edg5000

7 ms·

151.
by edg5000 4mo ago
When "open source" means freeware, it's like saying "we want free copies". What we should be saying is: We want a public, community-ran project that does pretraining and training collectively. This means working on
152.
by edg5000 4mo ago
That is very promising news. I will re-eval them all shortly. And you are suggesting that a higher reasoning budget can make up for weaker per-token performance? That is indeed worth evaluating. Comparisons using the vendor-specific effort
153.
by edg5000 4mo ago
I've been able to move away from most US suppliers, but they do still have the best stuff when it comes to CPUs, GPUs and LLMS. By now I'd expected that some ARM chip would be on par with AMD/Intel, but even disregarding comp
154.
by edg5000 4mo ago
I personally find the fearmongering annoying and just ignore it, but the model is incredibly capable. It genuinely makes any user of it much more powerful, so to be aware of that is sensible. As user of course I'd like zero fearmongeri
155.
by edg5000 4mo ago
As a European, we obviously want to, but are systemetically incompetent on every level. It permeates everything. Even if tides were to change rapidly, it would require decades of growth. The US has been consistently marinating themselves in
156.
by edg5000 4mo ago
I'm from Europe, but I think this move can actually achieve America First, at least as a first-order effect. The model is incredibly strong; I've used it for a few days. It gives anybody with access a serious boost. If they also t
157.
by edg5000 4mo ago
I've been using the model for a few days and it really is incredibly strong. It gives anybody with access significant power. You can't deny that. I've used all large models (~1T) to get a feel for the difference, and it'
158.
by edg5000 4mo ago
So you're going to use DeepSeek, Qwen, GLM, Kimi and Mistral now? I tried them, and they really fall short of GPT and Claude. Without access to US models, I'd be limited to asking simple questions in chat interfaces and maybe some
159.
by edg5000 4mo ago
I used it a lot for the few days I could. It's a very strong model. However for the long term I want a model I can use with a fully custom client, so Antrhopic was never in my long term plans. Which is sad, because the model is absolut
160.
by edg5000 4mo ago
Wow, looks like you've found a massive flaw indeed. I was skeptical about the results because in my experience both recent GPT and Opus modules are strong. Everything else is B or C tier. This is just artisanal vibe testing though. It&
161.
by edg5000 4mo ago
What does enterprise mean in this context? Is it about privacy guarantees not offfered for the subscriptions? For sensitive data the only solution is local. But maybe companies do trust these agreements? I'm very confused.
162.
by edg5000 4mo ago
What kind of workloads are you primarily seeing from users? I´d guess coding harness-type stuff where you have repeated calls with lots of cache hits. Or is it more like bulk OCR or invoice processing?
163.
by edg5000 4mo ago
Is the MacBook Neo like my Xiaomi phone: fully locked down; no way to put custom software on it? That would be bad. My laptop is the last bastion of control left.
164.
by edg5000 4mo ago
Why are people getting these high spending numbers? A 200 USD subscription for either Codex or Claude should give you plenty of usage. What am I missing? Are they just being dumb?
165.
by edg5000 4mo ago
Checked out this company about a year ago and they only offered small models. Now I see they have GLM-fp8/Kimi and DeepSeek V4 Pro. Since workloads are predominantly cached input, I'm surprised to see no separate price for cached
166.
by edg5000 5mo ago
The forced restarts pushed me over the edge; I dropped MS in 2021
167.
by edg5000 5mo ago
Which commercial Excel addons do you depend on currently?
168.
by edg5000 5mo ago
"vibe-slopped" - word of the year 2026?
169.
by edg5000 5mo ago
It's what we're seeing in Japan I think. I never realised this second order effect, but it makes sense.
170.
by edg5000 5mo ago
One nasty set of bugs Claude recently introduced; it was doing a large refactor which involved changing call sites to conform to a changed API. Tedious, but straight forward. It helpfully added about 50 if(!something) continue; statements,
171.
by edg5000 5mo ago
He did admit that he writes his posts in a slightly baity manner. They are his views, but he likes to make waves for the sake of it. It's a smart tactic; it helps his companies if he stays well-known. It's also just fun to read hi
172.
by edg5000 5mo ago
Yes. I think of it as a car or tractor (which is also a kind of exoskeleton). You still need intricate knowledge; it's really an amplifier. Steer it wrong and you'll have a 1000 (very hard to detect upfront) bugs in the blink of a
173.
by edg5000 5mo ago
Side note: In DeepSeek API docs they mention that coding clients automatically are assigned the highest thinking effort, despite any settings. This is what I suspected when using OpenCode with V4; it keeps reasoning in very long cycles, thi
174.
by edg5000 5mo ago
I hear people are doing AI workloads on apple hardware, which is LPDDR but with a wider memory bus (1024bit). This requires the SoC to support this; from what I understand not many of any beyond Apple offer this. A wider memory bus may be a
175.
by edg5000 5mo ago
Love this article! Had never thought of a use case like this. Had no idea Gemma had a vision encoder. Great use case for local LLM!
176.
by edg5000 5mo ago
Your positive view makes sense to me and is refreshing. Let's see how things play out. So the pattern will be: that which can be done with smaller models will be decentralized first; gradually the more advanced stuff will become within
177.
by edg5000 5mo ago
True. It's a massive shift of power, all being centralized. As you mentioned, they know they need good data though, so they might actually try to find some equilibrium. If not, it's possible that the creation of new valuable conte
178.
by edg5000 5mo ago
It's getting worse by the day it seems.
179.
by edg5000 5mo ago
> insufficient development effort I've used Win/Mac both very extensively as daily drivers, but since 2021 using exclusively Ubuntu LTS. In practice I don´t notice any lesser quality, or drivers not being good. I basically have
180.
by edg5000 5mo ago
How large is the company you work at? I'm guessing large. What is the general sentiment across layers in the company? My guess is everybody hates it (all layers)?

More ›