@Teslarati Have 2 teslas, one 2024 and 2019, both run Cabin Overheat Protection 24/7 in a very hot state. I think it used about 2-5% a day itโs turned on. Quite literally one of my favorite features.
Non-nerds are asking how Mr. Al-Qudsi (@mqudsi) is working to reconstruct redacted Epstein data. Here is a high-level summary that isn't as nerdy schizo
Mega tl;dr
> Send email
> Add attachment
> Emails no understand files
> Email turn files in text (Base64 encoding*)
> Image 1 is email turning attachment into text
> Send email
> Someone receive email
> Email reads add-on text
> "oh thats an attachment"
> Transforms into attachment you can see (Base64 decoding*)
> DoJ releases Epstein emails
> Didn't censor attachment stuff
> hehe big mistake, we can recover this
> Boom, all attachments "censored" now uncensored
> All hidden attachments now public
> Go to work
> Problems arise
> DoJ printed emails (???)
> Scanned printed emails back (???)
> Try to rebuild from email stuff
> Fails
> wtf.mp4
> Look inside
> DoJ printed as "Courier New" font
> L and 1 look the same
> Try to reconstruct
> Fails
> Computer can't figure difference between L and 1
> (Look at image 2)
> Can you even tell the difference???
To manually reconstruct all attachments from Epstein emails data forensic experts must find a way to programmatically determine which characters are L's and which are 1's. This is only a problem because the DoJ printed it as Courier New.
Proposed solution right now is bruteforce. Try every possible combination, swapping L's and 1's, check email thing, does it work? No? Repeat. However, this could take a long time.
Another solution is taking known email encoded thingies that work and compare it to Epstein files. Try to identify patterns and reconstruct it using machine learning.
Kimi Code has a similar plan, and you can use the plan for OpenClaw. I just set it up earlier today after burning through Openrouter credits as well using Kimi. Same model, just subscription based instead of credits. I also use OpenRouter still for subagents and have different models for different tasks but wanted a more cost effective way to keep Kimi.
@NathanMcNulty@FF_Freak Would you be willing to share what this policy looks like and maybe if it could be automated to be updated with threat intelligence for TLDโs?