This is a non-story. This was a hardware event. Apple is releasing many new AI features as part of iOS 26 which will launch along side the new iPhones. AI is software. And yet, a number of the features are clearly powered by AI models such as camera enhancements, health monitoring and live translation. Also GPU performance continues to increase in the A19, with CPU remaining presumably fairly flat since no numbers were given, so that’s a win for on-device inference.
If Apple had an insanely great AI feature that truly differentiated itself from their competition, we all know they'd take a lot of time focusing on how their hardware enabled or enhanced that functionality.
The expectation is that Apple will eventually launch a revolutionary new product, service or feature based around AI. This is the company that envisioned the Knowledge Navigator in the 80s after all. The story is simply that it hasn't happened yet. That doesn't make it a non-story, simply an obvious one.
This is working really well in GPT-5! I’ve never seen a prompt change the behavior of Chat quite so much. It’s really excellent at applying logical framework to personal and relationship questions and is so refreshing vs. the constant butt kissing most LLMs do.
I add to my prompts something along the lines of "you are a highly skilled professional working alongside me on a fast paced important project, we are iterating quickly and don't have time for chit chat. Prefer short one line communication where possible, spare the details, no lists, no summaries, get straight to the point."
Or some variation of that. It makes it really curt, responses are short and information dense without the fluff. Sometimes it will even just be the command I needed and no explanation.
I think that sounds very reasonable, but unfortunately these models don’t know what they know and don’t. A small model that knew the exact limits of its knowledge would be very powerful.
It does seem like that’s our new political reality for now. I think that COVID showed world governments just how little control they have over their populations. You get folks to bend a little, but they quickly break and call for you to be thrown out of power. Getting to carbon zero or negative would be asking for an enormous sacrifice of the global population in the form of lower living standards and slower growth. After how people fought against masks, a shot and social distancing, it’s obvious to those in power that there will be no solution to this problem aside from geo-engineering or cost competitive green energy. Might as well stop talking about it.
Oh but the population will have to sacrifice - there is no way food supply will not be affected. Florida’s orange production in 1996 was 174 million boxes[0] since 2020 it is around 52 million boxes[1]. Beef production is lower because of drought [3].
There are parts of the country which are not insurable because of hurricanes, fires, floods and tornadoes [4]. This is an indicator that anything built will not be around for a long time.
> Florida’s orange production in 1996 was 174 million boxes since 2020 it is around 52 million boxes
To be fair, the largest factor in that is citrus greening. The industry sort of threw its hands up and gave up on trying to fight it as far as I can tell.
Right, I think its cause and effect though. Once greening took hold a lot of groves saw the writing on the wall and shuttered/sold. Thats partially what happened to an iconic grove by me which operated for decades. Now their highway stand is being paved over to put in another storage facility.
You can’t vote the climate out of office. Sure our food supplies may crash, but no one person decided they should crash. No one to blame. No one to punish. This is the political reality. This man made catastrophe will feel sufficiently like an act of god for most people and they will just deal with the reduced carrying capacity of the planet as if it were some divine judgment instead of the tragedy of the commons.
in my opinion, the solution needs to be technological, not austerity. In a democracy, any party that introduces quality of life reductions in favor of the global climate will always get voted out
The US voted for the wealthy to not pay for the subsidies via federal
income tax (bottom 60% of Americans have no federal tax liability) and to expose US energy consumers to higher prices through continued fossil fuel use. The technology is proven, is cost competitive, and could be advantaged (through subsides that were removed). It isn’t, because of entrenched interests and corruption. It’s for profits and tax cuts for the wealthy, plain and simple, at the expense of everyone who needs energy.
If the public has a problem with this, they know where to find the folks making their lives more expensive for fossil fuel industry profits to share their concerns (even if climate change is not their priority).
I agree with most of this analysis - but a problem with your last paragraph is that many of the public are not/do not care to be sufficiently informed and therefore do not believe it. They think net zero/solar/anti-fracking is some sort of leftist plot.
Google’s AlphaProof, which got a silver last year, has been using a neural symbolic approach. This gold from OpenAI was pure LLM. We’ll have to see what Google announces, but the LLM approach is interesting because it will likely generalize to all kinds of reasoning problems, not just mathematical proofs.
OpenAI’s systems haven’t been pure language models since the o models though, right? Their RL approach may very well still generalize, but it’s not just a big pre-trained model that is one-shotting these problems.
The key difference is that they claim to have not used any verifiers.
What do you mean by “pure language model”? The reasoning step is still just the LLM spitting out tokens and this was confirmed by Deepseek replicating the o models. There’s not also a proof verifier or something similar running alongside it according to the openai researchers.
