Agree.
At the moment, the AnyType is missing many essential features, the editing experience is suboptimal, and the UI requires further refinement.
I’m late back to the party. Had to log for a while ‘cause life got in the way.
Yet another facet of this. Aside from the impact AI has on resources in the areas where data centres are established, we’re 100% being convinced to shoot ourselves in the foot.
To echo Bernie Sanders, AI has to be made to serve us. It has to be made to be factual regardless of whether that increases its usage levels or not. Strict adherence to factual information returns. Too few people actually verify the sources AI provides. When we’re out here still unable to create programs that effectively fix grammar errors for relatively low cost, we’re barging ahead way too fast on AI.
And its destructive impact on human creatives is incalculable. A few tech bro billionaires haven’t yet been made to understand their place in human creativity, and we absolutely need to push back on that.
I’m glad this subject brought out so much information about AI to the community.
Even if AnyType can be made to interface with AI somehow, it should never ship as a default part of the program. If someone wants AI to use they have access to it myriad other ways on their machine already.
Just discovered I can no longer successfully create links to external files and have AnyType open them in the program they use for me. Very miffed about that.
We’re a long way off from AI being a priority. It’s not a question of if any existing AI bubble is going to burst it’s a question of when. Once AI is much better governed and regulated, then we look at it here. Anything earlier is unhelpful to AnyType’s aims.
Yeah. I tried to access a new recipe from my computer on my phone the other day and the page wouldn’t update to the newest version and knock something off the back end in my alotted cloud storage. That needs to be happening. For sure.
Polishing what already doesn’t work should be well ahead of bell/whistle like AI.
I’ll have to hard disagree on this, Miguel. I’ll be using AnyType a decade down the road whether it has AI or not.
AI is non-essential. My brain is actually willing and able to do the work. We don’t need AI to automatically format text. Building in a function that recognizes the pattern of a page’s format and starts doing it for you automatically exists in Microsoft Word from before AI existed in any form.
A more advanced version of that is what you’re asking for.
Categorizing and tagging, well, that’s akin to an advanced sort and still doesn’t require generative AI.
As for summarizing information, hard pass. I don’t need AI to digest everything I’ve written into my space and spit back out ultra processed SPAM-esque information that I already have on hand because I TYPED IT IN MYSELF in the form I want to digest it in when referring back to it later.
I won’t deny it’s possible to enhance some uses, but with old features breaking on new updates and some basic quality of life improvements absent AI is NOT going to sell the program to more users. I’m so sick of seeing AI on every web browser I use I’m actively finding ways to strip AI out of my machine.
Unless it can start spitting out indisputably factual information about a topic I might be researching, which it cannot, it doesn’t do anything to enhance AnyType’s offerings.
Perfect and practical counter argument. Thank you. <3
AnyType’s streamlined way of managing objects, and organizing data reminded me of my OO Programming classes where creating single classes that were as simple as possible while offering maximum functionality… I remember thinking that the AT team really paid attention in those classes and had some teachers very worthy of being looked up to. That they had some real talent for making programming an art.
Taking a similar streamlined approach to AT’s development, applying coding principles to how they prioritize features would 100% be a boon to the team. But people and their hopes and imaginations are 1000% harder to corral than inanimate computer code.
This is one facet where taking on a mental form asceticism regarding what they try to do with the program could greatly help rather than hurt them. If one day AnyType is the programming epitome of ascetic decorative aesthetics, whose available functions worked flawlessly, that feels like the best version of the program they could possibly aspire to.
There MAY come a time when it becomes clear AI is critical for the program. We’re observably not there yet. When Notion added AI to its product, there wasn’t a single use I had for it. And what few queries I did put to it, came back with nothing useful. It’s not as essential to this kind of program as the team may be hyping themselves up to think it is.
I’ll put a counter argument on this, when someone goes through post-secondary education in the arts students are very often encouraged to find a style that comes from them and looks unique compared to most others. Something that, if it was practiced would make their work identifiable.
