giovanniwants

I see what this site is. You can have it. FUUUUUck this. I kiss no ring in order to give you a gift.

I'm actually glad I got you as the first response. It taught me something. I can't care about stuff like this. If anyone trolls me again I'll just post the github instead of calling the troll a dick or an asshole. Why point out the obvious? :)

https://github.com/giovanniwants/Yanah/releases/tag/v0.2.1 run that through ai and see what it says.

Last thing to say. not to this dick but to the people feeling guilty about using AI for coding. AI Coding is here. It's a genie that's not going back in the bottle. If you're not using it as a tool you're obsolete. You can sit in the corner writing code by yourself for 14 hours for a basic calculator app and feel proud about it because you built it all by yourself. No one will care. AI can make that app in 10 seconds. While you're sitting there feeling superior people will be making impossible things and leaving you in the dust with your calculator app. If you want to see what's possible: Yanahchat.com

Also Zero more replies to you unless you look at the code and criticise an aspect of it, and don't roast me on non cleaned up code because I'm going for speed and functionality over a clean workspace

Still waiting for a link to your work

I've been writing chose your own adventure games in basic (if "y" then line 20) since I was 9 years old on an TRS 80 in 1988. I know code. AI makes shit sooo much faster. I never went to school because I never had support, but that doesn't make me less capable.

Still waiting for your link to your work

Qwen 125b a6b with consultation with an 80b a3b model using a recurring loop. The whole point of the project is to get off the cloud, which is noble in itself-- running on my own custom daemon I might add

Opinions are like assholes, and I'm sure you have a reputation of dismissing everything without looking at it. I'm done with you because I have actual work to do. Seriously tho, what have you done? I'm not taking an ounce of your opinion until I see your work/projects.

Not sure why "reply" is not an option on your last reply, but no, It's my work. 14 hours a day for 3 months is my work. I give zero fucks about your opinion on my methods. Again, my work stands on it's own.

my work stands on it's own. If you're dumb enough to overlook it that's on you.

Are you talking to me? I have a 80b model running on an 8gb mac with unlimited context window, ingesting 1.5mb files in 10 seconds and able to digest and talk about them with zero lag. I used AI on a lot of if but if you're saying it doesn't work as described you're a fool

The problem with this is I'm searching like crazy for relevant posts, then I see some, I post on them, but I feel like I'm inflating HN with non-important posts specifically for the point of being able to show off my thing. I feel like I'm diluting HN because I'm a person who only speaks when I have something to say. I don't like forcing myself to comment on every post that I can say something on because it feels like I'm just adding noise not signal.

Same question. I've been working 14 hours a day for 3 months on something and I'm finally ready to show it.

I get that. I am developing some high level shit and I have zero education. I feel like I'm robbing PhD programmers of their worth because Claude takes what I ask for and just does it. You still have to have a very good big picture mind, but I excel at big picture thinking and therefore I'm making some powerful shit with zero experience. It's not fair for people who went to school forever to learn code.

I use a custom Daemon. I tried OpenClaw and Letta and it was like emailing your AI. I have a 64gb Mac m1 Max and the largest model I could run in openclaw was a 27b because its not built for local AI, it's built for API. I wanted something that you could chat to, collaborate with, build, evalutate the code, make corrections, build, repeat. I couldn't find anything that didn't come with massive latency so I just made my own. Now I'm running a 125B A3B model at 30 tok/s using SSD Streaming and my agent is instant with no latency and can do everything I ask. I use Aider for programming, exa for searching.. I'm happy with it.

This is a very very smart idea.

