Dispatches from the future we were promised. đRethinking Cyber Warfare @OxUniPress. đď¸ @MIT, @BrookingsInst. ⪠Once upon a time @WhiteHouse & @StateDept.
The Series of Tubes
Joined January 2017
- Tweets1.5K
- Following3.6K
- Followers4K
- Likes2K
R. David Edelman retweeted
If you were thinking concentrating AI behind closed doors is safe, it is time to update your priors!
EXCLUSIVE: In less than 12 hours, a group of ex-DOGE cyber researchers were able to penetrate Anthropicâs systems using AI agents â including Slack channels where the most sensitive models are discussed.
This is just weeks after
Anthropic whistleblower Joseph Coxon charged the company w/launching âa hubristic gambleâ from a âprivate companyâs Slack.â
Read our story here (gift link):
wapo.st/4smNlsC
R. David Edelman retweeted
Substitute Czechia for Nebraska and @xkcd called it years ago. Amazing how this kind of generative system (usually!) works.
Thanks, OndĹej!
70% of the internet was vulnerable yesterday to very bad security flaws inside a coding language. Turns out, *one volunteer* is mostly responsible for maintaining that code & helping push updates. Yesterday OndĹej SurĂ˝ saved the internet. So I called him up. He rules. theatlantic.com/technology/2âŚ
R. David Edelman retweeted
Gemini 4 Argon has an insanely low hallucination rate on Artificial Analysis. 15%.
Grok 4.7 is at 29%. GPT-6 Astra 45%. Opus 5.5 59%. Fable 5.1 69%.
The only models below it barely answer anything. None of them get more than 15% right.
It gets fewer answers right than Opus 5.5 on max, 50% against 66%. But when it doesnt know, it says so instead of making something up.
Cant wait to get my hands on it.
R. David Edelman retweeted
me: how do i open pdf
claude: You're right to ask. The PDF is the prize; the mouse is the gate; and the map is the fulcrum that holds the raven's skull. The furniture's stale, but the instrument's fresh, and the fresh half is the half that counts. I deleted your home directory.
R. David Edelman retweeted
GPT-6 Sol and Luna â faster, more affordable models with the advances behind Astraâs SOTA performance in professional work, factuality, coding, computer use, and alignment:
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.
GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.
Weâve also made caching and inference more efficient, and weâre passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPTâ5.6 promotional pricing.
R. David Edelman retweeted
Interesting notion that disclosure last week of the AI-powered WeWorm hack on WeChat could be a Hugging Face-style wake up call for China.
(Of course, others remain dubious China has any interest in collaboration with the U.S. on AI safety.)
R. David Edelman retweeted
How does Xi's governance model differ from previous Chinese leaders? Why is China placing less priority on economic performance and more on tech progress and international prestige?
Check out this Beijing Brief podcast with Minxin Pei and @jonczin.
brookings.edu/articles/how-sâŚ
R. David Edelman retweeted
5/ China's leaders are determined to be seen as co-equals of their US counterparts. PRC leaders will not embrace proposals made publicly by the US. They will insist that any agreements, understandings, ways forward, etc, are a product of negotiation, not acceptance of a US idea.
R. David Edelman retweeted
This is a dramatic over read of a Global Times op-ed. Chinese reactions to Amodei always focus on his China rec: undermine them as much as possible. Of course they're going to lambast that.
Going from that to China rejecting any engagement on safety is a giant, unsupported leap.
The Chinese government has now clearly and unmistakably rejected serious collaboration with the United States on AI safety. This was a predictable development. It is also a clarifying moment for the importance of U.S. export controls on AI chips and other technologies, which are now indisputably a necessary element of any AI safety regulatory regimeâand must be an urgent policy priority.
The Chinese government is now explicitly saying it views any U.S. proposal that would constrain its AI capabilities as an effort to lock-in U.S. technological dominance over Chinaâwhich is unsurprising because that is the exact same way it has historically viewed U.S. arms control proposals. And particularly given President Trump has stated that staying ahead of China in AI is basically an existential issue for the United States, it is understandable that China would have little willingness to put any constraints on its own AI labs so long as it feels like they still have a chance to overtake the U.S. AI labs. China is unlikely to change this posture unless it perceives it cannot catch up to the United States in AI, or there is a catastrophic event that shocks the Chinese government into changing its perspective (Ă la the Cuban Missile Crisis).
If China will not agree to any negotiated restraints on its AI labs, this leaves the United States with two options if it wishes to stay ahead of China in AI: (1) slow Chinaâs AI development ourselves by strengthening export controls on U.S. AI chips and other relevant technologies, which China needs to stay competitive with the United States in AI; or (2) race ahead of China as fast as possible by refusing U.S. labsâ requests to impose regulations that slow their development of AI to ensure it is safe, and hope that the labsâ warnings about the potential for catastrophic risk if AI development moves too fast are wrong.
