Post

OK, so bet. Someone sends you footage of the moon landing. You want to know if it's real. Turns out you're actually asking three different questions — and two of them are now solvable. 🧵 (1/5)

Has it been edited? Did it actually come from NASA — or someone pretending? Did the moon landing actually happen? We mash all three into "real" and argue forever. The first two now have mechanical answers. The third never will. That's not a hole — that's the whole point. (2/5)

Three results, live, against rmichaelthomas.com: ✅ real clip → ATTESTED & RECOGNIZED ❌ one byte changed → REJECTED ❌ signed by an impostor → ORIGIN UNCONFIRMED Each attack caught by a different check. Built on a couple of languages I made that compute what a file can do before it runs. (3/5)

A tool that claimed to prove truth would be worse than no tool at all — the checkmark would launder the fakes it was meant to catch. This clears away the answerable questions so the only one left is the honest one: do you trust the source? That was always the real question. (4/5)

No platform, no account, no authority in the middle. A person with a domain and a key can attest their own video and anyone can check it. (5/5) The longer write-up; thank you for rocking with me! rmichaelthomas.com/rmichaelthom...

rmichaelthomas.comThe Moon Landing ProblemYou get the moon landing footage. Was it tampered with? Did it really come from NASA? Did the landing even happen? Two of those you can now prove. One you never can — and that's the honest part.

1 repost1 like

i have been working on some stuff too for representing common transformation functions over byte segments so they're more robust, so you sign hash(f(blob)) instead of signing hash(blob)

Ah, ok, I see. OK I'll send you this doc for sure. I've been playing with polyglots and that's where some of this work came from and was inspired by.