Founding CEO of OmniLoop, Master in Media and Technology Management, Digital Solution Architect, Loving Life.
Bremen, Deutschland
Joined October 2009
- Tweets910
- Following507
- Followers119
- Likes1.4K
Christoph Müller-Dechent retweeted
The Binance $5.6B drain is a masterclass in modern EVM exploitation.
It wasn't a complex logic bug. It was an EIP-7702 delegation hijack:
1. Setup: Attacker funds 0x7c86 via Tornado Cash, deploys a sweep contract (0x63e8), and runs three live dry-runs on external wallets.
2. Execution: Attacker submits an EIP-7702 batch with 50 pre-signed authorizations (nonce n to n+49) to hit Binance 14's moving nonce window.
3. One authorization hits. Instantly, Binance 14 (an EOA) is reprogrammed to run the attacker's contract bytecode.
4. Execution looks totally organic: Binance 14 sweeps itself clean. Internal threshold alerts don't trigger because, from the nodes' perspective, the hot wallet is just executing its own transaction.
5. Binance 16 immediately auto-refills 28,000 ETH due to automated treasury rebalance scripts, feeding the exploit loop.
The cold truth: you cannot forge an EIP-7702 delegation list.
For this to execute, the attacker already held the raw private key or unconstrained signing access to Binance 14's HSM/KMS.
But hey, 5.64 billion drained across ETH, BTC, and Tron. 2.54 billion frozen by Tether, Circle, Ripple, and World Liberty within hours.
Code is law until the stablecoin admin keys log in.
Binance got bailed out by the very centralized choke points the cypherpunks warned you about.
Hardware wallets leak your home address.
Centralized exchanges leak the keys to the kingdom.
Self-custody on native, non-delegated chains remains the only honest baseline.
At least you know what to avoid.
Christoph Müller-Dechent retweeted
Bitcoin Core v32.1 was released!
It is available from: bitcoincore.org/en/download/
Release notes: bitcoincore.org/en/releases/…
Christoph Müller-Dechent retweeted
A brutal year for Jason Calacanis is just getting started.
Give me three minutes to explain why you should dump him now.
Outflows from the All-In audience are a signal and will continue.
We will test the “no one cares what he thinks” low again in the next 12 months.
Why?
Because there are far too many dynamic places to put your attention that are creating MUCH more value.
Why put your time into a dead take-machine that offers nothing beyond recycled 2014 startup wisdom?
What’s the point of placing Jason Calacanis bets if his contrarian-for-the-sake-of-it nature has been right one time, with Uber (and that's now failing).
The viral nature of Jason was his friends... if you didn’t follow him you were MOCKED!
Now? Completely reasonable, heck, SMARTER, to ignore Jason and place other bets!
No one is trying to convince people to listen to Jason anymore... the hot takes and “I told you so” are gone.
No one is saying JCal can change the world, because he won’t — and BECAUSE AI HAS!
Just shift from Jason to Super Intelligence because one provides little to no value, and the other provides unlimited upside for humanity.
This should be obvious to everyone right now.
Jason will never break his all-time high again.
Christoph Müller-Dechent retweeted
Claude Code just found proof that every app on your phone is listening to your conversations. TikTok, Instagram, Facebook, Snapchat. they're using what you say out loud to build your feed. i have the data to prove it.
i've suspected this for months but last week something happened that made me finally check.
i was on a call with my friend and told him i'm thinking about breaking up with my girlfriend. didn't text it. didn't search it. just said it out loud with my phone on the table. two hours later she opens TikTok next to me and her entire feed is breakups. "signs your relationship is over." "how to heal after a breakup." she looked at me and said "why is TikTok showing me this." i said i don't know. but i did know.
i connected my iPhone to my MacBook and ran Claude Code that same night.
it scanned every app's background microphone activity for the last 90 days. not just the apps that asked for microphone permission. apps that were supposed to only have access to photos or location but were activating the microphone anyway.
what it found:
→ TikTok activated the microphone 347 times while the app was "closed" in the background
→ Instagram → 211 background activations
→ Facebook → 189
→ Snapchat → 94
every activation lasted between 3 and 15 seconds. enough to catch a sentence. every one of them tagged a keyword and cached it locally on the phone with a timestamp.
Claude Code pulled the full keyword cache:
→ "breakup" — tagged 4 hours before her feed changed
→ "moving out" — from a conversation with my mom last Tuesday
→ "wedding venue" — from a call with my sister about her engagement that i never searched once
→ "protein powder" — said it out loud to my roommate, had ads for it within an hour
→ "flight to tokyo" — mentioned it at dinner, saw travel ads the same night
those keywords were sitting in a local cache on my phone organized by app with timestamps. TikTok had the most entries. Instagram was second. every single keyword matched a conversation i had out loud near my phone.
the scariest one was "wedding venue." my sister told me she's getting engaged on a private call. i never searched it. never typed it. within a day my explore page had engagement rings and venue tours.
then Claude Code checked a second phone in the same room. my girlfriend's. same keywords. same timestamps. both phones were picking up the same conversation and tagging it independently.
i asked Claude Code how to shut it down. it mapped every app's microphone access and built a permission audit.
what i did:
→ revoked microphone access from every app that doesn't need it to function
→ disabled background app refresh for TikTok, Instagram, Facebook, Snapchat
→ turned off "Hey Siri" which keeps the microphone in passive listening mode 24/7
→ revoked "nearby interaction" permission which uses ultrasonic audio between devices
Claude Code confirmed all four apps lost background microphone access after the changes. i checked her phone too. same steps. took about 3 minutes per phone.
within 48 hours both our feeds went completely random. generic content. nothing personal. nothing from our conversations. that's when i knew for sure it was real.
the full step by step and the prompt are in the first reply. takes 3 minutes. save it so you don't lose it and send it to anyone who's ever said "my phone is definitely listening to me." it is. now you can prove it and shut it down.