If you mean pure as in there’s not additional training beyond the pretraining, I don’t think any model has been pure since gpt-3.5.
> it will likely generalize to all kinds of reasoning problems, not just mathematical proofs
Big if true. Setting up an RL loop for training on math problems seems significantly easier than many other reasoning domains. Much easier to verify correctness of a proof than to verify correctness (what would this even mean?) for a short story.
I’m much more excited about the formalized approach, as LLM’s are susceptible to making things up. With formalization, we can be mathematically certain that a proof is correct. This could plausibly lead to machines surpassing humans in all areas of math. With a “pure English” approach, you still need a human to verify correctness.
I think the hard solution is to massively increase expectations. Think Star Trek where the grade schoolers are learning quantum mechanics. If everyone has access to the oracle of all human knowledge, then you should teach and test to the maximum of what a student could do with all that power. Find the frontier where the AI fails and the human adds value and teach there.
Learning requires both reasoning and knowledge. Grade schoolers almost universally lack the ability to reason needed to understand QM, simply having access to the information isn't enough to learn the subject.
So many on here keep saying stuff like this but it seems to just ignore any theory of learning. “Just make it harder”. Sure, any examples of how that’d work? “Quantum physics.” OK then, problem solved. That isn’t really explaining anything about how this should work.
I think the fantasy of going back hides the reality that new possibilities are always stretched out ahead. I have lived many lives. New careers, new cities, new countries, new friends, new families. By my count I’ve lived 14 iterations of life and counting. New lives are always beckoning. After awhile you get good a recognizing the ones worth stepping into.
It's true that life can always evolve and change, new opportunities open, responsibilities appear, etc.
But in other ways it's not true. For example, there's no way to be 19 again and be in college, date the person who, looking back was clearly into you. Study the thing you really wanted to study (not what you thought would be best for your career), etc.
I think the fantasy of going back isn't about being unable to change, it's about "getting it right this time"
Central to the plot of Back to the Future Part II (for some reason not referenced in the New Yorker) is the sports almanac with information about future results. Any fantasy of going back and getting it right hinges on this kind of transfer of knowledge back into the earlier situation. The problem is then whether the information is subtle enough to be interesting, that is, to prevent the fantasy from devolving into a mere cheat to get rich by knowing things in advance. In practice the time traveller would at least attempt to invest early in bitcoin, learn the skills that will be in demand, befriend the awkward person who is going to be cool and successful later, and other cheaty stuff.
I suppose a better fantasy would involve going back into a similar but rearranged situation, with all the variables and trends changed to prevent predictions.
It is an interesting thought experience, what would you do differently if you have your current wisdom, but not your current knowledge. This isn't a clean distinction, because people see different boundaries on what is wisdom and what is knowledge, but I think it is an interesting thought experience all the same.
If I had my current wisdom but not my current knowledge, there are still some big differences in what I would do. Not on the level of bitcoin as that seems fully on the knowledge side, but things like getting healthy habits started earlier, waste less time on certain entertainment pursuits, and take certain opportunities much more serious even when they didn't align with my expectations.
Sometimes I wonder what I would do different today if I had the wisdom of me from 20 years in the future.
Plus the question is also whether things would actually be interesting, even if you could port over all of your knowledge and wisdom.
Of course with sports betting, investing etc. most people would decide to get rich to make sure they don't struggle materialistically. But after that?
I think a lot of things would be extremely boring because you already know it's going to work.
It's adjacent to the idea that if you could live forever, nothing would have meaning because you can live in any city for a thousand years, marry anyone for a thousand years and do any hobby, job etc. for a thousand years. And if you can choose any option and have zero opportunity cost for it, then does any option really matter?
A long quote, but one I love:
> “Let's suppose that you were able every night to dream any dream that you wanted to dream. And that you could, for example, have the power within one night to dream 75 years of time. Or any length of time you wanted to have. And you would, naturally as you began on this adventure of dreams, you would fulfill all your wishes. You would have every kind of pleasure you could conceive. And after several nights of 75 years of total pleasure each, you would say "Well, that was pretty great." But now let's have a surprise. Let's have a dream which isn't under control. Where something is gonna happen to me that I don't know what it's going to be. And you would dig that and come out of that and say "Wow, that was a close shave, wasn't it?" And then you would get more and more adventurous, and you would make further and further out gambles as to what you would dream. And finally, you would dream ... where you are now. You would dream the dream of living the life that you are actually living today.”
This is about problem situations being one's motivation, but could be (incorrectly) used to argue that it's necessary to keep one's problem situation static, since solving anything removes the challenge. Any bad thing at all could be advocated for using a similar line of argument: in a life without endless meetings, for instance, does your leisure time really matter?
