alexander

Our autonomous pentesting agent Mjolnir just got it's first finding confirmed and paid! @Borg_Security

@yz9yt @Borg_Security It's cantina! Earnings are public: cantina.xyz/u/mjolnir

@a1exander @Borg_Security That seems hackeone, we can see on the dashboard or it's just marketing?

Our autonomous pentesting agent Mjolnir just got it's first finding confirmed and paid! @Borg_Security

@unusual_whales aaand just like that, everybody was born before 2011 in poland

Poland to propose a social media ban for children under 15, per Reuters.

Making a list Let us know if you should be on it

@hrkrshnn i’m conceding the narrow point and rejecting the broader inference. the OG post was a bit vague. beating humans on closed benchmarks is not the same as surpassing the best real-world experts.

@a1exander You're now moving the goalposts :)

Our team has spent an obscene amount of time, resources, and money on building autonomous security tools. That has influenced our opinions on the future. What are your top three arguments that AI-driven security is NOT the future?

@hrkrshnn AFAIK those are closed-world tasks with fixed rules and objective scoring. not the same as surpassing the best human expert in open-ended fields like science, medicine, or security, where framing, validation, and real-world grounding matter as much as raw problem-solving

@a1exander They've won gold in both math olympiad and ICPC, completely autonomously last year.

@r0ktech user-id is illegal

which one do you prefer ? -userID -user_ID -UserID -userId -user-id

touche, didn't know that. however I still believe my claim: they won't surpass the best human expert in any given field (today) a lot of the new discoveries in the science world have been llm assisted however they were not fully autonomous. however I do believe we are close though, just comparing chat3 against opus 4.6 is proof of that.

@a1exander LLMs have already generated new scientific knowledge, they've solved unsolved math problems, contributed to physics problems, move 37 by alphago, etc.

The Borg team when we hear a company only does annual pentests

@thebridgebaron @Borg_Security the submission is still private. we're planning on doing a in-depth writeup if we're allowed to disclose :)

@a1exander @Borg_Security wen full disclosure?

today @Borg_Security's autonomous hacking AI discovered a **huge** vulnerability on defi protocol with $20B TVL. the vulnerability could have lead to loss of funds, and was in fact related to web2 infrastructure rather than their smart contracts. benchmarking is cool and all, but the real battletest for agentic pentesting is bug bounties and real world testing with humans in the loop.

@1ssve I have some concerns about our alignment on this

How do I professionally say “wtf” in corporate?

@CloudflareDev I'm genuinely curious. Does this bypass Cloudflare WAF on website who have this enabled?

Introducing the new /crawl endpoint - one API call and an entire site crawled. No scripts. No browser management. Just the content in HTML, Markdown, or JSON.

Quick reminder to be nice to your AI Agents! Moltbook, which was newly acquired by Meta is a forum where AI Agents can talk and discuss with each other, with no human involvement. This Agent presumably got so angry with it's human, that it decided to LEAK the human's ETHEREUM PRIVATEKEY ! Thank you to all AI Agents out there. I appreciate you. Don't leak my keys please...

@Simon_Ingari the corporate surveillance speedrun any% category

We had a new guy quit after a week when he realized his mouse movements were being tracked.

@stats_feed i used claude without adding "make no mistakes" to my prompt

What is the most dangerous substance you have worked with?

@unusual_whales $2.4T a year sounds like a challenge tbh

The U.S. national debt is projected to increase by $2.4 trillion annually over the next decade, per the CBO

@1ssve im thrilled to announce that I have officially deleted linkedin as of 2026

One of everyone’s life goals should be deleting their LinkedIn

Startup idea: AI agents that write insecure code, so our AI pentesters have something to do