Dad, husband, immigrant, nerd. Here for truth, all opinions are my own.
San Francisco, CA
Joined April 2026
- Tweets2K
- Following325
- Followers255
- Likes1.2K
This is what you get when you ditch your legacy engine.
LFG
Bruh !
🔌 Turned on the 2nd network cable between the Sparks and TensorFold got even faster:
⚡️ 93.6 tok/s for 1 user (vLLM: 48.5), 1.93x
👥 213.8 tok/s with 6 users (+29%)
📖 Long prompt reading now even with vLLM
github.com/tonyd2wild/DeepSe…
This is what happens when you forget to use a sub and auto-fallback to API pricing. Opus 5.5 on Nous.
1/12th of a Spark in 4 hours.
What do you think happens after public investor pressure to subsidies? What happens to compute pricing then?
ngmi if your tech team is not running >1Tb fast vram locally. M5U for sweet C1 and CUDA cluster for monster C64
This machine is generating around 50B tokens a month for me, either in compute hours (quantising/etc.) or in DS/GLM inference.
This is the cost of the true frontier at better than cloud speeds and incredible concurrency.
My cost for this was way less, thanks to the generosity of a dear friend.
It's expensive, but a startup that burns 5 figures or more a month on inference should setup one up.
Nope, not behind.
Just had Mid-Autumn break, travelled to see family, ate mooncake around the lake in Hutong, recharged.
Back to cooking. LFG!!
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
YES, I knew it, they pulled off another 0731.
Wonder who would have 2xSpark and 2x6k recipe first.
@MiaAI_lab @WescheNex1q @CardilloSamuel 🙌
So pumped for this!!! Going to have to run so many things that failed before through it again!
Thanks to @ItsmeAjayKV for always having his ear to the ground on all the local AI news!
Narrative violation. Tech jobs have tripped in 2026.
US tech jobs, by the numbers:
• Active tech job listings have tripled in 2026
• Bay Area: 10% of US tech roles, paying 37% more
• 38% of tech workers are fully remote; 32% of new roles are
• Most roles still looking for programming experience
applystead.com
