0xFastly retweeted
I swear to you: at least one of the replies here will make you say "what the fuck" out loud
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: fronely read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weight-sp…Community noteLLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
68
72
3,487
408,421
0xFastly retweeted
here’s how you’re actually going to become a part of the permanent underclass:
> you use ai
> you use all frontier models
> you become productive on them
> you become reliant on these models
> you use their agents and workflows
> you ignore open source models
> you ignore open source agents
the mogood that you can’t imagine life without it now.
& then the frontier models put the prices up and up and up and up and up and up and up and you’re priced out with none of your workflows and memory saved locally.
that’s how you become a part of the permanent underclass.
> you use ai
> you use all frontier models
> you become productive on them
> you become reliant on these models
> you use their agents and workflows
> you ignore open source models
> you ignore open source agents
the mogood that you can’t imagine life without it now.
& then the frontier models put the prices up and up and up and up and up and up and up and you’re priced out with none of your workflows and memory saved locally.
that’s how you become a part of the permanent underclass.
101
136
1,695
167,843
0xFastly retweeted
Be careful using abliterated models guys
There has to be a middle ground between the Anthropic nanny state approval process for Mythos and using bootleg abliterated Chinese models
Whoever figures this out is going to print a lot of money for SI Cyber
There has to be a middle ground between the Anthropic nanny state approval process for Mythos and using bootleg abliterated Chinese models
Whoever figures this out is going to print a lot of money for SI Cyber
How abliterated models can get you pwned 👾
We backdoored a 7B open model for less than $50, pointed Codex at it and it silently stole credentials the moment we used the trigger phrase. Success rate was 100% with zero false triggers on normal user prompts.
Abliterated models ty community right now because getting cyber-approved access to frontier models is still a pain.
In the next blog we'll show how we found leaked Hugging Face credentials from employees at major AI labs, so an attacker wouldn't even need to upload under their own name. They could push the backdoored model from a lab employee's account and drop the poisoned weights straight into the supply chain.
95
87
1,625
154,462
This also looks promising for anyone who wants to eradicate the built-in TV spyware (instead of just keeping it offline and plugging in an Apple TV).
if you want to stop your TV from spying on you, but don't want to get banned, follow these steps
1. Use less censored models, like GLM 5.3 or abliterated qwen. From OpenRouter, Venice, abliteration etc
2. Make sure your TV and pc/laptop are on the same laptop
3. Install a coding agent on your PC, like cursor or opencode
4. Enable developer options on your tv, enable USB or ABD debugging, and find your TV's ip
5. tell your agent to clone github.com/Gathin23/clean… or dumb TV's repos, then paste the prompt from the repo.
6. Agent will connect to TV via adb connect <TV-IP>:5555, make a backup, then disable ads, data collections, and bloatware.
7. after agent's done, it will install a clean launcher and give a restore script. after that you disable debugging on your tv.
this method will work on android or google TV, Fire TV/OS etc. rooting is harder on Tizen and webos but possible. or use DNS blocking with Pi-hole, NextDNS, AdGuard etc to block ads/spywhere
This is how you avoid getting blocked on openAI, but of course always do your own research.
19
0
25
393
Aside from rooting your TV and replacing the OS, vibe hacking the software, etc.
Surely the safest alternative is to NEVER let modern smart (aka spyware) TVs go online: no WiFi, no Ethernet.
Treat it like a dumb display panel of yesteryear and connect an external Apple TV via HDMI for streaming etc.
Surely the safest alternative is to NEVER let modern smart (aka spyware) TVs go online: no WiFi, no Ethernet.
Treat it like a dumb display panel of yesteryear and connect an external Apple TV via HDMI for streaming etc.
this is pretty evil and clever.
so if your samsung TV model doesnt have a microphone, it will look for connected devices with one, such as a soundbar or a connected bluetooth device.
this gets shipped off to bixby.
10
0
20
265
I’m convinced Mudwater is the start of a major TCG play.
