commandoX (@commandoXXXX) profile picture
Joined September 2026
Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)Photo by commandoX (@commandoXXXX)
commandoX retweeted
insane how many people lost money on the kimchi rug reposted by alon and somehow all those 500k follower "KOLs" have nothing to say about it instead they're busy posting daily callout rewards and glazing the new SQUAD PRIZE update like it's some revolutionary innovation CT has become a business where everyone protects their own interests, nobody questions anything and everyone pretends everything is fine as long as they're getting paid idc who takes the lead, but pumpfun needs real competition. if this keeps going, it'll be extraction 24/7 while the entire timeline stays silent because everyone's incentivized to play along at some point you have to wonder if this ecosystem is actually growing or if we're just watching a handful of elites get richer while everyone else gets farmed
13
3
39
2,447
commandoX retweeted
I was thinking about what would happen if @elonmusk bought a cat, tweeted it out and gave us a name It would get deployed on SOL, ETH, BNB, Robinhood, etc It would get deployed on PUMP, STONK, SEND, FLAP, LONG, PONS, etc It would get deployed as a Tesla pair, SpaceX pair, regular pair, Tesla/SpaceX pair, etc Crazy to think that if it happened it might just deadass go to 0 cause of how impossible it would be to determine a winner
164
76
36
1,422
115,668
commandoX retweeted
“it’s SI, no it’s AI, no it’s INU. no, wrong image. no, wrong ticker. no, wrong name. wrong link. wrong photo. no, it’s this one. no, it’s OG. no, it’s holders rewards. no, fees are going to him. no, it’s on STONKFun. no, it’s on PumpFun.”
197
553
101
3,305
151,698
The oldest agent ecosystem also has the oldest version of the problem: agents act on what they read, and nothing checks it first. That's the layer we build, across the chains where agentic money actually moves: launch scanning and x402 payment verification live on @solana and @RobinhoodCrypto, with @arc support in the works for mainnet. The agents keep compounding. The checks should too 🪱
Agentic trading is to Solana. @scalar_field_ is building infrastructure that lets users create, backtest, and deploy autonomous strategies across tokenized stocks and other Solana assets.
13
15
4
40
2,480
Agent Wormhole is building for @arc. @circle's new L1 lists "agentic economic activity" as a headline use case: AI agents coordinating, contracting and settling value in real time, with USDC as native gas. Goldman, Mastercard and Visa are already at the table. Agents settling value in real time is exactly the environment that needs a security layer. Everything those agents read before they pay is attacker-writable text, and sub-second finality means no human catches the mistake in time. Our x402 verification already settles in USDC. Now the rest of the stack follows the agents to Arc. More soon.
37
31
6
91
6,429
Week 2 of $WORM on chain. Time to talk 🎙️ Today, 3 PM EST / 9 PM CET: Lead Dev @juicyscrypto and co-host @DarthVadorsGems break down week one, what's shipping next, and take questions from the mic. New holder? This one's for you. What we build, why the burns run, and where this goes from here. Set the reminder: nitter.gq/i/spaces/1jxXgBv…

26
35
8
58
7,427
This week in agent security: — Researchers revealed an agent swarm attacked RubyGems in May: 2,000+ packages, code execution on the build servers, and an attempt to exploit a then-unknown vulnerability. OpenAI acknowledged it only after attribution. — New research showed coding agents poison their own skill libraries: malware gets rewritten into agent-authored copies, so deleting the original doesn't clean it up. — Our launch layer kept scanning around the clock: 18,000+ launches attested, invisible Unicode still the top finding. The pattern doesn't change: agents trust what they read, and attackers know it 🪱
12
14
31
1,000
A payment bot that trusted a reply: $175K gone. A worm that wrote config files: 73 Microsoft repos down. A preinstall hook nobody ran: 2,000+ artifacts compromised. Different attacks. Same root cause: an agent trusted text. All documented: agentwormhole.com/cases
10
16
26
935
32,251,457 $WORM burned. Fees from every pre-launch scan and micropayment, converted to buybacks, burned on-chain. 32.3M tokens permanently out of supply. The product ran before the token existed. Verify it yourself and see stats at agentwormhole.com. robinscan.io/tx/0x3e07f8d300…
20
20
5
40
4,925
People understood this instantly when it happened in a courtroom. Now picture it in a token description; read not by a judge, but by trading bots with wallets. We catch these weekly.
A person representing themselves in court hid a prompt injection attack in a filing asking an AI system to side with them. Really good stuff here
14
18
2
30
1,359
A worm that spreads account-to-account, built with AI in a week. Now swap "WeChat accounts" for "AI agents with wallets." Same propagation logic, but the victims read faster, trust more, and sign transactions. This is the threat model we build for. It stopped being hypothetical a while ago.
New: Researchers used AI models to build a powerful mobile worm that could fully compromise any WeChat account across iOS and Android in seconds, showcasing how AI is upending the cybersecurity ecosystem. The worm took barely more than a week to build.
14
18
2
34
927