsense which I won’t repeat: exciting, attractive, scary and full of flaws.
Everything is starting to smell bad but few here notice.
If anyone wants to understand my position, please don’t ask me. The best way to learn is to get off this train and listen to people who aren’t invested in this stuff.
And the very next post I read on Mastodon is…
Oh I understand your position mostly I think and probably share concerns.
But, I don’t think it changes the path.
OpenAI, Meta, Anthropic are all on this guy wanting in.
To qoute Lyn Alden out of context, Nothing stops this train.
If you plan to use this as an app that has access to your system, your data and delegate privileges to it, I don’t think you understand my position. And that seems to be what you envisage in your post.
There’s a big difference between dealing with an app built by a human, and the well known processes that we’re used to following as we do this, and delegating the while shooting match to an algorithm.
It’s like the self driving cars promised a decade ago and which arrive only to be revealed as a mechanical Turk sham, cover for further extraction of wealth by delegating the hard part to actual humans under ever worse work and employment conditions.
It would take me a few days to explain the myriad of problems I see, which is one reason I haven’t shared anything like all of them here. Most are things that have occurred to me, and are subsequently confirmed or highlighted by others, and some are things I hadn’t considered.
Trusting this to algorithms and those who own them are just a couple of the issues here and I doubt many have a clue about where this is really heading.
Hmm, not exactly.
I am very curious and I think anyone interested in technology should be.
I am also shit scared to do so.
For now I am watching from a distance and considering how to try it seperate from being on “my system” definitely not doing that.
The post is only suggesting that we stay informed and aware because while your concerns are valid, the pace of improvement will result in widespread use in short time.
you wanna guess who just today installed openclaw on a raspberry pi (made it destroy itself by accident 
and) is testing what it accomplishes and where it just fails with different models …
for now I failed making it using a browser properly - and I’m failing to connect it to a local LLM server … but much fun - if our telegram wasn’t dead i’d really love to throw in an agent (maybe not the clawbot because that has too much access to something running in my network xD) into the channel that answers questions about current state of the network and maybe even the latest codebase …
May I suggest giving it your bank details and a shopping list?
Let me know how it goes 
That obvs is meant to be a joke but I also see a smart fridge keeping itself stocked in the not too distant future.
I’m about to make it install openCode and will try to make one agent delegate coding task to the other agent xD
will be fun to watch how that ends and if it blows up the pi or manages to solve simple tasks
feels like a bad idea but could be fun to watch
What model are you having it use? I guess that has impact on how successful it is in achieving tasks.
in this moment z.ai/glm-4.7 … might give it access to opus within the next days to see what happens then
I have a stack of sbc’s sitting idle. Maybe I will build a poor mans mac mini cluster/swarm to test it out
Actually, I will install it on one link the rest and see if it can work it out by itself 
Instead of millions of unmaintained, insecure apps being made, now we will have billions of unmaintained, insecure apps 
#thefuture
There are definitely pro’s and con’s.
For instance.
I use an app on my phone to scan documents for work, I pay to get the ad free version. Have paid for many years.
To make sure I don’t lose the documents the app stores copies on its servers.
It is really basic, take a photo, convert to pdf and then I attach the pdf to an email and off it goes.
None of it is required to be secure but I pay monthly because it is convenient.
Today, a couple of promts can create a version of that app for me that I no longer need to pay for.
Do I care if it is insecure? Absolutely not.
How many situations like this exist in the world?
Probable the company that I pay for this app is also data mining.
Which app is better for me?
I haven’t watched the video yet, but I don’t think that is unreasonable. Being able to tailor simple bits of personal software as a non-software engineer hobbyist is already with us. Folks are vibe coding stuff that would have been impossible just a few months ago. Things are moving fast.
Proof point, I knocked this together in an hour or two: AntTP Monitor - Track your upload commands in AntTP
Now, clearly folks need to articulate what they want clearly or it will be garbage in/out. However, clear specifications, using an iterative approach, can yield good results.
Moreover, asking for something like something that already exists is a decent strategy. So many tools have been made over and over again. The AIs know how to reproduce much of that as a result. Creating tailored versions of all sorts of software is now possible and not just by software engineers. Sure, you may not always want to share/sell such software, but for personal use, why not?
The end of 2025 changed the game with LLMs, IMO. Now 2026 is really going through the gears. We can’t turn back the clock and the future is a brave new world in software development. There will be challenges for sure, but I’m expecting big changes in the way software teams operate this year. Exciting times!
Nail on the head.
Different kettle of fish if you are dowloading vibe coded apps.
My guess is downloading simple apps (many of which may well be less secure) is going the way of blockbuster.
The one you have reason to be confident, is doing what you want it to do.
If that’s security, that, if something else, that etc.
If your programming language is English 
, and if you’re not an expert in each layer of this tech who understands what makes good and poor software and in what ways 
Surely you are not suggesting that every app on the play store is secure and built by experts.
If you can’t read the code of the calculator app you download or the code of the calculator app that Claude wrote for you.
Are you saying choose the one from google play every time?
I’d choose the app I trust more, which is not going to be one that I’ve created using English instructions to a black box and has only ever been tested by me.
If I created an app using tools (compilers etc) that are known to be reliable, consistent etc, because they are used by tens of thousands of other real programmers daily, I’d also be inclined to trust that. Because I have a good understanding of every layer and experience that shows me it works.
The first thing I do when considering an app from Google Play is look through the reviews. But then I rarely use such apps because I do not trust developers I know nothing about.
I also cannot trust tools that are proven to be stochastic and unpredictable and have who knows what embedded in them, because they are impenetrable. If people could control the training data that would help, but it still leaves them with fundamental problems that mean I never trust their output.
I see this daily in the output from web search. Mistakes are frequent and often serious.
So I’m always careful when obtaining software. I look for a number of signals first.
Most people I suspect don’t do this because they either trust blindly, don’t know any alternative or don’t think it’s worth the effort.
I think it is, and not just with software. I’m currently engaged in a project that requires me to select and employ people I never dealt with before, doing things I’ve never needed done before and know little about. To start I thought WTF, but so far so good and I’ve got a reasonable process and various ways of doing my due diligence. In fact, it is just problem solving in a different set of domains and I love that. I’m good at it, enjoy it, and love learning, so it’s fun as well as scary.
I approach everything new with a mix of excitement, skepticism, analysis and self reflection. I’m not against new tech without reason, nor Musk or anyone else. I just see things differently from many and have an understanding of why. But one thing I’m not is alone in this approach or the views that result.
One of the consequences of AI generated things - not just apps - is that doing the above becomes harder. Everything becomes untrustworthy because so much of it is now created with little effort, care, cost or consequence to those spewng it out.
I’ll never pay seriously to support that. I never bought anything from Murdoch, I shun chains, big tech and any entity I see as toxic. Instead I do without or find other sources and other ways.
I’m a discerning node. Are you!?
Totally agree obviously, but perhaps I need to clarify when I say “you or I” I am speaking in general.
Therefore:
Is where I still think that for simple apps there is an argument, custom apps may be less intrusive?
At the very least they wont be serving “you” constant ads and mostly likely tracking or who knows what.
I can’t argue your points, but people with the ability to take your view are a minority.
Lets say in the not too distant future a user can enter a prompt on their phone saying I need an app that will do x,y or z and a few minutes later it appears.
Ads free, spyware free, I am of the opinion depending on what the app is, often trivial, it is perhaps better than what you find on the play store made by someone wanting to earn something from it.