The people at Apple are likely just as furious as their users are. I anticipate that the Mac will get aggressively locked down by 28: hopefully only insofar as some system applications will firewall their data from everything else, at a level that not even admins can override. If you’re building apps (like OpenBubbles) that rely on this access, your days are numbered.
In a completely unsurprising turn of events, Meta's Muse AI is stealing Apple Messages, past and present, and uploading the contents to its cloud, even if explicitly told not to. By @Amber_M_Neely
appleinsider.com/articles/26…
666
Did you know that ~32% of federal student loans are currently in default or behind 31+ days on payments?
Trump is banning students majoring in degrees that don’t make enough money from taking out college loans. Degrees for social work, art, religious studies, teaching aides, and music will be hit the hardest. trib.al/NmmAzhf
495
After dozens of billions of tokens i’ve come to the conclusion that this slider is operating on the wrong dimension. I don’t think reasoning level actually matters that much; I’d rather this slider just be “Luna, Sol, Astra”. Leave the advanced menu in there to select a reasoning level, but just set each of them to High (or infer it from the initial prompt) and be done with it.
244
Big fan of the concept that when deciding which model class to kill, they chose The Earth.
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.
We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
299
Imagine telling others they need to defend their positions in public while blocking replies to your tweets.
I think you've done enough calling us paranoid and preposterous. The next step is for you to defend your position in public against someone who will push back against it. I'm happy to meet you for a debate anywhere, anytime.
You're a world-famous veteran of dozens of debates against the world's top intellectuals, and I've never argued in public before, so adjusting for the relative correctness of our positions, if you're a betting man I'm happy to put my $5000 against your $1000 (ie 5:1 odds in your favor) that I'll win by some standard of audience opinion change. Let me know if you're interested and we can hash out details.
1
1
361
Genuinely we’re past the point of it being irresponsible to use any Chinese AI product. Even if it’s an open model hosted in the west, the hypothetical threat of sleeper agent behavior is too great, and clearly all of their labs are dishonest.
We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.
Read the report: anthropic.com/threat-intelli…
209
The biggest reason why I feel that software engineers are still valuable is: I've seen what software engineers can produce with AI, and I've seen what many in other roles produce. Its a 10x+ difference, and I cannot explain why.
In one recent contribution I was doing high level review on, the contributor said "I used Astra so it should be good" brother what incantations did you poison Astra with to produce this? It invented new UI components for everything, including the *persistent global navigation bar*, such that the bar would totally change styling as you navigate to and away from this page. And this is not Astra's fault, because I can use Astra just fine and get great results, and its not because I'm prompting it "hey go reuse components". I don't know. Increasingly it seems like getting really strong results out of LLMs is even less transferable/teachable than coding ever was. It is, very truly, like casting magical incantations to an opaque magic box, even the caster doesn't know why the things he says works, and when someone less experienced says words into the box they do get little sparks to emerge from it, but that's nothing compared to the fireworks someone who knows what they're doing can elicit.
It is its own skill, which inherits a lot from being a good software engineer, but can definitely be picked up independently, with a ton of practice (and taste; the willingness to say "something about what I just told it to do didn't work, and its my fault, not the AI's fault).
1
146
At this point “real work” has become no true scotsman. If you’re building models and harnesses, that’s not real work. If your consulting business can take on 30 clients now instead of 12, that’s not real work. Autonomous driving isn’t real work. Validating new drugs in test isn’t real work. Designing real hardware humans don’t understand, but works, isn’t real work. I don’t know what real work even is anymore. I think, maybe, when LLMs through ten layers of application and abstraction lead to, I don’t know, a robot that replaces construction workers building housing, then people might say “ok I guess that’s real work”.
