Kanzie
christiankanzie.bsky.social
Kanzie
@christiankanzie.bsky.social
Reposted by Kanzie
What if we were valued for the good we have done and not the pain we have caused
March 5, 2025 at 11:13 PM
Reposted by Kanzie
"The Fuhrer is himself the bearer of the collective will of the people. In his will the will of the people is realized.” - Ernst Rudolf Huber, Constitutional Law of the Third Reich
send help, i'm in danger of becoming a "it's a republic not a democracy" guy
February 19, 2025 at 6:50 PM
Does anyone know why USSR is on the list, and where exactly do they plan to send them? Russia is also in the list so that’s not what is intended.

static.foxnews.com/foxnews.com/...
static.foxnews.com
February 4, 2025 at 7:49 AM
Let’s for a moment talk about something else than the (in context) nothing-burger DeepSeek-R1. How the active and retired team behind voyager-1 managed to remotely debug AND reprogram it to come back online and continue its work, 47 years and still going strong!

www.youtube.com/live/wpA8NBz...
Fixing Voyager: How NASA Restored Communications with Voyager 1 from Across the Solar System
YouTube video by NASA Jet Propulsion Laboratory
www.youtube.com
February 1, 2025 at 6:33 PM
And then this: xyzlabs.substack.com/p/berkeley-r...

I think we need to stop training to beat tests and start training on open ended reasoning tbh
Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Model RL Revolution
A Berkeley AI Research team led by PhD candidate Jiayi Pan has achieved what many thought impossible: reproducing DeepSeek R1-Zero's key technologies for less than the cost of a dinner for two.
xyzlabs.substack.com
February 1, 2025 at 10:41 AM
As impressive as the DeepSeek accomplishment is, the attention still is on the cost of training. As this deep dive points out, the cost is not accumulative. This means we are NOT comparing apples here. Last training run <> total training. www.thoughtworks.com/insights/blo...
Demystifying DeepSeek
We explore the technical details beneath DeepSeek to explore what makes it unique and distinctive in the world of AI models.
www.thoughtworks.com
February 1, 2025 at 10:29 AM
Reposted by Kanzie
My colleague Prasanna Pendse has written a deep dive into DeepSeek: first impressions of usage, highlights from their papers, and musings on what's next

www.thoughtworks.com/insights/blo...
Demystifying DeepSeek
We explore the technical details beneath DeepSeek to explore what makes it unique and distinctive in the world of AI models.
www.thoughtworks.com
January 31, 2025 at 5:16 PM
Reposted by Kanzie
DeepSeek R1 and R1-Zero Explained

thelmbook.com/articles/#!....
February 1, 2025 at 7:58 AM
Great reporting to put a light on the practice of outsourcing crunch and irresponsible offshoring. But just as relevant to western startups with narcissistic leaders. Gaslighting and manipulative practices so easily become normalized when we let abuse go unchecked: youtu.be/RR9HQ2C6h_4
100 Slaps: The Breaking News The Games Industry Ignored in 2024
YouTube video by People Make Games
youtu.be
January 31, 2025 at 1:36 PM
Reposted by Kanzie
They will do anything to steal your attention. Do not let them. Ignore the fuckers. Sing, dance, and create, every day. Make things. Give love. Be generous. Be kind. Focus on what brings you joy. Fuck them and their needy, insecure, hateful bullshit. You are in control of your attention. Always.
January 21, 2025 at 7:47 PM
Running DeepSeek R1 8B on my own server with a rtx3080 and it can’t answer any more complex prompt without collapsing into Chinese. I’ve dropped temp, top_k and p to no avail? Anyone know if it’s really useless at 8B or what do I tweak? (Ollama, openui) #deepseek
January 29, 2025 at 10:42 AM
Reposted by Kanzie
You know, I thought my challenge with Severance was how long it took to come back, but I think it’s more that the story mechanics are now so complicated that when I watch it, I feel like I’m playing a game with incredibly complicated rules that I have to keep looking up.
January 29, 2025 at 1:40 AM
Reposted by Kanzie
New Let’s Chat Games: Playing the Game the way the Developer Intended

christianspicer.substack.com/p/lets-chat-...
Let's Chat Games 019 - How the Developers Intended
Thinking about the notion that games should be played "how the developers intended"
christianspicer.substack.com
January 28, 2025 at 8:13 PM
Help me clarify this. Is the excitement that DeepSeek R1 performs so well or that it does so reportedly only at 30% of the cost/effort to train?
#deepseek
January 27, 2025 at 9:19 AM