SomeOddCodeGuy @SomeOddCodeGuy

Very new to X. Still figuring this site out. Created in 2023, but didn't start using until Sept 2025. Maintainer of WilmerAI. Tinkerer. github.com/SomeOddCodeGuy USA Joined September 2023

Tweets

16
Followers

0
Following

44
Likes

20

SomeOddCodeGuy @SomeOddCodeGuy

a day ago

Never been a fan of agents... so I'm not sure if the way that I develop with LLMs counts as Vibe Coding or not. But man, I can move fast. someoddcodeguy.dev/im-not-sure-if… #LLMs #WilmerAI

0 0 0 0 0

SomeOddCodeGuy @SomeOddCodeGuy

3 days ago

A breakdown of the new recursive Wilmer workflow doing a form of 'deep research' on my offline Wikipedia api. someoddcodeguy.dev/recursive-work… #generativeai #wilmerai

0 0 0 6 0

SomeOddCodeGuy @SomeOddCodeGuy

3 days ago

So Ive been toying with calling WilmerAI workflows from within themselves, creating recursive workflows with specific breakout criteria. Using this, I've been able to create an offline Wikipedia deep research =D This is going to be fun to play with. #WilmerAI #generativeai

0 0 0 6 0

SomeOddCodeGuy @SomeOddCodeGuy

4 days ago

So apparently VPN use really is a perm-ban offense on @Reddit Back in July I was using NordVPN with Reddit, made a comment/edited it and got my account security locked. Reset password, got back in, but found I was shadowbanned. I appealed daily without response until in Sept I…

0 0 0 3 0

SomeOddCodeGuy @SomeOddCodeGuy

7 days ago

Stanford dumped a whole bunch of lectures on their youtube a few days ago about ML and AI. Pretty good watch if you get bored. Really awesome of them to do this. It'll take forever, but I'ma try to get through as many as I can lol youtube.com/@stanfordonlin…

0 0 0 3 0

SomeOddCodeGuy @SomeOddCodeGuy

a week ago

Don't put off unit tests. When I first started building Wilmer, I barely knew any Python, and of course I didn't have Wilmer to help me build it lol. So the early code was nothing shy of a disaster; coming from a C# background, I first structured it all wrong, then someone…

0 0 0 9 0

SomeOddCodeGuy @SomeOddCodeGuy

2 weeks ago

I'll repost my Mac LLM speed benchmarks here for anyone who is interested. Ordered from newest to oldest (2025-08) Qwen3 235b a22b, GPT-OSS-120b, GLM 4.5, and Deepseek V3.1 on the M3 Ultra someoddcodeguy.dev/mac-studio-m3-… (2025-05) Deepseek R1 0528 on M3 Ultra someoddcodeguy.dev/running-deepse……

0 0 0 16 0

SomeOddCodeGuy @SomeOddCodeGuy

2 weeks ago

I wish that LLM providers could feasibly put out a range of the same model with different levels of context window. I'd much rather have a higher degree of confidence in 32k or 64k of context than have 128k or 256k but with a lower degree of confidence. Since the models are…

0 0 0 16 0

SomeOddCodeGuy @SomeOddCodeGuy

2 weeks ago

Hmm... first post. I created this account back in 2023 and only just now setting it up. Made the mistake of using reddit while on a popular VPN service, and promptly got shadowbanned. Guess this is my new home now? =D Step 1: clean up my feed. Good lord... what in the world?