Not really observations that others before me haven’t made, but:
- Despite the seemingly magical nature of LLMs, reflection over a >3 month timescale suggests my total productivity hasn’t increased by over 100%, or perhaps even by over 50%, and a lot of time is actually wasted because LLMs enable me to spend time on gratifying but low-productivity tasks that in the future turn out to not be useful
- Also, capabilities are incredibly spiky and highly correlated with the degree of investment poured into them, which my earlier tweet about math benchmarks implicitly points out
- From the above, it seems that the nature of LLM intelligence is wildly dissimilar to that of human intelligence and we won’t trivially get to something superior to human intelligence in all important respects just by scaling up existing approaches with various tweaks; even if AGI Is eventually achievable, this implies a significantly longer timeline
- Benchmark progress is almost definitionally guaranteed to happen because the process of constructing a benchmark is a direct precursor to the process of constructing a training dataset used for hill climbing that benchmark, but the scope of what can be captured in a benchmark is (at least for now) grossly lacking in terms of its relevance to real-world work, with maybe several limited exceptions
- Progress seems highly gated by data but the nature of model training means that each “next dataset” is significantly harder to assemble than what preceded it; some wins are possible through synthetic methods but those feel more like “patching up gaps” than “pushing the frontier forward”
At a higher level, I guess I’d say there’s a sort of refusal to think carefully about what models are or are not useful for in a rigorous way which I find personally quite annoying, and instead a reliance on some nebulous notion of being “AGI pilled” as a replacement for serious thought. I think people are very quick to anthropomorphize LLM intelligence because humans communicate through words and we infer the intelligence of human counterparties through comprehension of their language, but this leads them to wrong conclusions; for example if we observe that a new model proved some incredible mathematical theorem, some will say, “well, don’t we have AGI now, huh?” But to me, it’s actually more like, “well, given how hard it would have been for a human to do these mathematics, and given the limited economic effect of LLMs upon the world so far, isn’t it actually a negative datapoint vis-a-vis the generality of LLM intelligence?”
1
176
The value of the AI industry is not meaningfully founded on the presumption that everyone will start writing personal software. Some valleysurfers say stuff like that, but a realistic estimate is that it has already expanded access to meaningful software creation by 2x-4x. Beyond that, AI has tremendous multi-disciplinary value that this article does not address, yet immediately pivots the assertion “not everyone wants to create software” into “AI done”; reminiscent of a self-centered entry or mid-level software engineer who thinks their work runs the world, and is only slightly more epistemically rigorous than someone saying “AI writes bad code, AI done”.
146
The reason why I’m so distasteful of doctors, generally as a profession and absolutely not specifically/individually, lies in how they’ve abdicated almost all of their authority over patient outcomes to other parties who are definitionally less equipped to improve them. The noble profession has been hollowed out, to the point where most doctors are ChatGPT bots not because they want to be, because they’re forced to be.
42
So you end up with a situation where the doctor takes responsibility for being overworked (because they don’t control their schedule), their services are too expensive (because they don’t control billing), and oftentimes declined by third parties (because they don’t control payment). Everything is outsourced.
Yet I •do• levy blame on the medical profession (never individuals) because the •only• people who could change it are Doctors! Everything else is part of the system, but doctors are singularly in the unique position to forgo that system and try to build something better. And some do, and I love them for that.
But most don’t. Most would rather perpetuate a broken system because it’s less risky and, in my experience, their trophy wife needs a 4000sqft suburbs mcmansion, so they need to specialize and make $550k and thus anytime they show up in a procedure room it costs $10k/hr. Who pays for that new boat? You do.
23
One has to wonder if any of those “packed” weekly App Store reports contained any of the revenue metrics Apple claimed in court they do not track… (engadget.com/epic-trial-witn…)
For nearly 15 years, I sent @tim_cook a weekly report every Sunday night on the global App Store business. My team and I packed it with data, charts, and commentary on what was happening around the world.
Most weeks: silence. Every so often, a probing question would come back about a specific metric or market. Or, better, a note congratulating us on a strong quarter.
Those reports taught me a lot about discipline and accountability. No matter where I was in the world, I sent that report. Tim expected his leaders to know their businesses cold. Whenever I put anything in front of him, I needed to understand not just the numbers, but what was driving them and what we were doing about it. His expectations made me a much better leader.
On my last day at Apple, Tim and I had a 10-minute 1:1 scheduled. We talked for 30. Reflecting on my 21 years at the company, he told me my greatest accomplishment wasn’t helping build a $1 trillion App Store ecosystem or helping scale the business to more than one billion customers. It was the team I had built and the leaders I had developed who were ready to step up after I left. That meant the world to me. We took this photo together right after that conversation.
Today is Tim’s last day as CEO of Apple. Much has been written about what he and the company accomplished under his leadership. But perhaps his greatest legacy is the one he taught me to value: the people he developed, the culture he shaped, and the leaders ready to carry it forward.
I’ll always be grateful I got to be a part of it.
91