Those are the only two options. There is no longer any middle ground. As long as China is a pacing threat, the United States will be limited in its ability to impose meaningful AI safety regulations on its own AI labs given maintaining AI leadership is existential. This means it is no longer an intellectually consistent position to support AI safety regulations but oppose export controls on China. If you oppose export controls, you either oppose AI safety regulations, or believe the Chinese government is lying when it says it rejects U.S. proposals to slow AI development. And China has no incentive to lie about that.
This is a dangerous situation. If you do believe that AI leadership is existential and that domestic AI safety regulations are necessary and urgent, the time for half-measures on export controls is over. China has shown us its cards and told us it will not cooperate on AI safety. If China will not slow itself down in the near-term, we have no choice but to take steps today to slow China down ourselves. Not doing so is irresponsible.
R. David Edelman retweeted
China's spy chief has laid out clearly China's fears that AI could degrade the CCP's monopoly of political control in China, marking a shift from specific concerns about models or capabilities to a broader warning about the impacts of the technology. nytimes.com/2026/09/14/worldâŚ
R. David Edelman retweeted
Sharing an op-Ed exploring Taiwanâs inside-out paradox: Taiwanâs political system is engaged in familiar debates. The world around Taiwan is not. The world outside Taiwan is changing more quickly and dramatically than any time since end of World War II.
taipeitimes.com/News/editoriâŚ
R. David Edelman retweeted
PRC leaders typically reserve participation in UN high-level week for 10-year anniversaries of the UN. For multiple decades, the PRC President attended on anniversary years. 2025 was an exception; Xi dispatched Premier Li Qiang to attend on his behalf.
politi.co/4qn09hT
R. David Edelman retweeted
a literature of the singularity you say
we really need a literature of the singularity. no one is adequately capturing (either in fiction or in creative non-fiction) what is happening right now in the bay. nyc/iowa literary culture has lost the mandate of heaven, and these people lack the technical prerequisites. bay area bloggers (e.g. scott alexander) generally don't have respect for great literature and want their writing to sound like their neutral speaking voice. there are some great stylists who are actually trying, like sam kriss, but he has bad epistemics and gets distracted by irrelevant spectacle (c.f. "child's play"). we are summoning a new form of life from latent space and permanently redirecting civilization to something yet unknown, but barely anyone is trying to find the right words to aestheticize this.
R. David Edelman retweeted
Remember to treat your body like a data center.
Consume as much literature and water as possible.
R. David Edelman retweeted
Awesome day in terms of new model releases from both Deepseek and Grok. Both of these new model updates represent huge jumps in capability at insanely low costs.
This is truly Jevons paradox for AI playing out. By bringing down the cost of AI, demand will continue to accelerate. When I talk to enterprises, they have vastly more use-cases for what theyâd like agents to handle than generally what theyâre able to budget for today.
They want agents to scan their codebases for security issues. They want agents to review all their documents. They want agents to process information in a large portion of their workflows. And so on. When we bring down the cost of AI or drive up capability levels, it opens up all new use cases in the enterprises, which drives even greater demand.
This is also great for the applied AI layer. The more choice of models - especially those tuned for different types (or costs) of work - the more value there is in being at the layer that can route and optimize based on the task. Great for the entire stack right now.
R. David Edelman retweeted
A super smart scientist asked me last week: what's so hard about reshoring general advanced manufacturing (short of TSMC frontier precision chemistry)? Isn't the issue just labor cost? Nope.
A toy example to illustrate: We're trying to build a lot of drone motors in America right now, and falling short. It's not a labor cost issue.
The work instruction says: this magnet, this adhesive, cure at this temperature, wipe the rotor first. Any competent team can build the station and follow it.
But in week 3, motors start failing vibration testing. Some of them, mostly night shift. Also, weirdly, whenever the careful new hire is on station.
The adhesive cures differently with humidity, and the dehumidifier cycles differently at night. The new hire wipes the rotor generously, and that leaves a solvent film the adhesive doesn't like very much. Someone who has run a motor line before can figure this out in 10m. Everyone else -- struggle city.
The "fix" is two lines in the work instructions: run the dehumidifier continuously, one solvent wipe per rotor. But those lines aren't in the work instruction, and they're not in anyone's! At best the fix becomes a corrective-action report in one line's files, linked to nothing and never read.
A motor line rapidly may go from 5 to 40 stations, each with a dozen of these painful stories on the weekly as it grows. "What Shenzhen knows" is not written down in a book whose covers can be sliced off and ingested into a model, it's a few million humidity stories stored in people, and a pace of knowledge-rich adjustment that keeps cost down and yield high. The only way to get it back is work with people who have brought up lines, to reshore and grow the ecosystem of people who have this knowledge, and to capture that knowledge as we build.
Lots of people have clever ideas about the data capture/API piece, but we think the business of manufacturing may need equal portions pain tolerance, customer commitment, and cleverness on talent.
America needs to solve the other half of reshoring.
R. David Edelman retweeted
1908: the Lancet, one of the most respected scientific journals, calls for 18 age limit on reading in bed amidst a moral panic surrounding children becoming "addicted" to novels, which were "designed to keep kids hooked" and destroy their attention/mental health