Hey Christian. I've been building something similar for the last 3 months (yanahchat.com). I run 80B A3B Qwen3 Next and am able to get 5-9 tok/s on an 8gb mac. I originally downloaded 72 AIs, abliterated, think, instruct models and lots of others. I played with LMStudio first, then WebUI, then OpenClaw, then Letta. WebUI was decently fast but I wanted something more agentic but OpenClaw and Letta were crazy slow with local models. I have a MacBook m1 64gb and the biggest model I could get to run decently was a 24b. even then it was like emailing instead of chatting. I found LangChain/Smith and used that and found ways to make latency sub 1 sec. LangSMith/Chain was actually adding latency so I changed to a custom Daemon with an Sqlite memory system that the AI helped design by me asking the ai if you could design your own memory system what would you do. We calibrated it by tweaking it together. Then I started seeing SSD Streaming and I got excited. I immediately downloaded the 125B model and got it working well using Ollama.cpp. I set out like a madman to get this to work on an 8gb MacBook and came up with my own streaming architecture called MOE_Prefetch. That wasn't enough, so I took the largest model that could possibly work (the 80b A3B)and modified it to what other people are now calling 3bit (I thought I invented it but then found other people doing similar things), where we made the routers at 8 bit, the dense backbone and shared expert at 4 bit, then 2 bit on the rest of the experts. Claude and Antigravity have both tested the modification and show it retains 90-92 percent intelligence/ability of the full fp16. So that's how we got 5-9 tok/s on an 8gb MacBook.

Hello, I made a local AI Chat app that is as easy to operate as Spotify.

I know that doesn’t appeal to a lot of you because you’re mostly hardcore programmers, but this is for the people that aren’t as savvy and just have a normal MacBook. It’s running an 80B A3B Qwen model on any M-series Mac. Even on an 8GB Mac it gets 5-9 tok/s, so it’s not a proof of concept, it’s an actual program.

It has its own SQLite memory system. The 8GB version has fewer features than the 16GB and above, but even the 8GB will chat at faster than reading speed with zero internet, never touch a cloud, and synthesize info between (small) documents.

The 16GB or more version is actually pretty robust—you can upload as many docs as you want and it will cross-reference all the info.

I know not a ton of you will have use for it as a daily program, but I’m thinking about doctors' offices and lawyers' offices that can’t upload their files to the cloud. This would get them ChatGPT-level AI for free. Right now companies are having to spend $5000 and a monthly service fee just to get local AI, and this will do it for the price of a used MacBook Air with 8GB.

I really would like brutal feedback (about the actual operation of the program, not the idea) and to have people test it out to see if there are bugs that I'm not catching. Don't roast me on the code, there are some messes in there from failed experiments that I will clean up later. It's Mac Notarized and safe, and any info you want to know is at yanahchat.com

Also want any advice on how to get your app noticed and off the ground for any of you that have been at this stage before. Thanks! Github: https://github.com/giovanniwants/Yanah/releases/tag/v0.2.1

Hello, I made a local AI Chat app that is as easy to operate as Spotify.

I know that doesn’t appeal to a lot of you because you’re mostly hardcore programmers, but this is for the people that aren’t as savvy and just have a normal MacBook. It’s running an 80B A3B Qwen model on any M-series Mac. Even on an 8GB Mac it gets 5-9 tok/s, so it’s not a proof of concept, it’s an actual program.

It has its own SQLite memory system. The 8GB version has fewer features than the 16GB and above, but even the 8GB will chat at faster than reading speed with zero internet, never touch a cloud, and synthesize info between (small) documents.

The 16GB or more version is actually pretty robust—you can upload as many docs as you want and it will cross-reference all the info.

I know not a ton of you will have use for it as a daily program, but I’m thinking about doctors' offices and lawyers' offices that can’t upload their files to the cloud. This would get them ChatGPT-level AI for free. Right now companies are having to spend $5000 and a monthly service fee just to get local AI, and this will do it for the price of a used MacBook Air with 8GB.

I really would like brutal feedback (about the actual operation of the program, not the idea) and to have people test it out to see if there are bugs that I'm not catching. Don't roast me on the code, there are some messes in there from failed experiments that I will clean up later. It's Mac Notarized and safe, and any info you want to know is at yanahchat.com

Also want any advice on how to get your app noticed and off the ground for any of you that have been at this stage before. Thanks! Github https://github.com/giovanniwants/Yanah/releases/tag/v0.2.1