deepfake clones your cfo, authorizes a wire, the bank's hsm cryptographically signs the theft. signature is valid. intent is fake.
verdico is client-side. secures the moment of authorization before the packet hits the rail.
banks solved double-spend. this solves double-speak.
i build air-gapped hardware that verifies wire transfers before they hit the banking rail.Microsoft's VALL-E 2 clones any voice from 3 seconds of audio. someone deepfakes your CFO on a video call, authorizes a $50M wire, the bank sends it. Arup lost $25.6M this way in February 2024. the phone callback is dead.detection is a losing arms race. so: physics.credit-card-sized device. no wifi, no...
It is fascinating to see the evolution of the 'threat signature,' though. You were flagged for proximity to literal nuclear physics in 1998; I was flagged for a vibe-coded prototype that just looked like physics in 2026.
I have a sinking feeling that, like you, I've just unlocked the 'secondary screening' achievement for the next decade. I’ll take the permanent Swiss police record over the Five Eyes watchlist any day. You win.
To be precise with my terminology: I showed the forensic expert the terminal history and compiler output in my VS Code/Cursor logs.
Because I was 'vibe coding' with LLMs, I had a long scrollback of cargo build failing repeatedly with ownership/borrow errors. 'Chris' (the forensic expert) reviewed that timestamped history to verify that I was genuinely struggling to compile a harmless display driver in a hotel room that morning, rather than deploying a pre-compiled malicious payload.
His logic was essentially: 'A terrorist brings a clean binary. A developer brings a terminal full of red text.' The broken build state was my alibi.
I’ve walked around SF and NYC with prototypes for years without issue. I genuinely failed to code-switch for the environment. I expected a bag search; I didn't expect a 13-hour detention and a forensic code audit.
It felt like 'pantomime' at first, but once the forensic team arrived, it became very real. They weren't performing theater; they were genuinely verifying the interrupt handlers in the code.
That said, your point is the correct one. In a normal setting, leaving a bag near your table while you check in is fine. In a high-security zone during WEF, placing a black box with wires on a bench is a massive red flag. I treated it like a backpack; they treated it like a potential threat. It was a hard lesson in situational awareness.
They were incredibly professional. Once we moved past the initial 'threat assessment' phase, the officers and the forensic expert were fair, logical, and treated me well. The system worked exactly as it should for a suspicious package in a high-security zone.
You are absolutely right about the optics (and the blinkin' lights). The forensic expert made the same point—unintended drama is still drama. I definitely didn't aim for this (just wanted to demo the hardware), but once the process started, I had to respect the thoroughness. Their protocols are no joke.
That said, the scrutiny shifted quickly from the 'black box' visual to the actual tech. It wasn't just a physical inspection; they brought in a forensic expert to audit the Rust borrow checker logs and interrupt handlers to verify the triggers were benign. It turned into a very technical interview, just with higher stakes.
To get released, I had to walk a forensic expert ('Chris') through this codebase line-by-line. He didn't care about the pitch; he audited the Rust borrow checker logs, the specific hardware interrupts, and the encryption implementation to prove it wasn't a trigger mechanism.
It was the most aggressive code audit of my life. Happy to answer questions about the stack, the 'vibe coding' workflow I used to build it, or the Swiss prison lasagna."
I'm planning to index and store 1.5 trillion tokens using Faiss and would love some feedback on my approach:1. Partitioning: I'm thinking of using distributed k-means and inverted multi-index quantizers for efficient data partitioning.2. On-Disk Storage: Due to the scale, I'm storing everything on disk using a Compressed Sparse Row format.3. Distributed Search: I plan to implement a...
Check for yourself: https://huggingface.co/openai-detector
No one can blame the drug companies for operating as an institution and it's hard to pin that blame on the individuals operating in it, too. You're right that the incentives that form these companies haven't been around forever and we should seek to find ways towards a better solution.