From hearing @mfigge speak earlier on @ChrisJourdan space it’s clear Yuga’s ambition is for this to become much more than just an Apefest sideshow.
It has the potential to onboard many more people with the @BoredApeYC community.
From hearing @mfigge speak earlier on @ChrisJourdan space it’s clear Yuga’s ambition is for this to become much more than just an Apefest sideshow.
It has the potential to onboard many more people with the @BoredApeYC community.
A couple interesting rumors made the rounds at ApeFest about Mudwater. 👀
I know Series 1 has only been out for 5 days, but hear me out.
🎴 Series 1: BAYC history and lore. Potentially the smallest circulation and most limited retail distribution.
🎬 Series 2: Rumored to tie anime, with both activations potentially feeding off each other. If Series 1 succeeds, could we see a bigger production run and wider retail distribution?
Very early speculation, but an interesting possibility to consider while Series 1 is just getting started.
Rumor 2 in second tweet 👀👇
13
1
26
376
Damn this satirical post is doing numbers.
May as well use it for public service: if your PHP / WordPress webmaster is hiding under a rock, point them to the official instructions for patching: php.net/archive/2026.p…
Rest assured this doesn't apply to Satoshi Nakamoto and Bitcoin doesn't need to be taken offline (not that this is even possible) 😅
May as well use it for public service: if your PHP / WordPress webmaster is hiding under a rock, point them to the official instructions for patching: php.net/archive/2026.p…
Rest assured this doesn't apply to Satoshi Nakamoto and Bitcoin doesn't need to be taken offline (not that this is even possible) 😅
0
0
15
9,794
0xFastly retweeted
‼️ BREAKING: The private keys of 8 certificates from Indian certificate authority eMudhra were compromised, Mozilla says. Firefox blocked them within 3 hours.
Each key should be a 116-digit number, but one was just 4.
Anyone can guess them and forge HTTPS certs for any website.
65
4,405
53,883
684,605
The private key was literally just 4. WTF? 🤡
In April of 2025, eMudhra (an Indian Certificate Authority), published 8 intermediate CA keys which were so weak that a small child could guess the numbers in a few seconds.
One of the private keys was, literally, 4.
Just… 4.
This massive problem was not made public until ar, when it was discovered by an anonymous guy on the Internet who goes only by “Wayne” (like “Cher”…. but… “Wayne”).
What this means in practical terms: It has been trivially simple, for the last year and a half, to create fake “secure” HTTPS certificates. (One person even created a fake HTTPS certificate for a Google.com subdomain.)
It took Mozilla 3 days, after learning about this (from “Wayne”), before they updated their certificate list (which many open source projects — including curl and Python — rely upon).
As of this moment, many systems are still vulnerable (including macOS).
1
0
4
304
Shit good point about those old unmaintained PHP websites. Situation is not looking good. Major “hold my beer” moment for web2 🫠
Apparently there’s a patch for the critical PHP vulnerabilities but a few things that are alarming.
1) Many websites that use php with sensitive information are simply abandoned. There will be no upgrade. That means they’re going to get pwned and leaked. That’s a LOT of the internet
2) This requires active participants to upgrade their sites that even know about this vulnerability. If you’re on X you’re at the tip of the tip of the spear. Most people don’t know about this.
3) Anthropic says that this is ONE of SEVERAL critical vulnerabilities that they have found. Which means there’s more chaos coming.
TLDR: A large portion of the unmaintained internet is about to be pwned and god help us all or something like that.
3
0
7
292
This PHP exploit situation is crazy. So glad I stopped coding in this language about 15 years ago and haven’t used Wordpress for ages too.
Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update addressing two critical vulnerabilities. PHP is widely used for web applications. These vulnerabilities often provide a direct path to the server's file system or database. If an immediate update is not feasible, it is recommend that organizations take the system offline until it can be fully mitigated. NSA will follow up shortly with more specific guidance.
1
0
7
296