News for hackers. Stuff that matters.
Re:Intro to DeepSeek's open-source week and why it's a big deal by deepdarkforest on Thursday March 06, 2025 @03:56PM (#43284992)
Love when someone that actually knows their stuff does a deep breakdown like this. Super useful

I wonder though if all it matters is the last punchline. The profit margin vs competitors. If llms truly get commoditized and do not benefit from economy of scale(like infra/aws) because they would be run in house, then would all companies that use them be in a nash equilibrium? Like everyone has access to the same stuff, same reasoning, same benefits. The only people who will really win in enterprise are consultants that find the best way to leverage and implement agents, not the model makers themselves. And of course data will still be king

Re:Intro to DeepSeek's open-source week and why it's a big deal by DanteBLN on Thursday March 06, 2025 @02:12PM (#43284013)
Interesting read. Great to see ppl actually delivering with these open-source releases that really are what openAI should have done…
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @02:16PM (#43284053)
in a parallel universe...
Re:Intro to DeepSeek's open-source week and why it's a big deal by Couch_Surfer on Thursday March 06, 2025 @03:49PM (#43284942)
Great stride ahead for OS, curious to see how inference providers respond
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @04:06PM (#43285076)
If they don't use some of the released kernels, then they are leaving money on the table
Re:Intro to DeepSeek's open-source week and why it's a big deal by augustinemp on Thursday March 06, 2025 @01:41PM (#43283605)
I wonder if this openness will inspire other teams to share more of their infra. Imagine how quickly the field could move if more labs took this approach!
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @02:12PM (#43284008)
The level of DS's openness has been really useful to actually understand what LLM folks are working on day in day out
Re:Intro to DeepSeek's open-source week and why it's a big deal by lukaspetersson on Thursday March 06, 2025 @02:21PM (#43284106)
Cracked team vs export controlls, who will win?
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @02:45PM (#43284349)
Never bet against a cracked team
Re:Intro to DeepSeek's open-source week and why it's a big deal by linklater12 on Thursday March 06, 2025 @01:41PM (#43283615)
So system design is still relevant and it's not all ML and math?
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @02:08PM (#43283956)
Indeed. Modern LLM research is more about large-scale systems design than deriving gradients by hand. Math matters, but systems win.
Re:Intro to DeepSeek's open-source week and why it's a big deal by tsunego on Thursday March 06, 2025 @01:30PM (#43283449)
Uff this is quite a dense post
Re:Intro to DeepSeek's open-source week and why it's a big deal by nazgulsenpai on Thursday March 06, 2025 @01:36PM (#43283543)
The best kind for reading on that conference call you have no idea why you were invited to! I mean yeah, I agree.
Re:Intro to DeepSeek's open-source week and why it's a big deal by t55 on Thursday March 06, 2025 @02:13PM (#43284017)
agreed, quite ambitious to cover this entire week in one post
Re:Intro to DeepSeek's open-source week and why it's a big deal by sadboots on Thursday March 06, 2025 @02:01PM (#43283885)
better than all this bulletpoint fluff these days