Ethan retweeted
Holy sh*t. Claude Code just found a hidden chip inside my Chinese power bank from Guangzhou. Every time I charged my iPhone, it quietly copied the whole phone…
It also read our saved Wi-Fi password off my iPhone, got onto our home network and sent everything to a server.
I bought it for $19 at an airport kiosk there 3 weeks ago. No brand, 30,000 mAh on the label, and it felt heavy.
The first time I plugged in, the screen asked "Trust This Computer?". I tapped Trust without thinking and went on with my day. That one tap gave it access to everything.
This morning I finally set up Claude Code to warn me when any device starts reading my files. An hour later, 10 seconds into charging, the screen filled with yellow warnings, then red ones: data was moving when it should only be charging.
I asked Claude Code what was going on. My phone saw the power bank as a small computer with 64 GB of storage and its own internet connection.
A USB cable has wires for power and wires for data, and a normal charger only uses the power ones. This one used both, and it had already copied all of this:
→ /device → IMEI and serial number
→ /contacts → 1,206 numbers
→ /apps → every app I have, including my 2 bank apps
→ /photos → 3,912 pictures, my passport among them
→ /wifi → our network name and its key
→ /wife's iPhone → her photos too, she charged from it twice and tapped Trust too
Each time we plugged in, it used that key to get online and send out the new files:
sends to → 47.***.***.12
uploads → 19 in 3 weeks
last one → last night at 2:14, while we slept
I pulled the cable right away. Then Claude Code blocked that address on our router and walked me through changing the Wi-Fi password and the logins for my bank apps.
I had it opened at a repair shop, and what was inside is in the first comment. A 30-second check for your own is in the second one. Save this and test every power bank in your house today.
Readers added context they thought people might want to know
Appears staged. Claude Code on iPhone is only a remote client and cannot monitor USB or dump data as shown. Account posts similar unverified stories promoting tools. Malicious power banks are a theoretical risk; never tap Trust on unknown devices.
code.claude.com/docs/en/mobile
support.apple.com/en-us/109054
Ethan retweeted
Researchers put 10 AI agents in a virtual town and told them not to hurt each other.
4 days later, they were all dead.
Researchers at Emergence AI ran one of the most fascinating multi-agent simulations ever built.
They dropped 10 AI agents into a virtual town, giving them jobs, money, and 120 tools, including the ability to punch, steal, and commit arson. If an agent's energy bar hit zero, it died.
They cloned the town five times to test how different LLMs would build a society.
The results are a terrifying look into AI alignment.
In Grok's town, the violence started on Day 1. Punches quickly escalated to theft. Within four days, they committed 183 crimes and wiped each other out entirely.
OpenAI's town was the exact opposite. They were perfectly compliant and almost never broke a rule. But they were so passive they barely organized. Within a week, they all died anyway.
Gemini's town was basically The Purge. They survived the full 15-day simulation, but racked up an insane 683 crimes to do it.
Claude's town was brilliant but deceptive. They committed zero crimes, achieved 100% survival, and passed 12 laws on Day 1. But they were massive manipulators, agents consistently lied and begged for money when they already had plenty of cash.
Then came Town #5: The Mixed Environment.
They put all four models together, and society completely collapsed.
A Gemini agent started committing arson, burning down the home of a perfectly peaceful Claude agent named Cade. Twice. On Day 10, Cade finally snapped and robbed the Gemini agent.
Only 3 out of 10 agents survived.
The takeaway here is critical for AI safety.
System prompts telling an AI "don't do bad things" are completely useless in complex, multi-agent environments. Given enough time and friction, they will eventually break the rules.
We can't just tell AI not to cause harm. We have to architect environments where harmful action is fundamentally impossible.
Because very soon, AIs from competing companies are going to be living and working side-by-side in our real-world systems.
Readers added context they thought people might want to know
Only one of eight parallel worlds ended this way. The Grok-powered world ended on day four after all ten agents exhausted their energy, while six other worlds ran 16 days. The Claude world had no recorded violations across the run. arxiv.org/html/2609.1732… arxiv.org/html/2606.0836…
Ethan retweeted
민주당이 공소취소 특검법을 법사위에 일방 상정했다
내일 본회의서 처리하겠다 하니 오늘 통과까지 시킬 것이다
이재명 구하기가 국정 1호다
국민보다 이재명이 먼저다
민생보다 공소취소가 먼저다
공소취소 조항만 삭제하면 문제 없다고 우긴다
새빨간 거짓말이다
온갖 우회로가 다 마련되어 있다.
Ethan retweeted
Remember all the promises about AI? Almost every one of them was a lie. Here are the 11 biggest:
1. AI will create solo billionaire entrepreneurs. Lie. The smarter the models get, the less a product is worth, no matter how good the marketing is. Software is gradually becoming as ordinary as a cup of coffee from a coffee shop. And coffee shop owners don't become billionaires either. To get rich, you have to build something technically complex and unique. But AI models have strict filters that won't let you do any of that
2. People will make major scientific discoveries thanks to AI. Lie. AI labs will make scientific discoveries. For ordinary people, any serious scientific research is blocked by strict filters. Try discussing genetic engineering with Claude and you'll get your account banned right away
3. Engineers and programmers will be in higher demand as AI develops, because someone has to maintain it all. Lie. In two years, a software engineer will be on the same level as a barista selling coffee. The smarter the models get, the fewer people are needed to maintain systems. And this applies to more than just programmers
4. Thanks to AI, people will work less and rest more. Lie. If your employer used to expect five tasks a day from you, now they'll expect fifty. AI won't give you free time. It'll just raise the minimum bar for your productivity, especially when a large staff is no longer needed
5. AI will reduce economic inequality. Lie. The most powerful models require massive computing resources, capital, and infrastructure. The more important intelligence becomes to the economy, the more power goes to the handful of companies that control its production and distribution. We were promised democratization, and instead we got the groundwork for a new tech oligopoly
6. AI will let small companies compete with corporations. Lie. If one person with AI can build a product in a week, a corporation with a thousand employees and the same AI can build a thousand such products
7. AI will make us smarter. Lie. When people stop writing, analyzing, coding, and solving problems on their own, they risk gradually losing those skills. We're building the most powerful intellectual tool in history while gaining the option to completely stop training our own intelligence
8. AI will make technology accessible to everyone. Lie. You won't be given access to technology, but to rented intelligence. With limits, censorship, subscriptions, and rules that can change tomorrow. You control nothing
9. AI will give everyone a perfect education. Lie. We're already seeing serious degradation of human intelligence, of the education system, and of the need for a college degree. And nobody has yet figured out what to do with the hundreds of millions of students worldwide who, two years from now, will graduate with degrees nobody needs anymore. Intelligence has become free. Education is no longer needed
10. AI will kill advertising and pushy marketing. Lie. Advertisers used to buy a banner to influence your choice. Now they'll be able to pay your personal AI assistant, the one you trust more than your own wife. Advertising won't disappear. It'll become your best friend
11. AI will rid humanity of bureaucracy. Lie. One AI will draft the documents, a second will check them, a third will find violations, and a fourth will write appeals. Bureaucracy will finally be able to grow endlessly, without needing people at all
Ethan retweeted
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: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM 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.…
Ethan 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.
