An "explorative" hex editor where you can do "fuzzy" searches, e.g., searching for a header with specific values for certain fields. (I thought ImHex should be able to do this (and still think it might), but haven't really figured out a good work flow...)
Love this one! My immediate thought for a build: just a camera with on-device VLMs and LLMs. You could point it at a normal whiteboard (or a wall of sticky notes), and the model could interpret the handwriting, track the tokens, and sync the state digitally in real-time without needing any special "smart" hardware.
It feels like a much more flexible approach than an expensive proprietary whiteboard. Would a camera-only setup cover your use case?
Yeah, I figured a camera oriented solution would probably be the simplest way to actually implement, especially for copying the writing over.
A loooong time ago I had one of those notebooks (rocketbook I think) that you would take a picture of and the software would detect it and upload a digital version of it, it was clunky though. But the thing I liked about it was the little icons you would color in that would denote what kind of note to save the page as.
One thing I want: something like a function call graph: you load some source code files, open a function in window A (windows A now shows Only this function with all functions calls and structs highlighted). Click on any function call then the callee shows up in a new window with proper highlighting. Click on a struct object it immediately shows the struct def in a panel stick to the bottom of the screen.
Essentially, it’s like VSCode Peek definition but with a different visual style, and similar to the same functionality of source insight but free and in Linux.
The purpose is to read xv6 source code. I have found some parts of the kernel, e.g. the file system to be convoluted, and I need to follow quite a few jumps to form the system mindset. Having such a small tool is very helpful.
Actually it doesn’t look too hard to implement. Maybe I’ll write it myself in QT.
Although it would be manual, that’s very doable in Emacs and Vim. It’s the very reason I use these editors as you can have a screen fulls of opened files instead of the tabbed model of common editors. One example having all the files you need to edit when you’re implementing a feature in a multilayered architecture. Easier that switching tabs again and again.
I experimented with this idea a couple years ago, using the Language Server Protocol to make it some what universal. The output gets messy pretty quickly, maybe these days an LLM could be used to only show the interesting calls.
Planners. Something to help me make sense of what my priorities are during the day. Something to help me determine what needs my attention.
I have a lot of systems, but I could really use a Jarvis at work right now and it seems like the whole damn promise of AI is to deliver that. I’m waiting.
There's something about having so many planners online/digital that sucks.
I want to plan, but I don't want to be on the computer yet. But if I plan non-digitally, it's slower, I can't copy/paste, and then I don't really have good access to it unless I have it with me.
But if I plan too much digitally, it's stored in digital memory, not my memory.
I'm really struggling to find something to mitigate this. I wish I had a tactile miro board that also created the miro board online.
An LLM tool that can sit on a CI pipeline to propose what tests should be blocking.
Instead of brute-force method of selecting the appropriate test suites by path or similar, have LLM analyze changes and propose the set of test suites that is relevant to the change.
If there are new complex tests added to the change, estimates how many times to run them to ensure they are not flaky to begin with (hundreds? or thousands?).
I hate DevOps. I have to do multiple commits to implement something.
I would love to be able to have access to the same env as the CI so that I could prototype the script/job on my own machine before committing to git. Most things are using Docker anyway, so it should be possible.
I hate that I need to write commands in Yaml files, commit (or use the browser) and then look at the result.
Perhaps the ability ability to stop at a specific point in the script and being able to modify any commands and execute the step and then continue the script until it fails again.
You know .. debugging interface would be a killer feature that would save so much time developing scripts.
reply