aitization 𝕏  @aitization

AI, ML, research, data, cybersecurity, privacy, learning and improving, solution focused, critical thinking, startups, business, VALUE CREATION ＄ Global Joined August 2023

Tweets

10K
Followers

189
Following

603
Likes

44K

The Hacker News @TheHackersNews

4 days ago

🔴 Most SIEM alerts are junk. 74% of breaches had warnings — but analysts ignored them, buried under 5,000+ daily false positives. The price? Burnout, wasted millions, and missed threats. Why the SOC’s beating heart is now its weakest link → thehackernews.com/expert-insight…

1 29 126 14K 37

Download Image

NSA Cyber @NSACyber

a week ago

Operational Technology (OT) cyber incidents can have severe consequences for OT systems, especially those in NSS, the DOD, & the DIB. Increase your asset & vulnerability awareness by checking out the recently released joint CTR with @CISAgov and others. media.defense.gov/2025/Aug/13/20…

18 48 151 8K 33

Download Image

alphaXiv @askalphaxiv

a month ago

Iterative reflections for LLMs can outperform heavy RL? This paper shows that having the LLM reflects on its own trajectories, rewrite its own prompts, and evolve a diverse pool of candidates beats RL w/ GRPO so far on four reasoning tasks . 10% improv with 35x fewer rollouts!

6 35 227 15K 165

Download Image

Jackson Atkins @JacksonAtkinsX

a month ago

LLMs can now self-optimize. A new method allows an AI to rewrite its own prompts to achieve up to 35x greater efficiency, outperforming both Reinforcement Learning and Fine-Tuning for complex reasoning. UC Berkeley, Stanford, and Databricks introduce a new method called GEPA…

48 337 2K 204K 2K

Download Image

Yi Ma @YiMaTweets

2 months ago

To me, the real danger of current artificial intelligence is that the whole society increasingly relies on "knowledges" generated by machines that in fact do not understand them. It is not the action of current AI systems that would doom us but our own misjudgment and behaviors…

10 19 175 16K 44

elvis @omarsar0

2 months ago

One Token to Fool LLM-as-a-Judge Watch out for this one, devs! Semantically empty tokens, like “Thought process:”, “Solution”, or even just a colon “:”, can consistently trick models into giving false positive rewards. Here are my notes:

13 122 700 99K 640

Download Image

Yoshua Bengio @Yoshua_Bengio

2 months ago

The future of AI governance may hinge on our ability to develop trusted and effective ways to make credible claims about AI systems. This new report expands our understanding of the verification challenge and maps out compelling areas for further work. ⬇️

Ben Harack @benharack

2 months ago

3 31 85 31K 42

Download Image

12 21 118 10K 27

Andrej Karpathy @karpathy

2 months ago

I often rant about how 99% of attention is about to be LLM attention instead of human attention. What does a research paper look like for an LLM instead of a human? It’s definitely not a pdf. There is huge space for an extremely valuable “research app” that figures this out.

Michael Levin @drmichaellevin

2 months ago

267 319 3K 721K 973

284 428 5K 609K 2K

Michael Levin @drmichaellevin

2 months ago

I'm constantly irritated that I don't have time to read the torrent of cool papers coming faster and faster from amazing people in relevant fields. Other scientists have the same issue and have no time to read most of my lengthy conceptual papers either. So whom are we writing…

267 319 3K 721K 973

Anthropic @AnthropicAI

2 months ago

New Anthropic research: Why do some language models fake alignment while others don't? Last year, we found a situation where Claude 3 Opus fakes alignment. Now, we’ve done the same analysis for 25 frontier LLMs—and the story looks more complex.

70 270 2K 455K 1K

Download Image

Anthropic @AnthropicAI

2 months ago

Today we published a targeted transparency framework for frontier AI development. Our framework focuses on major frontier model developers while exempting startups and smaller developers to avoid burdening the broader ecosystem.

51 181 1K 133K 368

Download Image

The Hacker News @TheHackersNews

2 months ago

⚡ Google just got hit with a $314M verdict — for secretly using your mobile data while idle. Phones sat still. It sent 900+ background transfers a day — and you paid for it. Details here → thehackernews.com/2025/07/google…

20 141 318 104K 90

Massachusetts Institute of Technology (MIT) @MIT

2 months ago

Happy 4th of July weekend, everyone! We hope the fireworks are spectacular.

17 57 239 15K 13

Download Video

Quant Science @quantscience_

2 months ago

This guy literally shows the code for how to automate ingesting 90,000,000 rows of financial data for his quant hedge fund:

8 61 497 22K 790

Download Video

unusual_whales @unusual_whales

2 months ago

In more than 230 U.S. cities, even a starter home costs an average of $1 million, according to Zillow

382 866 6K 703K 398

Ai2 @allen_ai

2 months ago

AI practitioners need to know if their models are performant, safe, biased, or hallucinating. That’s why Ai2 offers a range of evaluation frameworks and benchmarks, all of which are open and accessible. 🧵

3 5 36 3K 16

aitization 𝕏  @aitization

2 months ago

I'm in full compliance with all community standards & user policy, I only use @LinkedIn as intended by the founders for professional networking only @reidhoffman @LinkedInHelp WHY did you log me out for NO REASON TODAY. I never use any third-party tools or apps. 1 clean device ⚠️

0 0 0 21 0

The Hacker News @TheHackersNews

2 months ago

🚨 AI agents are leaking sensitive enterprise data — and most orgs don’t even know it. GenAI tools often connect to S3, SharePoint, Google Drive… without proper controls. The risk? Silent breaches. Join the webinar on July 7 to learn how to secure your AI workflows ↓…

4 22 57 16K 17

Download Image

EFF @EFF

2 months ago

This bill doesn’t close gaps in California privacy law, it blows them wide open. S.B. 690 would roll back decades of hard-won protections. Tell your lawmakers: vote NO. act.eff.org/action/ca-stop…

0 12 28 6K 2

Andrej Karpathy @karpathy

2 months ago

Love this project: nanoGPT -> recursive self-improvement benchmark. Good old nanoGPT keeps on giving and surprising :) - First I wrote it as a small little repo to teach people the basics of training GPTs. - Then it became a target and baseline for my port to direct C/CUDA…