A human touch unique to that human being. Yes, this is a thing. I can look at an oil painting by a fantasy artist and know its theirs by the look of the piece. If two artists collaborate on such a piece, I can tell which parts of the painting were crafted by who.
Generative AI is absolutely a threat to that. That’s something I’ll defend to my dying breath.
I don’t think it’ll take nearly that long to get an understanding of what AI can do for/to us.
There are already people committing suicide because AI makes them think they should. Too many do not have the psychological defenses in place for this stuff. And the kind of AI you can have a conversation with is nothing like what a program like AnyType could benefit from.
If the team added someone on a completely volunteer basis to painstakingly tease out AI implementation and it didn’t affect core development, I can see that. But the core team needs to remain focused on the core program.
I think the amount of discussion that has happened since the post first went up shows that I am open to back and forth on the subject.
While I started from a position that came on too strong for the likes of some, calling firm opposition to AI a religious stance was unwise at best and a very poor choice of words at worst. I’m obviously no luddite. I have a computer station with 6 screens I spend most of my day on my computer whenever I can for various purposes. This however, is one aspect of information technology that has CLEAR and PRESENT dangers.
Not the least of which is compromising the proof of origin for someone using the program for creative writing and academic writing. Hopefully I’ll find some time to keep the thread going.
I only have personal experience to go on here, so if the usefulness of AI in a program like this can be judged by my experience with it in Notion, I can unequivocally say I never used it. And the few times I did, what it fed back to me was less useful than the information I had filed away in the space.
And Notion’s devs were absolutely DRAGGED for including AI before basic quality of life fixes that had persisted for 3-4 years. Rightfully so.
I’m hoping this team will learn from Notion’s example and understand that AnyType for many users who migrated away from Notion for reasons like wasting time on AI were hoping for a less infuriating experience here. My initial experience of AT has very much been that. Things should work just DO.
I truly hope the dev team might take some time to get a full picture of why folks are migrating away from other platforms and come to the conclusion that a lack of AI is one of the reasons some folks are actually here. A development process that prioritizes the right stuff and fixes bugs. That seems to be the overwhelming sentiment of migrants to the program. At least to my eyes.
It wouldn’t surprise me if some users may very well be here pushing AI in bad faith. As a way to sabotage their development focus. Software development can be as cutthroat as any other business.
I can only hope their gut assessment of the community takes us down a no-integrated-AI path until a much higher percentage of bugs are put to bed.
The only “bug” for me, right now, is how poor the search is. Adding “AI” or vectorization + vector search, would help immensely.
For better or worse my skin is tougher than that. AT is too useful as is without AI for me to walk away. A reply like that is the lowest point in my interactions with the team, thus far.
Otherwise, they’ve been damned good. This is something I’m willing to give a little grace for an imperfect “human moment” for.
Unless I’m wrong AnyType, to a variable (by variable I mean object variable/property) level, searches ALL the content of a space for the exact word you search for in everything in the space no matter how long a single page may be.
I’m almost 100% certain improving AT’s searchability can be accomplished without AI. I actually ran a few tests for what AT can and cannot find. It searches every letter and word in your space for the content you’re looking for. It’s search is actually BETTER than Notion. Which is one of the reasons I’m here.
At one point Notion limited its searches tot he first 500 words of a page. I was using it to record plot ideas with thousands of words and images on a single page. That was quite literally meant I lost the ability to find an incredible amount of plot information later on. Essentially destroying the data. They might as well have deleted it. I absolutely flipped my shit when I discovered Notion had imposed that limitation. I haven’t seen anything that would suggest AT does anything other than search a space word for word. That’s something they can be proud of.
I can’t remember what the use case(s) was where their search failed, I know there was at least one I found and I outlined the conditions for that failure in another post. If I ever find it again, I’ll link it here.
