Rudy

Proficient in Finance, Sports, Tech, Security and Blockchain

[Rudy @Rudy4FutureTech]

@Yuchenj_UW That’s exactly what someone with a $10 million net worth would say. Making $10 million is not a piece of cake in the Bay Area either.

[Rudy @Rudy4FutureTech]

@miramurati Same ol’ boring stuff. Look at the latency with which the model responds to an interaction. Also, this is nothing which OpenAI can’t dismantle in a week of time.

[Rudy @Rudy4FutureTech]

@cvewhen Did you finish it?

[Rudy @Rudy4FutureTech]

@ToBScottA @SEJeff @trailofbits Did the team consider wrapping Joern as an MCP server to provide data flow in addition?

[Rudy @Rudy4FutureTech]

Difficulty for humans is not the same thing as depth of problem structure. I’ll leave it at that.

[Rudy @Rudy4FutureTech]

Not surprising to watch the security community go through the five stages of grief now that LLMs are bulldozing the busywork layer we used to confuse with skill.

[Rudy @Rudy4FutureTech]

@LiveOverflow AI compresses a lot of the search/execution layer, but real pentesting still depends on judgment: what to chase, what to ignore, and where depth actually matters. Broad automated coverage isn’t the same as a real assessment. Hard CTF challenges still bottleneck on correct insight

[Rudy @Rudy4FutureTech]

@ItsKickyCares @naiivememe No way

[Rudy @Rudy4FutureTech]

@BrettBreakerZ @naiivememe She is absolutely ugly

[Rudy @Rudy4FutureTech]

@lucacarettoni Nice! Please don’t leave us hanging. Could you share which three platforms you tested?

[Rudy @Rudy4FutureTech]

@T5yll_ So you did find a way to squeeze in religion somehow into this. congratz!

[Rudy @Rudy4FutureTech]

@pashmerepat @deedydas Elon musk will help you. Go cry outside his home you loser!

[Rudy @Rudy4FutureTech]

@ParikPatelCFA @pashmerepat @deedydas what matters is the end result

[Rudy @Rudy4FutureTech]

@sracha What baseless reasoning and an attempt to make her choice seem rational and objective? The height difference between the couple has no bearing on the mortality rate during childbirth. @theweb3jess

[Rudy @Rudy4FutureTech]

@smackingg They definitely treated you differently by assuming you’re Kurdish or Middle Eastern. Japanese people can be incredibly racist (passive-aggressive) toward those who fit that appearance.

[Rudy @Rudy4FutureTech]

Well, ofcourse the agent architecture and orchestration is not a minor thing but the point here is you can only go so far due to model limitations.

[Rudy @Rudy4FutureTech]

LLM wrapper “security” startups:

1.“AI autonomously finds vulns + variants + patches”
2.“AI gives SMEs vuln primitive checks to assist”

In reality: a pitch deck, prompts, tools, rented intelligence from the same model providers with no means to solve deep security problems.

[Rudy @Rudy4FutureTech]

Even with the patch diff and a clear win condition of code execution, turning that into a real exploit is still hard work, not autocomplete.

[Rudy @Rudy4FutureTech]

This!

Security research isn’t “vibe coding.” You’re working in an unbounded problem space where you actually have to connect dots and understand systems E2E. There’s no cheap “success” signal like in software dev (builds pass, tests green) telling you you’re done.

#React2shell

[Guillermo Rauch @rauchg]

When the POC comes out, it’ll be a humbling moment for LLMs and how we use them. What’s circulating is extremely naive and incorrect.

Experienced engineers are sharing plausible-sounding hallucinations from frontier models.

Reminder to bump React, Next & frameworks.

[Rudy @Rudy4FutureTech]

Pretty sure every LLM-wrapper startup has a few “human researchers” tearing into this right now, purely so they can brag their agent “discovered” the exploit in the launch blog.

#React2Shell #CVE-2025-55182 #CVE-2025-66478