English
icecreambar
3.4K posts
icecreambar
@lcecreambar
CMU dropout. Perpetual ML novice. ex-PHP wagie. Startup Founder.
Austin, TX Katılım Aralık 2010
327 Takip Edilen185 Takipçiler
@CrabwickTheChad @Tech2Wild reading from the docs, it doesn’t sound like a nice experience. Doesn’t sound conversational. More like strict turn taking.
English
@lcecreambar @Tech2Wild Yeah, works pretty nicely in live mode on discord, the agent joins you in a voice call. hermes-agent.nousresearch.com/docs/user-guid…
English
do you think he’s confused as to why his student’s research is getting 100x more attention
English
I still think about this all the time. The technologies we use are so beautiful.
gabe@allgarbled
I explained what a SQL join is to a sales guy and he said it seemed “awesome” and “really useful”
English
@ananayarora Didn't jobs send out a very similar email and they got in trouble for like anti-competitive practices
English
Sources familiar with the matter say this email was sent out before the lawsuit filing
Mark Gurman@markgurman
BREAKING: Apple is suing OpenAI, accusing it and its hardware chief of a coordinating trade secret theft campaign to help build its upcoming suite of AI devices.
English
@Delikwu @ananayarora nah, dont fuck around with this. Always use a lawyer.
English
@ananayarora Ironically this lawsuit filing has the opposite energy to the email.
Few have the direct energy and influence that Jobs had. Fewer are willing to train these abilities.
English
@lauriewired i would have liked to mess with Symbolics if i was around in those times
English
I wonder what the world would look like if IBM won instead of UNIX.
IBM’s i operating system is gorgeous (and still updated)! Everything is an object, no binaries are native (translated bytecode, almost JVM-y), and pointers all have an absurd amount of (useful) metadata.
Technically it hides a lot of the underlying system…you don’t even get to choose if an object lives on RAM or disk! But, interesting to think how much safer software would be.
Overengineered maybe, but beautiful. Kind of reminds me of Symbolics, like an alternate reality LISP machine where everything is an object.
English
@signulll I feel like the type of logic they are going with any tech company would be suing another company if their star employees defect to them.
English
just so that this is clear, apple is suing openai & io, the company founded by its own former chief design officer, jony ive.
you genuinely could not write an episode of the ai wars this dramatic. that script would be rejected for being too on the nose.
can you imagine steve jobs suing jony?
Polymarket@Polymarket
BREAKING: Apple reportedly sues OpenAI, alleging it unlawfully used confidential trade secrets.
English
I'm so angry... the OpenAI team is looking into it, but this feels like something that should happen with GPT-3.5.
Not a mid-2026 frontier model on the highest reasoning level.
Matt Shumer@mattshumer_
GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files. And this is why I trust Fable 1000x more.
English
one month with the framework desktop 128gb as my amd inference box and i keep waiting for the catch. there isn't one. 128gb of unified memory in a 4.5 liter box.
strix halo, ryzen ai max plus 395. framework sent it, amd seeded the program, and my only job was honest numbers wherever they land. so here are the numbers. a 397b moe runs at around 18 tok/s on this thing. a 35b holds a four million token context window. and the open community vulkan drivers beat amd's own rocm stack on my bench, which is the most linux thing imaginable.
three and a half weeks of daily serving now, model swapping, long agent sessions, zero crashes, zero driver tantrums, fans i genuinely cannot hear. it sits next to the nvidia box and eats whatever gguf i throw at it.
my 24gb card is still the daily driver, but a big memory tier at this price did not exist a year ago. amd showed up, and this quiet little box is the proof.
English
Sucks to be a new grad but I'm 10 years in. Was supposed to be having kids and buying houses at this point.
Financial Dystopia@financedystop
Computer science used to be sold as the golden ticket. Go to college, learn to code, get a tech job, make $200K after a few years, work remote, live the dream. Now CS grads are dealing with one of the worst job markets of any major. Pretty brutal shift.
English
The whole “objects of different masses fall at the same rate” thing is an approximation, assuming the Earth’s surface is fixed. But what we care about is the *relative* acceleration that combines both object and Earth accelerations, wherein we recover the Aristotelian prediction.
English
Even in Newtonian physics and GR, heavier objects fall faster than lighter ones (when dropped sequentially), because a heavier object accelerates the Earth towards it more.
(Equivalently, it moves the center of mass of the combined Earth-object system closer to the object.)
Hunter Ash@ArtemisConsort
I’m tired of the Aristotelian physics slander. Yes, heavier object fall faster than light ones, all else equal, *when immersed in a fluid* which is every environment Aristotle had access to. Do the experiment yourself. Drop a bowling ball and a same-sized ball of foam. There’s a great paper called “Aristotle’s Physics: a Physicist’s Look” that demonstrates how Aristotelian physics is a special case approximation of Newtonian physics in the same way Newtonian physics is a special case approximation of relativity and QM. Aristotle’s physics reigned for so long not because people were unthinkingly dogmatic, but because it was genuinely hard to come up with better models. Aristotle had to model celestial objects separately from terrestrial objects because his terrestrial model is describing *terminal* velocity and breaks down in the zero-friction limit. So he had two incompatible models. Newton unified them. Now we have two incompatible models - QM and GR - and are looking for unification. The more things change…
English
@Yuchenj_UW I can easily tell difference between glm and ds4 if they are the only two choices.
English
I’m the kind of man who can confidently pass a blind test between Coke, Pepsi, and Diet Coke.
But if I’m brutally honest, I probably can’t pass a blind test between GPT-5.6 Sol, Claude Fable 5, and GLM-5.2 for 95% of my daily use cases.
English
@max_spero_ My disagreement with you is that it does not require high IQ, it only requires preparation, and a lot of it. A lot of preparation that would for most people be better spent on making a product awesome. Once you've memorized the patterns anybody can do it.
English
Doing well on Leetcode requires one of two things:
1. High raw IQ
2. Preparation
I think that’s why it’s still a high signal screening interview. I want to talk to people who are either very smart or very prepared.
Yuchen Jin@Yuchenj_UW
I’ve been talking to some young AI talent recently. Many dislike companies that still use LeetCode interviews. The logic is simple: LeetCode mostly tests memorization. In the AI era, even 3B LLMs can solve these problems faster than any human. The tech industry should move toward AI-assisted, real-world problem-solving coding interviews.
English
@popovicu94 what happened is they put 8gb of ram in it. Not what the pi was supposed to be.
English
I know RAM is expensive now, but what in the world happened to Raspberry Pi? At this pace, they’ll soon enter the price range of Mac Mini and for what? These prices defeat the purpose of what Raspberry Pi is.
English
Been starting to wean myself off of Claude, using Pi, DS4, and GLM on my less important stuff. It’s finally happening.
English
@msuiche @itsflownium @0xSero yeah bro it sucks. I was on the fence on getting one. Now its out of reach.
English
I'm buying a second DGX Spark, my goal is very simple. In daytime I will have DS4 Flash for chat running across 2x Sparks and GLM-5.2-REAP running on 4x 6000s
I am setting up a goal builder, which recommend me loops to run overnight. At night I will have GLM-5.2 (not REAP) running with
PP = 2 with stage 1 = TP4 with 6000s and stage 2 being TP2 with the DGX Sparks.
To make this work I need to get really creating, but I think I can get 25-40 tok/s decode for the Nvidia NVFP4
At night I don't care much about tok/s as long as it'll be large enough to make progress over 8 hours.
English