Cyber Coach | Story Teller | Internet Gravedigger | Mad Data Scientist ❤️ food, cats, bad movies, and the worst TV

The Golden City
Joined March 2018
#crypto solves 0 problems for finance just like #tesla solves 0 problems for transportation and climate change. They are modern tools for the bourgeoisie to financially control the proletariat. Happy Holidays!!!
7
Cobalt Strike is a crutch
2
2
160
I can't spend a Toyota to store a few system logs, sorry.
126
We hit a massive milestone for the company over the holidays, we’re at a point where we can dogfood it and use it to secure the access to the repositories that we build it in.
1
1
16
967
One of the last red team ops we ran was called floss boss and was meant to be a returning actor from a previous op, molar bear. We got branded dental hygiene kits made as propaganda to spread the lessons learned.
2
4
2
33
4,863
After 6 years our offensive security team is shutting down and today is my last day at Yahoo. It’s been a great ride and couldn’t have asked for a better team to work with.
14
19
3
387
39,824
Wow, you must be natural language processing, because you've got me at a loss for words
1
1
3
592
Sigstore is aimed to ensure privacy & scalability, integrates technologies for seamless signing, verification, & provenance checks. 🔏 Explore how Yahoo utilizes #Sigstore alongside Athenz as an internal Certificate Authority for container image security: openssf.org/case-studies/202…
12
23
3,180
🔑Pushing new container images at scale is hard. Verifying those images' authenticity is harder. Read how Yahoo uses an open-source project — Sigstore — to sign tens of thousands of images a day transparently! 🔑yahooinc.com/paranoids/scali…
2
1
13
1,563
Wild to see this company still cringe posting their "hip" ads when every other advertiser is pulling off the platform because of elno's support of antisemitic conspiracies. xitter.cf/vicariusltd/status/171…
Lil CISO remixes @CISAJen's four tips for Cybersecurity Awareness Month!
1
347
Introducing our newest game: Quizzle! 🎉 Can you guess today's word in fewer than 20 questions? Play now 👇 quizzle.game/
2
1
423
💯 Smaller models, tuned and configured to reason, coupled with RAG and knowledge graphs, will absolutely eat a lot of white collar jobs. Maybe not in the next 6 months, but it’s inevitable. That’s why it’s so important that the open source / open LLM community start *now* on standards for data governance, provenance, and licensing. We have the biggest opportunity for democratizing and resetting the scales of economic equity, since the rise of consumer internet and seminal open source like Linux and GCC.
You thought that you can go to sleep now?? Orca 2 Just dropped. Paper: arxiv.org/pdf/2311.11045.pdf Results: Orca 2 13B beats LLaMA-Chat-70B TL;DR: Training smaller model to reason by using multiple techniques: step-by-step, recall then generate, recall-reason-generate, direct answer And determining the most effective solution strategy for each task.
15
108
4
661
277,180
I've been a pretty vocal skeptic of AI and it's importance but recently started playing a bit with attacking LLMs after some nerd sniping from a co-worker. Did a few CTFs to learn some of the basics and decided I wanted to try it out for real. 1/n
1
2
1
5
2,567
jojo characters: "anyone in this crowd could be the enemy stand users." the enemy stand users:
What happened to masculine men?
218
9,793
83
99,530
3,944,426
You can generate the entire navy seal copypasta with this demo xitter.cf/yvrjsharma/status/1721…
Developers can now generate human-quality speech from text via the text-to-speech API released by OpenAI earlier today. @Gradio Demo -huggingface.co/spaces/ysharm… Explore all the available voices and models from OpenAI in this demo.
1
211
The hackings will continue until morale improves
7
153
Lot's of discussion about security keys disrupting the attacks against Okta. I don't have enough information to make a claim about that. They definitely are effective against phishing but it's not clear whether phishing was used. 1/n
1
1
3
373
The year is 1988. Earliest documented hostile exploitation of a buffer overflow. We learned that failure to distinguish code from data leads to serious issues. The year is 2023. We rapidly deploy LLMs where we intentionally co-mingle code with data. We'll learn that...
2
176