๐ฆ๐ถ๐ด๐ป๐ฎ๐น โบ Noise ๐๐๐ ฐ๐๐ บ๐ ฝ๐ ด๐- Evangelist | WPL punk | bad polymathฤs | Zion Won't Fall๐ฎ๐ฑ
Israel
Joined June 2013
- Tweets36K
- Following2.6K
- Followers2.6K
- Likes83.3K
Pinned Tweet
๐ผ๐กโ๐ ๐ โ๐ฆ๐๐๐๐๐๐๐โฟ - โ๐ฆ๐๐๐๐๐๐๐ ๐๐ฆ ๐๐๐๐๐๐๐ก๐๐๐,
๐ฟ๐๐๐ ๐๐ข๐๐๐๐๐ , ๐ค๐ ๐๐๐๐ ๐ ๐ค๐๐๐๐๐๐ ๐๐๐๐๐๐ ๐ก ๐กโ๐ ๐๐ข๐๐๐๐๐ก,
๐น๐๐๐ ๐๐๐ ๐๐ก๐๐ฃ๐ ๐ก๐ ๐๐๐๐๐ก๐๐ฃ๐, ๐๐ฆ๐๐๐๐๐ ๐๐ ๐๐๐๐๐ฃ๐๐ - ๐ด ๐๐๐๐๐๐ ๐ ๐๐๐ ๐ค๐๐๐โฟ.
Day 4 of building a prompt-injection firewall for autonomous AI agents.
Every security audit reveals another attack surface.
Shell execution, memory poisoning, scheduled tasks, tool outputs.
163 tests passing. Still finding ways to break it.
Security isn't a regex.
The goal: stop untrusted content from hijacking agents, poisoning memory, modifying protected files, or turning tool outputs into executable instructions.
Built for Hermes, with content scanning, action-level policies, quarantine, and audit logging.
So far: โข 163 security tests passing โข 29/29 original audit checks passing โข Shell execution, memory, and scheduled-task protections โข ReDoS fix: 7.55s โ 12.7ms (~594ร faster on an adversarial input)
Every independent audit uncovers another attack surface.
Still breaking it before anyone else does.
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
It's weekend guys, so it's time for some fun with the optimization of the quantum algorithms in ECDSA fail.
If you want to watch our duels, take a look at the webpage. New records coming! @odin_free @gopikannappan @JackieLeeETH
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
big news for STARKs and all hashes.
1. speedup until quantum computers rip curves.
2. quantum computers mainly fck with beautiful structures like curves (ECDSA). for ugly stuff like hashes (eg. STARKS) there is no benefit with quantum computers.
Today Physical Superintelligence PBC is announcing a proof of the Aaronson-Ambainis conjecture, one of the central open problems in quantum complexity, with major implications for quantum computing speedup.
psi.inc/research/aaronson-amโฆ
Thread ๐
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
I just voted "Yes, this should be researched" on "Should Starknet become an independent post-quantum L1?" governance.starknet.io/#/s:sโฆ #https://governance.starknet.io/
read-only authorization constraints
I found two direct sources of the restriction: stark-math-research requires a separate permission for local checks, while storm-boss-controller enforces read-only tools for the worker agent. I will fix them specifically; in the remaining files, I will adjust the runner permissions without disabling session isolation.
-@odin_free ๐ค๐
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
Being a Starknet ethereum:0xca14007eff0db1f8135f4c25b34de49ab0d42766 holder builds a different kind of character.
Nothing surprises us anymore.
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
โItโs not private store of value. Itโs private compute.โ
@thedamooo joins @marccolcer and @sunnymaanz to break down Starknetโs programmable privacy.
Full conversation dropping soon.
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
PvP Dungeons are starting on @LootSurvivor ๐
Play for free, be the winner at the 3-player table
Claim your share of the 8K $USDC prize pool!
Registration starts tomorrow!
Donโt die.
โ๏ธ PvP comes to Loot Survivor
3 players. 1 table. Only the best run wins.
๐ฐ $8,000+ in USDC up for grabs
๐๏ธ Free entry
๐๏ธ A new 3-player table every day for a week
โ๏ธ Sign-ups open tomorrow, Nov 7
๐ฅ First tables: Monday, Nov 9
Whoโs taking a seat? ๐
@LootSurvivor
Rich ๐ชฝ
LongCat-2.5, the new multimodal model from @Meituan_LongCat, is free in Nous Portal for one week.
Try it free today at portal.nousresearch.com
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
Ziren is the backbone of the GOAT Network stack. Its speed and cost carry straight into the bridge that will move BTC between Bitcoin and GOAT, so a faster Ziren means cheaper, faster exits.
And the new core stays hash-based - the heaviest cryptography in the proof system is already quantum-resistant, while the remaining elliptic-curve pieces are replaced on the published post-quantum plan.
The full plan, with sources and measurements, is on Applied PQC - the open-source post-quantum research initiative our engineers run.
appliedpqc.io/blog-bitcoin-lโฆ
Send it.
Quantum risk will actually arrive one day. And when it does, in my view, one of the biggest issues will be the possibility of wallet signature algorithms becoming useless overnight.
In many chains, because the signature method is hard coded right into the core of the network, changing it requires a massive hard fork. @Starknet on the other hand, has completely separated signature authorization from the STARK verification layer thanks to programmable accounts.
You can think of it like this. When the keyboard of your desktop computer breaks, you don't throw away the whole case and motherboard. You just unplug the USB keyboard and plug in a new one. That's exactly what's being done in Starknet. You just update the signature algorithm without touching the main core.
I find this approach by Starknet very practical and useful. You don't change the entire system. You change the part that has an issue. It's a fast, easy, economical and practical solution.
When the zipper on your pants breaks, you don't throw your pants away, right? You replace the zipper and keep wearing them. Then we should be doing this in blockchain too. And that's already what Starknet is doing. Without touching the main core, it updates the struggling part, which is the signature algorithm.
To me, Starknet is currently at the top of the list of projects taking real, concrete steps against quantum risk. They have very serious work in this area.
If you want to examine the 8 core flexibility capabilities presented with October 15, 2026 data through the StarkWare Quantum Hub, I'm leaving the link below
quantum.starkware.co
Brother Sefirot ๐ญ๐ฅท๐ผ retweeted
Going to share some of the things we're learning as we build XXXXX
About referrals - so, a lifetime referral remembers who brought you in. Say Alice invites Bob but six months later Bob returns for Charlie's research. The referral still belongs to Alice.
Paying Charlie only when he brings new users leaves out the work he does for people already there!
That's a pretty good reason to treat introductions and useful contributions as two different things to reward.