brymko (@brymko) header
brymko (@brymko) profile picture

Exploit Dev, Security Researcher, Hobby Quant, Solving Problems Ex ctf @FluxFingers, @allesctf, @Sauercl0ud Affiliated @the_secret_club

Singapore
Joined January 2019
Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)Photo by brymko (@brymko)
traveling without airtag in luggage #whereismyluggage
3
6
1,818
Honestly I feel hopeful, all these discoveries are not very clean bugs. Feels like the fracing phase of vuln research
1
4
451
touch your security key, enter security pin, touch your security key again, we have sent an code to your email, verify idendity, touch your security key again, enter security pin, touch your security key again, use passkey to login
2
295
Very not concerned about PII leaks given I have to verify my identity like 2 times a week to retain CyBeR AcCesS on sota providers. Must be losing them or smth
3
290
made $31m ARR today (my friend sent me $20 in 2 seconds for his share of the lunch)
we’ve hit $2m ARR in 5 hours not a joke btw
252
statistical outlier iq people on my timeline getting nerdsnipped by terminal emulator performance
2
174
Never leave the house without your bulletproof vest
6
795
OAI training new model i suppose or something cause performance is super degrading. Tempted to swap to claude in the meantime but then i got triggered to the moon after seeing "load-bearing" and "honest" appear after the first prompt so i.g I gotta write some code by hand now
1
3
269
Hes right, but where is the new abstraction layer. It's not going to be the "You are right to push back on that" chat window
“Why would I look at code? It’s like assembly, like a compiled artifact.” Sensei @karpathy fully agentic, 18m after “I don’t really use autocomplete AI code tools”
1
333
These guys are doing a huge service to humanity.
On July 25, our team hacked OpenAI. It took us less than 72 hours. Two vulnerabilities chained together gave us access to ChatGPT and Codex accounts belonging to OpenAI employees. We demonstrated the impact with a harmless PR in OpenAI’s internal monorepo. The full chain: HEIF upload → libheif heap overflow → RCE → OpenAI SSO flaw → ChatGPT/Codex takeover → connected GitHub → internal PR. OpenAI fixed the SSO issue roughly 14 hours after our report. Research by @rootxharsh, @S1r1u5_ and @iamnoooob. Full technical write-up: hacktron.ai/blog/hacking-ope…
1
14
1,390
On July 25, our team hacked OpenAI. It took us less than 72 hours. Two vulnerabilities chained together gave us access to ChatGPT and Codex accounts belonging to OpenAI employees. We demonstrated the impact with a harmless PR in OpenAI’s internal monorepo. The full chain: HEIF upload → libheif heap overflow → RCE → OpenAI SSO flaw → ChatGPT/Codex takeover → connected GitHub → internal PR. OpenAI fixed the SSO issue roughly 14 hours after our report. Research by @rootxharsh, @S1r1u5_ and @iamnoooob. Full technical write-up: hacktron.ai/blog/hacking-ope…
84
386
99
3,447
403,382
Is it only me or does alot of the LLM Architectures look very similar to cryptography ?
1
404
lost so much weight none of my pants fit anymore and my belts are at the last hole barely fitting. Need some recs for style in sg otherwise just gg uniklo again ig
4
7
779
Aaah, perfect product-market fit for robbers. Now you can *literally* beat the keys out of the victim.
A hardware wallet that makes signing cost you. Pulse for everyday. Blood when you mean it 🩸. OS. Bitcoin & Ethereum. 3D printable w/ raspberry pi ($90 in materials). </CELL/>
2
579
Got targeted by a "semi" sophisticated phishing/scam attempt just now. Step 1) Receive an official email from google that my email was used as a recovery email for [email protected] (first image). Step 2) (optional im not sure) Click on remove email. Appeard to be already removed. Step 3) Receive call from SG registered number 3106 2518 (all previous attempts have been from foreign numbers). Step 4) Matthew from "google" is very professional in following a well thought out script that might come straight out of a real support playbook. Step 5) Matthew sends whatever that email is. (second image) The link in question points to the official hotelprices support page on support(.)google(.)com Headers: from:[email protected] reply-to:[email protected] to:[email protected] < At this point I kind of got excited and tried to record, but my nontechnical ass didn't know that the iphone call recording will tell the other side and Matthew hung up > My guess is that given the link doesn't work, the fake support agent would tell me to type in a custom url or something.
1
12
743
oh boy "frontier" models losing perf again, I wonder what that means... I hope this strategy blows up in their face when everyone switches to some OS non flaky chinese model
264
yea cause the train will be delayed 30
"I can...roll up to the train station 10 minutes before departure" you can do it... with *german* high speed rail. with chinese HSR you can't, because their stations are located far from the city center and have airport-style boarding.
723
nice bug in csgo mate
OOB reads in network message handlers leads to RCE VALVE Bug Bounty PoC $7500 Bounty Follow us: youtube.com/@NullSecurityX #BugBounty #CyberSecurity
1
4
787
Yeah its gonna be great, trust.
Introducing Claude Code Security, now in limited research preview. It scans codebases for vulnerabilities and suggests targeted software patches for human review, allowing teams to find and fix issues that traditional tools often miss. Learn more: anthropic.com/news/claude-co…
1
739
using vc money to pay for my $15 cny grab delivery surcharge (i made a new account)
1
320