If people were helped to understand the mechanics or limitations of their search function, I think they might find they can use it more effectively, but it’s true that it would be nice if the learning curve was much more user friendly.
Exactly that is the problem. I don’t always remember how I named things. And semantic search can find what I’m looking for even when I don’t know how EXACTLY I have written it a while ago.
I don’t give a f*** about Notion nor about comparisons with it, to be honest. I just need a tool that helps me do the job.
So the idea here wouldn’t be to include a generative AI, but one that can digest and understand context of what it’s looking at.
That’s something the team can keep in mind perhaps. No generative AI, but one that’s entirely sandboxed on your machine that can digest content to return better search results.
At the very least, I find that less offensive than including generative AI. I’ll admit some details here are a bit outside my understanding of AI, so I’ll have to read a bit more into it. Thanks for coming back with the clarification. 
And if you want to see if AI can make that difference you might want to give Notion a test run. See if it can deliver what you’re looking for. Notion did finally offer an offline mode that repatriates user data so less risky than it used to be.
That’s just my use case. Vectorization or tokenization is part of the AI field. Probably the very first step.
Other people have listed different use cases, like summarizing, interpreting, etc. depending on what they use AT for. I could also agree that it would be nice to have an interface where I can ask a question and the “AI” would answer based on the stuff I have in my knowledge base (which is my primary use case for AT). Sometimes I accumulate a lot of material (like in the last week, while setting up a VPS with various servers) and there are overlapping sections and all kinds of chaos as the material grows and branches out into separate articles. I don’t have time to comb it while I’m working on something and, after some time, I don’t even know if I’ve covered everything or not. Of course, a session of restructuring the new material would help but there I’d need better navigation tools, i.e. block links, collapsing sections / easier overview, semantic search (that identifies similar concepts), etc. I think AT have lots of space to expand in that direction. And then, once I have all the instructions on how to set up NextCloud, for example, I’d love to be able to just ask “what do I need to do to enable connecting to mailserver in another pod on the same host” instead of reading through bunch of text. Although, admittedly, with right navigational tools, the correct section should not be hard to find.
Note that LLMs can also link concepts together (probably exactly due to vectorization) and find that you need to enable certain NextCloud setting, but also turn on on Network=pasta for the mailserver pod (which is written in another article and is related to mail server and pods). I think this would be extremely useful as the size of collected material grows.
And, likely, I’m pretty sure there are dozens of other interesting use cases that we are only about to discover. With visual models, one could find similar images, or images with the same person, or a sign on them. Or sound files, and so on, and so on.
And right there is one of the features I’ve been screaming for, for ages: block links. They’re essential to a program like this. And like you say, it would go part of the way alleviating part of the appeal of AI, to be able to effectively build our own web of logical references that takes us directly to where we need to be on a page.
In a minute I’m gonna post a link inside this reply to any requests for that feature here so someone who wants that feature and sees this can go give it an updoot on its own post.
I would like AI features however…
- I can wait, would rather have other features before AI
- AI should be optional and opt in
- Option to use private self hosted AI models such as Ollama LLMs
How is a web app like Lumo more secure than a local first AI assistant? I feel like the image one of the Anytype devs shared on product hunt sounds really promising, and I hope that the team keeps pursuing this idea.
I think that the big AI companies are far too nonchalant about the broad implications of this technology, user privacy, job loss risks, AI singularity, yadayadayada…
…buuuut a local first AI assistant that helps you synthesize your own knowledge that you yourself created? That’s about as harmless and safe as this technology gets.
The way this post is written just sounds like a vitriol filled ad for Lumo, an app that I have never heard of in my life. Please just like, stop and think about what you’re posting for a second. I could probably hear that piece of advice sometimes as well.
We literally have an API, MCP server, and headless mode… You have everything you need to integrate AI right now… I’m doing it myself…
Now can I finally press backspace and have it work correctly?!… No… OK then…