steve caldwell 👨‍🍳

@stevecaldwell

Dad. Founder/CTO @hellogumbox. We help businesses actually get ROI on their AI spend. Building and launching software products for 20 years. OSS enthusiast.

Louisville, KY

hellogumbo.com

Replies Reposts Media Articles

Pinned
steve caldwell 👨‍🍳
@stevecaldwell
Aug 3
DeepSeek v4 MTP + DSpark in llama.cpp shipped this weekend. My updated config takes DeepSeek V4 Flash from 7.5tps to 40+tps in OpenCode on my Mac Studio. local model lovers, rejoice👇
Article
Speeding up DeepSeek V4 Flash on a Mac Studio with DSpark
This weekend, DSpark speculative decoding landed in llama.cpp master, and it took my local DeepSeek V4 Flash setup from 7.5 tokens a second to 40+ on code -- same machine, same weights. Flash 0731 is...
4
steve caldwell 👨‍🍳
@stevecaldwell
6h
just got back from vacation, what did I miss?
steve caldwell 👨‍🍳
@stevecaldwell
Sep 2
yawn. I need a vacation.
Claude
@claudeai
Sep 1
We’re introducing Claude Fable 5.1 and Claude Mythos 5.1. They're the world’s most advanced models for coding and knowledge work.
00:00
steve caldwell 👨‍🍳
@stevecaldwell
Sep 1
being forced to read the work of a shitty author every day is exhausting
steve caldwell 👨‍🍳
@stevecaldwell
Aug 31
now running every meeting through Granola, Wispr Flow, and Pocket. major differences in the transcription quality, especially when my LatAm colleagues speak.
1