This thread is for anyone working on personal projects to share their progress, and hold themselves somewhat accountable to a group of peers.
Post your project, your progress from last week, and what you hope to accomplish this week.
If you want to be pinged with a reminder asking about your project, let me know, and I'll harass you each week until you cancel the service
Jump in the discussion.
No email address required.
Notes -
I drive a 2010 Honda Fit and recently the driver side door handle started feeling "soft." I assume something broke due to age but the door handle and lock still work. I decided not to wait until it fails completely so if the new door handle gets here in time, this weekend I will pull apart the inner door so I can get to the handle hardware and replace it. I have access to the Chilton manuals through work and it doesn't look too impossible. Famous last words.
I would recommend using, or picking up, a pair of decent work gloves. I've done four or five car door repairs, and every single time I've either cut up my hands or gloves. Even if it's miserably hot, they'll be worth making sure it's not your fingers.
The gloves were finger/palm saver. My hands are bruised to bits but no blood was shed.
Auto work is a reminder that men are so much stronger than women. The recommendations that a particular task might take a little force really means I need to use all my strength.
More options
Context Copy link
Thank you for the recommendation. The Chilton manual also recommended gloves!
My lasting memory of a car door fix is when I was a kid I was a second pair of hands for my dad replacing a window in his Datsun Z, and when he tapped the last bit of trim into place, he accidentally shattered the just-replaced-window. The second replacement went faster.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Homies Ride or Die update
Been consumed by that rendering bug where the lighting seemed all wrong on the tires and I began to suspect my model data was fucked up.
I was using this ancient format for loading 3d models (".obj") that I downloaded from sketchfab.com but it turns out when it says this format is the native format uploaded by the artist and the other formats are derivative, it's kind of lying. A few 3d people have told me you want to get the ones in .glb or .gltf format, even if sketchfab says these were auto-converted . (I do not know how to make sense of what sketchfab says)
Anyway I thought maybe the .glb models would have more reliable model data and they do seem to. So then I got the .glb model loading mostly working (using tinygltf) but I was getting this random error where the car would render fine for the first frame and then vanish. I spent about 4 days trying to debug the bgfx state machine until I realized I was creating vertex and index buffers in the GPU and sticking handles to them in a class, and then I created a destructor for each class that would free those handles up. Fine, except I was not realizing I was copying the class around and the old class, a transient object, was getting destroyed which was deleting the memory held by the handles.
Since the handles are like, integers into a GPU resource table and not pointers, I wasn't getting any kind of pointer derefence error. It was just accidentally working for one frame and then failing to work after that.
I had to spend a long, long time staring at this stuff and remembering C++ copy/move semantics before I was able to track this down. Lesson (re)learned: don't put things like file/resource handles in object destructors unless you're extremely careful.
Finally making progress again. I'm getting a lot more texture data and the tires look a lot more realistic now, matching what's on sketchfab. Maybe I'll get spinning tires and stuff working again by next update!
More options
Context Copy link
The queries fetching the content are finally officially unretarded, which had a few fortunate downstream effects for things like fetching tweets with tweets along with their responses. Managed to move on to unretarding the import queries as well as fixing some minor bugs.
How are you doing @Southkraut?
No tinkering whatsoever this week. Family maintenance consumes all. Thanks for asking and please keep it up.
More options
Context Copy link
More options
Context Copy link
Long time lurker, first time poster on the thread. I'm currently halfway through learn python visually. I know the python parts, I'm doing the book to get a different sense of programming. My mentor wanted me to see things from a different side. So far I'd state the book is a great tool for anyone young who wants to learn programming because making things go off on screens is a fun experience.
The hardest parts are the non programming ones, it's a mild detour, knowing the math behind things is a plus for me. I'll start data structures in python after this.
Also gophercon is happening in my hometown this year in September, so if any mottizens write go or may be panning to visit, do lemme know. I'll be there for all three days.
I use ratatype for typing and my current keyboard is a terrible sub 6 dollar piece of trash where using the pinky finger to press enter or backspace would mine in sub 5 minutes. I keep missing a lot of keys on that side since I have to reset my hands anytime I use backspace or enter. Any suggestions beyond getting a split keyboard. I'll get a new slightly better mechanical keyboard soon which should ease this a bit.
More options
Context Copy link
Starting a greenfield project to build a stable software environment for ai coding. I can't think freely when operating within legacy codebases. So starting from scratch to build my ideal ai-native scaffolding.
I'm starting with a tool to auto-redact pdfs. Simple, useful, well constrained. Would appreciate suggestions on software paradigms that have worked well for ai development.
Stack:
Some ideas:
I have a basic demo ready. Codex is already raising PRs. The redacted bounding boxes are off. And the LLM redaction logic is wonky. But, so far I am impressed at the LLM's ability to build a greenfield project by itself.
I'm a serviceable software engineer. Cracked engineers of the motte, what are some software systems paradigms that you think I should play with ? I would especially like to know paradigms that make it easier for agents to understand, write & verify auto-generated code.
Sounds pretty good. I'm a noob engineer so can't offer any feedback whatsoever. My only experience with llms and pdfs was when we tried to build something that dealt with large pdfs and the biggest hurdle was the tables. I've heard pcr got better in the latter half of 2024 and I stopped my startup LARP right about that time.
Haha, yep, tables and rich extraction is pretty bad out of the box.
In this case though, I can confidently say I'm an expert on PDF extraction for llm use.
ANy tips and tricks you picked up regarding this not available out there on the web? I basically just throw the most powerful vision model at it and YOLO it.
Why not just use one of the many existing commercial solutions? That's what we did last I dealt with OCR'ing PDFs, just used Azure's API and then processed the data. Would be surprised if a raw vision model is cheaper or higher quality.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link