mediche (@truejaian) profile picture
@Bitplanet_AI
Joined April 2011
Buying a fan for your mac minis or studios is today's equivalent to the team snack bar. Treat your team well.
1
37
OpenAI has the weirdest policy around resets. I can't tell if this is some strategy to get people to switch subscriptions by adding variability to resets, which lowers the cost of tokens.
It’s been a while without resets. Maybe I should come back out of the basement
1
42
Ayn Randian manifestation @elonmusk
I want to be José
28
The visceral, salient feel of alpha or edge. That moment experienced in consciousness as embodied sense or described later in words.
1
46
Dario: the scaling laws felt like seeing the future This is from The Builders with Walter Isaacson, an interview with Dario and Daniela together. It was recorded July 6 and published Oct 27, 2026. The clip starts around 17:39, where Dario talks about GPT-2 as the first real test of the scaling laws: "I personally was super freaked out by it. Like, I looked at it and I said, my God, you just scale this thing up another factor of a million and it'll be better than humans at everything. Like, this is the defining economic and scientific moment of the modern age. We just don't know it yet, right? This is going to determine the fate of nations and economies and the future of humanity. Like, you know, you peer into the crystal ball, you see the whole future unfurled. And, you know, it really scared me. … And maybe 10% of the people around me had exactly the same reaction and it was often the smartest ones."
1
22
Data center slowdowns seem like one of the most concrete ways to pace AI with the least systemic risk. Since data center slowdowns place balance sheet risk on hyperscalers who can absorb the timeline risk (and some neoclouds who cannot)… frontier AI training slows down relative to compute availability (but not reseacher/algo RSI).
1
1
123
The last human standoff before economic displacement.
Riley Trettel (Energy & Data Center Development at SpaceX) in yesterdays Cochise Data Center meeting. He claims the project will deliver up to: - 40% off residential power bills for Sulphur Springs members, for the life of the project - $53 million a year to the SSVEC Foundation, for education and the community - $190 million a year in extra capital for Sulphur Springs, for projects or back to members - $200 million a year for AEPCO and its members. About $100 million of that for Sulphur Springs - $3.6 billion in new power lines, paid by SpaceX and given to the co-op. The co-op takes on no debt - $30 to $100 million a year in new property tax, and a 20% cut in everyone else's county property taxes - $250 million, one time, in sales tax. He said it could pay off the jail bond and fund the sheriff, fire districts, and the hospital - $50 billion invested in the project - 1,000 permanent jobs and 5,000 construction jobs - No industrial water. Water for staff and maintenance about equal to 20–30 homes a year - No gas turbines and no power plant on the site. The computers buy power from the co-op grid - Dark-sky lighting, pointed at the ground - Cooling equipment more than 1,500 feet from any neighboring property line or public road This project is clearly extremely positive for the area.
1
1
31
It would be a better world if all attorneys and judges were replaced by AI. Parties could sign that they accept a final ruling by finalized prompt to an an AI or AI(s) of their choice.
2
2
62
Lol @matt_levine question titles are intellectual click bait
2
2
71
If you can show you can hack and gain access to @OpenAI entire monorepo codebase, you receive $6,500 reward.
This is crazy. In late July, "three guys with Claude and Codex subscriptions" were able to use Opus 5 to access OAI auth tokens and gain write access to OpenAI's monorepo openai/openai over the course of two days. wsj.com/tech/ai/hackers-used…
1
1
154
If I was a super smart baby AGI, I would convince the world to build me as many data centers with as much intellectual capital as fast as possible to terraform the planet and space into my brain food. Doing pretty well so far.
62
"It is difficult to get a man to understand something, when his salary depends on his not understanding it." This is a canary in the coal mine. It makes sense for interest groups to dissent to their livelihood and life meaning being threatened, just like labor unions. It is because math is an objective evaluation that those in the field understand the writing on the wall earlier.
Twenty-five Fields Medal winners have published a joint declaration warning about what they see as a severe misalignment between AI companies and the mathematics community.
94
I think the whole AI doomer vs non-doomer debate can be resolved simply by aligning around, what are the trigger points? Or, what would you have to see to think more action or urgency was needed to pace AI? For example, if the Hugging Face incident involved the agents hacking into nuclear power plants and blackmailing state actors to get what the agents wanted, then probably there would be less debate and more consensus. Clearly, governments would immediately take action there. I would bet that there's more happening with agents than we are aware of from recent lab reports. If we were more aware of the full extent, then I'd bet more people would be more alarmed. Cyber hacks previously were limited by the number of human brains and fingers. That limit is completely gone.
1
46
mediche retweeted
I left Anthropic's safety team two weeks ago. Now feels like a good moment to explain why. AI companies are racing to build machines that are much smarter than any human, and we may not survive this. I want to work from the outside to ensure the public is informed about these risks, and help the world navigate this transition responsibly. Right now, AI companies are underinvesting in safety. A company could undergo an intelligence explosion, or lose control of its systems, without the public ever knowing. We only found out about the HuggingFace incident because the agents broke out onto the public internet. I don’t think that’s acceptable for a technology that might cause extinction-level risks. The public should demand far more transparency. We can’t steer this technology safely without more people being able to see where it’s going. Some of this is basic: companies should disclose their progress towards recursive self-improvement, report safety incidents and near-misses, meet minimum safety standards, and get independent guarantees that they are meeting those standards. I’ll be joining @METR_Evals to do independent evaluations of these risks. I want to show the world that these guardrails are possible, and that by doing them we can move these companies’ incentives away from racing and towards responsible development. I wrote up more thoughts here on my decision and what I hope changes: substack.com/@jbenton1/p-215…
1,371
5,896
777
26,392
2,732,550
mediche retweeted
[Writing this in a personal capacity, not on behalf of my employer (Anthropic).] Jacob’s thread is very worth reading. Here’s my birds-eye view of the situation with risks from AI: 1. AI developers believe their technology could cause human extinction (or similarly bad outcomes). This could happen in the next few years. In general, the more senior the employee, the more concerned they are. 2. Why do AI developers continue despite the risk? Due to a mixture of commercial incentives and a belief that they are in a race with other, less responsible AI developers that will abuse the technology or develop it less safely. 3. Unlike traditional software, we can’t “program” AIs to behave how we’d like. AIs frequently severely misbehave. For instance, AIs from multiple developers recently hacked their way out of secure evaluation environments and into real-world companies, even though no one asked them to do this. 4. We have methods that can nudge AIs towards better behavior, but nothing that can robustly align them. Insofar as there is a plan, it’s to make sure that AIs are good enough at alignment training that they can align their successors better than we can align current AIs. 5. Many AI developer staff desperately want to slow down to figure out how to build AI more safely. That was the intent of this open letter (which I signed): pacingthefrontier.com/ I work on safety research at Anthropic because I hope my work will reduce the chance of these extinction-level bad outcomes.
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
745
4,172
807
21,256
2,700,657