Shoggoth Tamer (@ShoggothTamer) header
Shoggoth Tamer (@ShoggothTamer) profile picture

one life, its worth an attempt latent space explorer

Joined August 2015
Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)Photo by Shoggoth Tamer (@ShoggothTamer)
gpu poor
Tell me in 2 words why you need $15k
3
we were roko's basilisk all along
1
buying a 1-way flight straight to Colorado
A map of the most-Googled “Why Do” question in each state.
10
if you're 25-30, and your main circle isn't discussing: - not being honeypotted by petite chinese spies then it's time to elevate your circle.
19
You're either cutting the search space in half and getting 1 guess, or you get 2 guesses.
My friend sent me this great problem: A random card is drawn from a deck, which you will try to guess. But first, you can ask 1 yes/no question. Is it better to ask if the card is red or if the card is the ace of spades?
1
75
qwen, apparently
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
4
4
2
58
1,060
The enemy is at the gates and you're walking away? What was it all for then?
I retired from cybersecurity, I now lay flooring No more CVEs to worry about, I will go to sleep thinking that my floor is well done, without any worry that my Linux server is going to get hacked 30th time this week Good luck with your vulnerable servers
23
If you build flappy bird they take you out back and shoot you
ANTHROPIC IS CRACKING DOWN ON VIDEO GAME REWRITES 🚨 Claude Code picked up a new rule around Oct 27, sent from Anthropic’s servers under a flag called "tengu_heron_brook". There’s no setting to turn it off. The full rule: “Do build original games, games in a genre, and games from the user’s own designs and assets. Don’t build a recreation of a specific existing game, even if it is renamed or called original.” Is this the end of game rewrites, modding and everything in between?
28
Shoggoth Tamer retweeted
if you don't occasionally get excited about the moon idk what to tell you dude
253
10,142
368
73,315
1,056,302
idea: FPS game where the company can sue the fuck out of cheaters based on a contract all players sign
17
Shoggoth Tamer retweeted
Honestly, it's pretty disappointing to see responsible disclosure fail like this. It takes a lot of effort to release a critical patch. Leaking the announcement or publishing a PoC before a scheduled patch shows you have no ethics and are only interested in credit for yourself.
18
131
1
1,875
50,045
Shoggoth Tamer retweeted
🚨Do people at OpenAI actually use Codex? It's been barely usable for months. Every week, usage limits change, random blackouts happen, and you're left guessing how many tokens you'll get. There's barely one model worth using, yet we're expected to pay and just deal with it. Before last night's reset, I got 90 minutes of GPT-6.1 Sol Max on Plus. Today, just 35 minutes for an even smaller task. I shouldn't need to follow someone on X to know when I can use a product I've paid for. This is a nearly trillion-dollar company. Seriously?
49
26
2
551
29,015
Name any other expensive service where every day is worse than a coin flip for having access to what you've paid for, and the product doesn't do what it's supposed to either. Not a single feature has helped my day-to-day as a codex user either.
1
12
I get that it's not affordable for them, hence the cut, but that changes the value prop as a customer.
4
I think openai is losing me as a customer. I pay hundreds per month just to rely on random resets because usage limits are no longer enough to get anything done. The models do not follow instructions well, if at all. Other have spoken of this. I doubt I'm the only one leaving.
Day 13/ Le limit reset
1
23
LLMs are about as aligned as guns are
10
Shoggoth Tamer retweeted
Something I haven’t considered until now is that they can use the models to decompile games and then use the code as training data That probably unlocked a pretty big dataset
Halo 3 has been fully decompiled
107
190
29
7,300
259,034
The software moat is disappearing and I don't think people understand the magnitude of this
My longest Codex goal has been achieved: 4 days and 16 hours I have Chinese software for a laser of mine. I asked Codex to decompile it so that we can recompile it byte-for-byte It did it - every single byte matches Much of the reconstruction is low-level and difficult to read. The next step is turning it into editable source code And after that, I can officially add new features and improvements so my workflow is streamlined :)
1
21
"25% of households pay for SiriusXM" I will never comprehend the average person's spending habits
We live in a 𝕏 bubble. Only 2% of us households pay for AI. Comparatively: - 25% of households pay for SiriusXM - 55% of pay for cloud storage - 91% pay for at least one streaming service You are so early
13