SamInTheShell

Junk. Just another company that doesn't understand why I pick up a phone or use a support chat or send an email. Before I even started in corporate, I stopped checking email. Now when I call in to get support for things, everytime I get a clanker, I'm guaranteed to talk to a human customer service person to chew them out and cancel the service.

This is pretty good to know today. I’m currently using cosine similarity in some of my code that could have just been dot products.

Just use a harness that discards all except the most recent thoughts. This model is probably the first real small model that does well on long horizon tasks.

I did. Distributed too and supports erasure coding. The reason is because nothing quite had the architecture or features I wanted in a deployment pattern that matched my use case. All major databases that are validated are complicated to operate. I built something that operates one way for a non business use. It’s great, doesn’t need to serve a million people, just about 5 or so. Doesn’t need to be the most optimized thing ever.

I’m opting out of any corporate targeted code for personal use, anything not designed for user experience gets scrapped when I’m the user and I have Software on Demand as a Service.

It's worth running, even quantized. I liked Meta Muse Glimmer's outputs, but qwen3.8-27b@q4_k_s kinda seems way better. Haiku/Sonnet kinda pairing in workflows?

The proxy caches and package solution for Go has had its own drama.

My big contention with this project is that if I wanted to use node (or npm; or vite) in any capacity, I would have chosen node for that. I don’t want to use those tools in this context because it serves me worse than a well thought out solution.

The thought out solution is to isolate your frontend into its own directory. You can embed and serve that as part of your Go server. That frontend could have been done with react, vite, jsx, whatever typescript nonsense you want; without the Go parts necessarily needing npm or node in the build process.

I literally have toy projects illustrating the idea (this one is a toy I had AI make with some of my own patterns months ago): https://github.com/SamInTheShell/social/tree/main/frontend

Projects like grafana do stuff like this already. It’s a known good pattern.

I'm kinda in the camp of wanting nothing to do with node if I'm building in Go. We have our own stdlibs for serving.

You go to person's profile. View their submissions. Use find via CTRL+F to seek out their "Show HN" submissions (or in this case, it's the top of the list because it's my last submission https://news.ycombinator.com/item?id=48970916 )

Done. You have 2/3rds of Discord or Slack solved as just an insignificant fraction of my Private Cloud Project I just shared. Not interested in video or voice right now, it's easy to solve though. Also you'd have to solve for the mobile apps.

Edit: If you keep checking back, I'll have a video showing off that application. It's a bit beefy to just get up and running.

Sure, I'll make a Show HN post this weekend.

At least code isn't a moat anymore. Have a weekend and want your own Discord? Doable.

How much does it actually matter? I use to bother with these settings, but never really seems to fix the problems I commonly see. The issues I commonly see are small models getting stuck in a repeating loop, or they just fail at follow up calls for editing.

Build a distributed system using raft groups. You will see Opus fail.

Opus can make harnesses fine. I barely have to instruct it.

This is probably the first small model I got through some simple web game tests without having to reset the context. It tends to opt to overwrite an entire file instead of doing edits... which editing is where most of these small models fall apart along with getting stuck in repeating loops. Only 24k tokens in so far, it did some decent newbie work.

Tool calling at Q8 for this model is kinda impressive. It's doing patch edits like a champ. It gets stuck looping and doing the wrong things though as the context window grows.

Related/Dupe: https://news.ycombinator.com/item?id=48494697

Also, I had Claude Opus read through some of the mailing list threads and it made this script to check a few things: https://gist.github.com/SamInTheShell/c3609743d742f002f1e8c7...

Lazy check script by Claude Opus: https://gist.github.com/SamInTheShell/c3609743d742f002f1e8c7...

It seems my stubborn resistance to using yay saved my bacon here.

Good luck everybody else.

Scratching my head. Wondering why I would reach for this over just running a Yugabyte cluster.

It’s just stuff I play with. My main reason in publishing this is to give people that otherwise lack the knowledge can have somewhere to start.

This should be safer than relying on seatbelt on Mac OS. Docker and Podman being in a VM is pretty much peak security imo for AI.

Showed me teal but only have options for blue and green. Teal ain't either.

Edited: I saw the multiplier table changes for annual subscribers. Not happy about it and planning to demand a pro-rated refund.

I don’t even read the conversational garbage. Only care about the outputs I demanded it give me.

I actually changed this tonight. Most of my feedback so far has been this issue, so I just changed it so the ship fires where the mouse is, movement is still WASD.

This was a small side project I worked on this last week. All the code is AI generated.

The actual link from the initial reddit post without the GOOG tracking: https://www.congress.gov/bill/119th-congress/senate-bill/136...

Edit: Also does this mean OpenAI can bring back the Sky voice?

Yeah, I plan to do a follow up comment with data and results.

Thought about this during the morning. I'm run the posts through ministral3:3b, mistral-small3.2:24b, and gpt-oss:20b this weekend to build a sentiment mapping and see what I get. I'm optimistic about ministral3:3b, but the other two are pretty good at this type of stuff.

To me this seems to be lacking in usefulness. Maybe if I was hard "no AI", it might look useful, but how do you trust people anymore? You can't really follow the "trust but verify" philosophy here.

The only way to really have a GenAI free codebase today is simply by not accepting any external contributions.