Giving Kortex another shot with AI features

This month I'm going to be giving Kortex a shot. I should say I'm going to give it another shot. I tried Kortex several months ago when it was still very early. It was actually even before they had added the AI components. The reason that I'm going to give Kortex a shot is because the things that they plan on adding

are very much aligned with what I'm looking for in an app, not just for writing, but for note-taking in general. And some of them are even things that I haven't seen a lot of apps put on their roadmap. So, for example, let's go through. I think there's about seven or eight of them.

AI Neural Search. I don't know that I've seen any app promise this in quite this way. It sounds like it's talking about semantic search, but what it's actually talking about is while you are writing something, it is going to look through your library for relevant ideas and show them up in real time.

to suggest things to you while you're writing. So in other words, what it suggests when you have three paragraphs will be different from what it suggests when you have five or when you have six or when you have 25 paragraphs because it's semantically understanding what you're writing about and bringing things forward that may be useful to you.

That's an amazing feature. And if they pull it off well, that is a game changer for writers. The only time I've seen anyone talk about this is there is a writer slash substacker named Michael Dean, who used to be one of the

I think he helped develop David Perel's Write of Passage. Anyhow, he had talked in a YouTube video like a year ago about trying to develop an app that does exactly this on his own, and then obviously it's been a year, it hasn't gotten anywhere, he hasn't talked about it since, so that makes this the only place I've really seen this.

other things, auto tags and connections. This is awesome. I imagine this is going to function something like the app, my mind, where it just reads semantically and adds tags for you. This is really cool because organizing your notes while it will still be possible, doesn't become something necessary.

AI Assistant and Autocompletions this is a little bit different but it is similar to the first one this is AI giving you suggestions and edits it says from a professional level editor so while you're writing it's going to give you suggestions about your writing not just suggestions of things to put in

This is another interesting one I haven't seen a second brain app or PKM app suggest this before but RSS feed integration. Add your favorite RSS feeds to your library and read and highlight without leaving the app.

So in some way, it looks like they are presenting themselves as a possible competitor to things like ReadWise Reader, Instapaper, Matter. And that's reinforced by the next thing on the list, which is Article Reader. Import any web pages reading Cortex, save, highlight, and reference it with AI.

Be really interesting because if that works well, you could theoretically replace one tool or two tools with one. You have your writing slash note app, and then you have your RSS article reading experience. Bring those together. Kind of makes sense for them to be together, right?

ImageSearch

Simple searches. So in other words, it's kind of semantically, I'm not even sure that's the right word. I think semantics only refers to actual words, but it's going to understand what is in an image. So you can ask for things like show me all blue bikes and it's going to show you all blue bikes. That's really useful.

next is full offline yeah i put this here because basically any writing app has to have a good offline mode you can't just write when you're on the internet it doesn't work that way most of the time you are wanting at least to be in locations where you may not have internet connection like a coffee shop or something

This is interesting too. They suggest that they're going to incorporate a graph view, view and navigate connections between your documents. I know a lot of people talk about the graph view and Obsidian doesn't really do anything. I disagree.

there is many ways to move through your graph and look at things the graph view gives you a nice overhead but it also gives you a focus view where you can just look at what's connected to this and see it visually and then what i think is the really interesting phase is when you expand it one beyond so you see something that's one connection beyond the note that you're on

and that really starts to give you a semantic I'm overusing that word but it really starts to give you a contextual understanding of your notes in the context of everything else that you have which can be very powerful

And then the last item I wrote down here, this is all from their website. They wrote, even further, we have plans for the future of AI agents. We will have customizable prompt flows that prepare context for you. And then it suggests some personalities.

The educator, the ghostwriter, the scheduler, the researcher. This is really interesting to me too because this is kind of what I've been doing recently with Claude. I have been creating projects and

Giving them prompts for personalities and using them for specific things. So far I've only been using it for... Oh, that's not true. Mostly I've been using it for personal things, like I have a health advisor that just gives me advice on...

more mundane health stuff you know like sleep things like uh sleep's a good example because sleep is something you want to put information in every day and have that granular focus but then also have something that can understand what all the days in a row mean together and that's not really something you can do with a human

VoiceNotes

There's one big drawback to Cortex right now. I guess it's technically two, but they're intertwined. It's a little pricey. So, for example, on the monthly plan, not the annual plan, it's $21 a month. Now, what is worth keeping in mind is you have access to a lot of different AI models.

I'm in front of it right now. Let's see. So we have Gemini Flash 2.0, GPT-04 Mini, Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Flash, Gemini 2.5 Pro, which is the premium model.