Readers added context they thought people might want to know
2 days before, OP claimed to have found spyware on his "wife's phone":
x.com/starmexxx/stat…
Now, in this post, he talks about his "girlfriend":
x.com/starmexxx/stat…
OP has a history of fake AI content:
"Spy camera" found in hotel room: x.com/starmexxx/stat…
Spyware on "friend's computer": x.com/starmexxx/stat…
OP selling AI products: x.com/starmexxx/stat…
Flagged as AI: x.com/i/grok/share/b…
Christoph Müller-Dechent retweeted
You are just a day away from shipping slop coded convenience that others don’t know how to create themselves with a basic AI subscription. This window probably will last another 2-3 years.
Christoph Müller-Dechent retweeted
Andrej Karpathy spent 8 years at OpenAI and Tesla
Last week he compressed everything he knows into one free 2-hour lecture
Prompts → Agents → Harness → Graphs → Loops
People spend $15k on bootcamps that teach less than this
You probably don't have 2 hours right now
Don't let it vanish from your feed
Watch it, then read the article below ↓
Christoph Müller-Dechent retweeted
for the first time in a long time i can feel my mania coming back. in the best way possible. my hunger for life. excitement for building things. vision for what they could become. i feel myself regaining the child like wonder i had lost for so so long
life. what a lovely thing..
Christoph Müller-Dechent retweeted
This is your sign to get off your phone and go live. @ChrisWillx
Christoph Müller-Dechent retweeted
网景创始人 、硅谷顶级投资人Marc Andreessen 在访谈里坦言，他看到了科技史上最可怕的一道认知鸿沟，
传统巨头以为 AI 只是个降本增效的插件，而新一代 AI 原生团队，正在用完全非人的组织效率掀翻整个商业版图。
当年写出网景浏览器、见证了整个互联网三十年兴衰的 Marc Andreessen，坐在 Ben Horowitz 面前直接发出了感慨，
如果 1994 年我们做 Netscape 的时候就有今天这种 AI 工具，那简直不可想象。
现在绝大多数传统巨头和现有企业，依然在用改良软件的旧思维去看待大模型。
在他们的组织逻辑里，AI 只是一个高级功能，用来给现有业务做个插件、在钉钉里加个智能助手、或者把客服团队砍掉 20% 省点成本。
他们以为自己在拥抱变革，其实只是在给老旧的蒸汽机打补丁。
但真正恐怖的事情，正在那些刚成立的 AI 原生团队里发生。
那些创始人对 AI 能力的理解，跟大厂完全不在同一个平行宇宙。
他们每周都在看着底层技术飞速进化，而且他们压根不会把自己定义成一家软件公司、硬件公司、甚至是一家 AI 公司。
以前创办一家公司，你的组织架构是按人类的协同瓶颈搭出来的：
招产品经理写文档、招工程师写代码、招法务改合同、招运营做投放。
因为人类的脑力带宽有限，所以必须靠庞大的部门和管理层级来支撑业务。
而真正的 AI 原生物种，是用机器智能作为最底层的骨架。
两个人加上一整套自动化 Agent，可以直接把从市场洞察、软件工程、全栈设计到商业销售的链路全部跑通。
他们不是在效率上快了 30%，他们是在组织形态上彻底跨越了一个物种。
大厂的盲区，从来不是买不起算力，而是他们无法理解组织本身才是最大的负债。
当现有企业还在开会讨论如何把 AI 塞进现有的业务流程时，新一代的野蛮人正在用完全超越行业界限的方式，把传统的业务流程连根拔起。
未来拉开人与人、公司与公司之间差距的，根本不是你用了哪个模型，而是你到底是在用新工具维持旧世界的体面，还是用新世界的底层逻辑，重新定义了自己能做的事情。
Christoph Müller-Dechent retweeted
@start9labs is criminally underrated.
Their domain setup + any DNS provider (like DuckDNS) is a total game changer. Point any service to your domain and boom. Automatic Let’s Encrypt certificate with zero hassle.
This level of simplicity for self-hosting is not shilled enough.
Christoph Müller-Dechent retweeted
Your pre-frontal cortex is the harness for the rest of your brain
Meta-cognition is like a self-improving harness – at San Francisco, CA
Christoph Müller-Dechent retweeted
As a German living abroad and looking at things from outside the German media bubble, it’s tough to watch such a great nation hold itself back like this.
Christoph Müller-Dechent retweeted
OMG… the Tesla Roadster just went upside down on a half pipe and landed!
Christoph Müller-Dechent retweeted
"Why did they waste so much time making a flying car no one can use in the real world?"
Because not everything is about whether you can imagine it being useful to you. Because like art, sometimes we make things for the glory and challenge of making them. Because we don't know if what we'll learn will make your toaster work better and you'll never realize it because you take for granted that things get better without thinking through how that happens. Because you can't inspire people with toasters, but a barrel-rolling car that can drive upside down is absolutely setting someone on fire right now.
I'm not a car person and you shouldn't trust me with a rocket, but I want a roadster because I make art and that thing is absolutely a thing made for love and passion and I get that in my heart.
(Also because I'd look really cool in one.)🤣
Christoph Müller-Dechent retweeted
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
Judging by my tl there is a growing gap in understanding of AI capability.
The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code.
But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along.
So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions.
TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.