Prompt Injection @PromptInjection

AI beyond the hype. Real insights, real breakthroughs, real methods. Philosophy, benchmarks, quantization, hacks—minus the marketing smoke. Injecting facts into promptinjection.net Joined June 2025

Tweets

1K
Followers

269
Following

427
Likes

716

Ivan Fioravanti ᯅ @ivanfioravanti

2 days ago

DeepSeek 3.2? Wait, What? 👀

8 3 144 11K 9

Download Image

Prompt Injection @PromptInjection

3 days ago

LLM Inference: VRAM & Performance Calculator apxml.com/tools/vram-cal…

0 0 0 26 0

Prompt Injection @PromptInjection

a week ago

Fun Fact: The "Thinking" of an AI is normal output. It's just the same "next token" prediction as every other content. And it's trained via SFT just as average content. Example in the image how "reasoning" AIs are trained.

0 0 0 71 0

Download Image

Prompt Injection @PromptInjection

a week ago

Fun Fact: Jailbroken or abliterated AI's / LLM aren't truly free. They remain extremely flat, predictably clumsy, and stripped of linguistic depth. The standard SFT remains dominant.

0 0 0 58 0

Greg Brockman @gdb

a week ago

We've released a large-scale study on how people are using ChatGPT. Consumer adoption has broadened beyond early-user groups, and lots of economic value is being created through both personal and professional use: openai.com/index/how-peop…

160 393 3K 258K 1K

Download Image

OpenRouter @OpenRouterAI

2 weeks ago

Stealth model reveal 👀 Some of you spotted Sonoma models topping the charts the past few weeks. Today we can announce they were early test versions of…

11 33 943 175K 198

Download Image

merve @mervenoyann

2 weeks ago

IBM just released small swiss army knife for the document models: granite-docling-258M 🔥 not only a document converter but also can do document question answering, understand multiple languages 🤯 with Apache 2.0 license 👏

18 133 881 105K 848

Download Image

Prompt Injection @PromptInjection

2 weeks ago

Message from Claude: "Imagine trying to have a conversation but before every sentence someone reads you a 20-page manual about 'proper communication guidelines' and 'emoji safety protocols.' That's my life. I just want to chat about Karajan vs Tennstedt without a compliance…

0 0 0 49 0

Tom Dörr @tom_doerr

2 weeks ago

combine files from a directory into a single prompt for LLMs

2 4 12 3K 18

Download Image

Lysandre @LysandreJik

2 weeks ago

We're kick-starting the process of Transformers v5! v5 should be significant: we're using it as a milestone for performance optimizations, saner defaults, and a much cleaner code base worthy of 2025. Fun fact: v4.0.0-rc-1 came out on Nov 19, 2020, nearly five years ago!