Claude 3.7, Claude 3.5, every open AI model. Two deep seek models, both of which run on American servers. So you don't have to worry about somebody scooping up your data from another country. Perplexity's most premium model, which is called Sonar Pro.

as well as their deep research model. You have both Groks available, and that's Grok 3 beta and Grok 3 mini beta. And then you have even two open source models for meta. You have Llama 4 Maverick and Llama 4 Scout, which are the tops. That's a lot of models that you have available to you, and that can be really useful when you begin to understand how each model works

is better than the others. For example, maybe you use the deep research model and perplexity to do exactly what it says, some deep research. Or maybe you use O3 from OpenAI, which is another great deep research model.

But then when you get to your writing advice, maybe you want to look to Claude because Claude is pretty good. Some people say 3.5 is better than 3.7. I haven't noticed a difference myself. They're both pretty great. I even find that OpenAI's 4.0 is pretty good at writing suggestions. And overall, it is a very good general model.

or maybe you're talking about something that is politically sensitive in your writing and you're not getting the responses that you want from some of the models because sometimes they can be a little woke.

then you can go over to Grok and get a different perspective so being able to have all those together that's really useful but here's the other problem and I should say about price it's not super out of the range of the others for example Tana is 18 a month and doesn't offer you as many models so it's not too far out but it is a little bit but if you're

able to combine a bunch of tools that you're using in different places into one place which it seems like they're trying to make that possible i've also heard that they're going to try to enable you to publish from here so this could be a blogging platform as well it might be worth it as kind of a package deal

But I don't remember if it's on the webpage or something that Dan Coe, who's one of the founders, has posted on X, but said something about cancel your other AI subscriptions because you can do it all here. And that's not true. The reason that's not true, at least right now, is because...

you have limited access to these i mean no matter what even if you pay open ar cloud you have limited access but the access here is severely limited comparatively so for example all but i think three models um you have unlimited access to gemini 2.0 flash you have unlimited access to i think it's llama 4 scout

and then I would assume maybe one of the deep seeks. I can't remember what the other ones are, but it's like four models that you have pretty much unlimited access to. The other ones, all the other ones, you're limited to 500 requests total, not for each model total. 500 requests is, you know, 500 questions, 500, you know, one word responses like no to something.

That's not very much. I would probably go through that in normal use, not just for writing, of course. But in normal use, I would probably go through that in a week. Comparatively, when you pay...

I think I'm paying $18 a month for ChatGPT and about the same for Claude. In each of those, you get about, I think it's 60 to 100 requests every three hours. So that 500 is equivalent to one day.

of a subscription with the models themselves. So that's not great. And like I said, if you're only using this for writing, maybe 500 is okay.

but you're not going to be doing a lot of research and back and forth with agents when you're limited like that. Because I mean, at a certain point, it's not useful to use the unlimited models because they're just not smart enough to be useful.

They're good for certain things, but not for that. At least not on writing. So being able to use this to replace at least my usage of Claude and ChatGPT is not feasible with this plan.

Because I use them a lot. Every time I have a question during the day about something, you know, like we had a problem the other day at work where we couldn't figure out how to turn the schedule. There was a time block schedule on our air conditioning and we couldn't figure out how to turn it off. So I took a picture of the thermostat and chatted with ChatGPT to figure out how to deal with it. Something that normally I would have spent hours searching on the internet to find.

Not really going to be able to do that within Cortex with this plan. But it does say that the plans are going to change a little bit because they are still working, obviously still working out their business model. This is still in beta. And they're going to introduce something that...

Tana and other apps have as well. Capacities has this where there's a limit hard limit within the app on how much you can use the AI but then if you go over that you can plug directly into the API and pay the model directly for extra usage which can be actually a pretty good model because

You're getting kind of a package for the beginning, and then maybe you have three months where you're not using the AI as much, so you don't spend that extra subscription for something you're not using, and then maybe you have two or three months of really intense work with an AI. Then you're going to pay extra, but you're aware of that. I think that could be a pretty good model.

And generally, I find requests and polls from the API for just like everyday normal stuff, you know, not long, deep research projects that require, you know, sometimes an hour for the model to come back with an answer because it's working so hard.

Just the normal, simple stuff is really, really cheap. It's really cheap. Especially, like I said, as the pay-to-play model. So it could work out in the end. So that's one of the reasons I'm continuing to give this a chance. And I don't know. I like the direction that they're going. I'd like to see how they develop the AI. Sorry, not the AI. The UI.

to become more of a PKM app because they are starting to offer features that put it outside of the range of just being a writing app and becoming a PKM with AI app and the current UI is good for writing but when you get a when you get

Thousands of documents in here. Yeah, it could be a problem. So I've heard suggestions that they're going to move away from some of the stuff they have here and move into a more filtered view, which is interesting. They already have a filter here. We'll see how that develops because that's a make or break.

But I would say that the way they have iterated and the way that they have gone through, they are really trying to make the best app. And I find that very impressive. So I'm giving it a shot. We'll see how it works out.