An Untitled Microblog

Thanks ChatGPT, I wouldn't want…

Nate — Thu, 18 Sep 2025 02:03:24 GMT

Thanks ChatGPT, I wouldn't want to offend a fictional metal door.

#ai #chatgpt

Toyed around with Gemini's advanced…

Nate — Sun, 14 Sep 2025 01:34:04 GMT

Toyed around with Gemini's advanced photo mode, seems to do pretty good with replacement and generative fill.

Combined two of my photos:

Into the following:

Seems to have worked real well. A bit low resolution, but still pretty good (the input images I uploaded have been heavily compressed, their originals were higher res).

However, it doesn't seem to like to do basic edits. I gave it the unedited original (at least the JPEG, it won't touch RAWs) of the milky way version and told it to do similar edits I did to the one I shared (dehaze, saturation, contrast) - and instead of editing it is just replaced the milky way portion with an AI generated image of the milky way that was fairly similar but obviously a new image.

#photography #ai

I've decided to jump on…

Nate — Tue, 08 Apr 2025 10:29:07 GMT

I've decided to jump on the bandwagon and try vibe coding, if anybody's interested in checking out the results I made a Discord/IRC inspired Nostr web client.

It's a single static html file and everything runs everything locally in the browser, so it's pretty flexible. Currently got a version running on Github pages:

https://0n4t3.github.io/n_cord/

I've also got a git repo:

https://github.com/0n4t3/n_cord/

For my very poor development skills I think it turned out pretty good, even if it's probably due for a bit more polish. At the very least I'll probably use it to keep track of tags I want to follow without needing to spam my timeline with them. It'll let you follow tags without an account and/or login with a Nostr browser extension.

#vibecoding #ai #nostr

Meta Engineers See Vindication in DeepSeek’…

Nate — Fri, 31 Jan 2025 02:47:30 GMT

Meta Engineers See Vindication in DeepSeek’s Apparent Breakthrough

https://archive.ph/cqpJP

#news #ai #deepseek

Finished a new blog post, basically…

Nate — Tue, 15 Oct 2024 10:22:27 GMT

https://nate.mecca1.net/posts/2024-10-15_non-generative-llm-uses/

Finished a new blog post, basically just a verbose version of my notes that accumulated while experimenting with local LLMs and similar AI tools for translation, proofreading, summarizing, and transcribing content locally.

#ai #blog #blogstr

Hey #nostr can I waste your…

Nate — Fri, 09 Aug 2024 01:34:15 GMT

Hey #nostr can I waste your time for a sec? It's a mildly amusing waste of time. Anyway, I've been playing around with Suno, an AI bot that can make songs, and also been working on a revised version of a blog post on microblogging protocols since the old one had a few things wrong and was getting out of date. Wound up writing some random lyrics and told it to make me a sea shanty, so without further ado I bring you the "Silly Goose Protocol Song"

Had a couple of points where it got hung up on pronunciation or messed up the lyrics a tad, but it turned out surprisingly well. Thought it was funny and didn't want to wait until the updated blog post to share.

#ai #music

Local LLMs and AI Ethics (mine…

Nate — Tue, 26 Mar 2024 12:15:26 GMT

Local LLMs and AI Ethics (mine makes nukes)

https://nate.mecca1.net/posts/2024-03-26_ai/

A #blog post by me about #AI. It's split into two parts, the first part is about my experimentation running entirely #foss models locally. The second part is more of an essay/opinion piece walking through the conclusions I've reached on AI and it's ethics as well as predictions for the future.