Ok, editor is added. It supports vim keybindings as well if you enable them in the settings. Just press 'e' to activate the editor.

@Ly Pro LMK if you have ideas of what to add to it. I added editing earlier today. And you can just put instructions in your setup to have it open markdown files in the editor when it creates / edits them.

As a programmer with decades of programming experience, programming before the days of the Internet (who remembers the Commodore 64?), and working at tech startups in Silicon Valley I can offer my perspective from that angle. I would say that it is more about systems thinking and engineering rather than writing code. I started off very low level with electronics and embedded systems, writing assembly language on a daily basis. I think a good way to put it is that AI allows a non-programmer to go from 0 to 1. But a programmer with AI can go from 2 to 10. A programmer is going to be able to leverage AI much more. AI enables smaller sized apps to be built. More complicated apps are still not possible without solid engineering backgrounds. And especially anything that needs to serve thousands of users you need solid engineering skills for sure. In terms of not knowing what you don't know that is certainly an issue. AI is going to give you A solution, but it might not be the best solution. A good software engineering is instantly going to say, "no, don't do it like that, do it like this". Or they are going to be able to spot the weaknesses in a plan that the AI didn't consider. Good engineers also anticipate problems and cut them off before they even happen. AI can fix the problems once they are identified. But it may be a lot of work to undo a bad design, rather than just don't build it that way in the first place and save yourself a lot more time and headache. Engineers tend to break problems down into smaller systems and then know which patterns and architectures to use (decomposition / composition)—basically applied systems thinking. AI tends to build it the most direct path possible and only concerned with completing things, long term consequences be damned. Engineers know from experience to slow down and build up the lower layers and make them solid and then build on top of that. AI is going to have a more "hold my beer" cowboy style of programming and not care about long term consequences.

All that being said, AI is the best teacher there is for a motivated learner. I'm picking up things that would normally require a college degree and tons of background. Knowledge and skills can be acquired in a fraction of the time because of AI. It can build courses. Here's a course I put together specifically for teaching me the fundamental internals of AI. Sometimes I'll ask for a topic to be added if I don't understand something or it can revise a lesson or create separate Jupyter notebooks as spin-offs when I need more help understanding something. https://courseai-two.vercel.app/app/lesson/embeddings-and-position So keep building, and if you start feeling the frictions I described above, then ask AI to solve that problem as well and teach you. Oh, and here's one for teaching music. I didn't really put much time into this one. But you get the idea. Don't just ask for a prompt when trying to learn. Build yourself a custom course for what you want and how you want to learn in the medium you want. This one has a build in sheet music visualizer and player. https://orchestral-iota.vercel.app/app/lesson/cinematic-progressions

@Carla Bosteder I don't think it was JUST a marketing ploy, I think it is BOTH. They have a history of using fear as a marketing ploy so it fits their M.O. At the same time, they also got more inference capacity. And maybe the enterprise demand isn't as high as they hoped at the prices they are charging so they are willing to at least make some of it back on subscriptions. Grok 4.5 and GPT 5.6 Sol are now serious alternatives for developers and enterprises; when Fable first came out they were the undisputed best model. Now that's debatable.

@Benjamin Chen I can't DM either. Looks Skool level related. Telegram: photoshootninja or IG: photoshoot.ninja

@Carla Bosteder No, just Fable from Anthropic. I use Fable for its taste, problem solving, and nuance. It catches a lot of things and seems to just be smarter but the cryptic communication and poorer execution prevent me from using it more. It says it's done when it's not way too much. 5.6 Is my main go to and daily driver. Grok 4.5 for quick fast tweaks or tasks. Occasionally I'll load context from and prep things for Fable with a weaker model though (Opus or Sonnet).

Are you guys finding that you need to tweak things when using a different model Fable vs Grok vs 5.6 Sol? Or when a new upgraded model release comes out? Or can you use the same setup across multiple models? I'm increasingly using multiple agents depending on the task and even experimenting with agents shelling out to other agents. So I'm wondering if and how that changes things.

I find Claude writing to memory files more annoying than helpful. It usually creates something overly specific and when I see what it saved I often tell it, no don't do that, delete that memory, that's just going to make it worse. It's typically so specific that you're unlikely to ever run into that scenario again or it doesn't understand the nuance of things. Sometimes I try to improve what it saves, but more often than not after a 5-10 minutes of trying to get it dialed in and it still can't create the right memory I just tell it to nvm, just delete the memory.