Alex

cybersecurity/ai/design

[Alex @Alexrcalz]

AI agents are getting scary powerful… but most still open random URLs with zero safety checks.
Phishing, malware, token drains — one bad link and it’s over.

Built a lightweight pre-browsing gate called Safe Web Confidence Protocol.

Before any agent loads a page it:
• Checks local allow/block lists (instant)
• Hits VirusTotal, Google Safe Browsing, urlscan, OpenPhish, OTX
• Returns a clear score + verdict: allow / sandbox / block

Fully explainable, cheap, and open source.
If you're building agents, this should probably be in your stack. https://t.co/CHikDKnMal

What safety measures do you actually use for your agents right now? 👇
#AIAgents #AISafety #OpenSource #Cybersecurity

[Justin Schroeder @jpschroeder]

Fable isn't the first.

In 1999 the department of defense blocked exports of the PowerMac G4 for crossing the 1 gigaflop threshold.

Steve Jobs turned it into an ad.

[Alex @Alexrcalz]

@andrewmccalip genius

[Alex @Alexrcalz]

@Jason My friends are all in their early 20s and are just misinformed due to reels lmao

[Alex @Alexrcalz]

@testingcatalog would be so nice if they add search for bookmarks

[Alex @Alexrcalz]

I ❤️ Austin

[Alex @Alexrcalz]

@viemccoy @justalexoki You can play multiplayer on Java now

[Alex @Alexrcalz]

@sporadica What about your future kids?

[Alex @Alexrcalz]

@ramankhatri did u see the leaked LeBron diss lmao

[Alex @Alexrcalz]

@ajambrosino we soon will be able to do whatever we want

[Alex @Alexrcalz]

@adcock_brett @hark_labs when are you hiring security positions?

[Alex @Alexrcalz]

@viemccoy I’d like to speak with Scully bc I’m not reading all that

[Alex @Alexrcalz]

@JoshKale is this a website?

[Alex @Alexrcalz]

This protocol is strongest against:
- known phishing and malware landing pages
- suspicious login flows on unfamiliar domains
- domains with existing threat-intelligence coverage
- repeated investigations where local policy can short-circuit the obvious cases

[Alex @Alexrcalz]

AI agents are getting scary powerful… but most still open random URLs with zero safety checks.
Phishing, malware, token drains — one bad link and it’s over.

Built a lightweight pre-browsing gate called Safe Web Confidence Protocol.

Before any agent loads a page it:
• Checks local allow/block lists (instant)
• Hits VirusTotal, Google Safe Browsing, urlscan, OpenPhish, OTX
• Returns a clear score + verdict: allow / sandbox / block

Fully explainable, cheap, and open source.
If you're building agents, this should probably be in your stack. https://t.co/CHikDKnMal

What safety measures do you actually use for your agents right now? 👇
#AIAgents #AISafety #OpenSource #Cybersecurity

[Alex @Alexrcalz]

Here’s the full protocol flow 👇

[Alex @Alexrcalz]

@adcock_brett Curious: how will the police be able to identify the owner of said robot if they need to contact its owner?

[Alex @Alexrcalz]

@rihim_s @viemccoy I’d say so, from my understanding llama excels at detection while nemo is a more complete system that excels at prevention + recovery

[Alex @Alexrcalz]

@rihim_s @viemccoy open-source tools like NVIDIA NeMo Guardrails is a great place to start.

https://t.co/A6Iv4IAUBV?
sortBy=developer_learning_library%2Fsort%2Ffeatured_in.nemo_guardrails%3Adesc%2Ctitle%3Aasc

[Alex @Alexrcalz]

@skupor @USTechForce When will applications open?

[Alex @Alexrcalz]

@beffjezos @elonmusk @pmarca Just gotta put trust in Moore’s Law