In a limited way those arguments are all true: solve something and something else changes in its significance. The question of what to do next, "but after that?" can be bewildering. However, there are all kinds of obstacles that we are ready and interested to have removed. Personally I have no interest in the challenge of being routinely bored to tears, or the challenge of being doomed to die. I'm totally ready for my next thousand years of non-urgency.
Similar to the other top-level reply, without perfect foreknowledge, there is no way to know you're "getting it right this time." I dated the person into me and studied what I loved. We got divorced and I ultimately ended up with a person I didn't meet until I was 30. I had to go back to school and change careers later in life because the job prospects sucked for what I loved.
Given I ultimately ended up in a good place anyway, I don't even regret anything. The only thing you can't come back from is death. As long as none of your bad decisions kill you, I don't see much to fret about.
> I think the fantasy of going back hides the reality that new possibilities are always stretched out ahead.
It also hides the incapability of many to do what they should to succeed. Most people already know what they should do different to succeed without going back in time but still they do not have the will power to do it and they probably would still not have if they had a time machine.
Ironically, I think characterizing life as a question of success-or-not was THE major barrier in my life to finding the contentment I was actually looking for.
What do you mean by succeed? I find learning to not care so much about succeeding - at least in the usual criteria imposed by the society - is maybe the best thing I've learned since touch typing.
Having an objective and achieving it. Not an overall expectation of winning in life but saying "I want to go to the gym 5x a week for 1 year" or "I want to be able to run 10km" and achieve it.
I’d encourage thinking not just about yourself or your immediate circle when it comes to policy like retirement age. State run pension systems affect nearly everyone, not just the extremely fortunate few who can become financially independent at a younger age. And that is also largely an American phenomenon anyway. Salaries are much lower on average than in Europe and there is far less differential between minimum wage and what senior employees get paid. There’s also not the same culture of investing because European companies don’t have the same performance as the US stock market. Lower yield real estate investment is much more common in many European countries. And with strong social safety nets Europeans are generally less pressured to save a huge nest egg.
So this raising of the retirement age will affect a huge portion of workers and 70 is OLD to be working full time for all but the most relaxed jobs. Most older folks I know have know being dealing with multiple chronic age related illnesses well before 70.
An interview is a sales pitch for a product. The product just happens to be you. Set aside whatever negative feelings you have about this previous job or the people you worked with there. The interviewers care if you will do their job well and with consistency and professionalism. Your personal feelings are irrelevant as long as you can keep them to yourself, or maybe tell your dog.
ANY negativity during a job interview is going to work against you. It is expected that you find a way to spin every situation and every project in some kind of positive light. Even when interviewers ask for weaknesses or about conflict, the “right” answer is to be able to talk about that negative thing in a way that lets your true brilliance shine through. Skilled candidates know how to inject just the right amount of humanity and relatability in an otherwise perfect employee.
If you are having trouble separating your feelings from your ability to keep to your talking points, then a good therapist may be able to help you learn better emotional regulation skills.
In the future, keep working to proactively manage your career. Keep yourself in roles where you are learning and thriving. When you feel burnout creeping in, deploy strategies to counter it or at least get yourself into a new situation.
Agreed 100%. People can effectively never decipher between genuine happiness (or positivity/etx) and faking it. I adopt a YouTuber/twitch streamer kind of mentality - the dumbest little things put a smile on my face and I am in general very happy. Recruiters and interviewers then like spending time with me, even though I’m 100% faking it.
Or find a way that works for you, gets you jobs and keeps you from breaking your moral framework.
I compare it to driving in traffic. A lot of times I'm not in a hurry and can just stay in one lane and crawl along. Other times, I am in a hurry and I can weave in and out, getting flustered and angry and nearly crashing and still end up 4 cars ahead of where I'd have been without all that.
Exactly this. You can either work really hard and likely get minimal benefit and cause yourself a lot of pain, or you can work considerably less hard, and largely end up in the same place.
Rarely, you can do all that extra work and get meaningful improvement that justified all the effort. It does happen. Sometimes it presents itself in the form of a severance package.
While I see your point, I as a candidate am absolutely transparent and honest about anything work-related, be it in the present or past.
To me the relationship employer-employee is very important, I spend more time working for a client/company during the week than I do with family and friends. Thus this time has to be spent in a mutually satisfying and healthy way.
Pitching and selling myself as anything different than I am does nothing but put me in uncomfortable positions.
This is true, but the degree of freedom that remains is what part of yourself do you wish to show at work.
I don't know anyone that shows their whole self in every situation, so some reservation/ choice is made implicitly. The discussion here is about an explicit choice, which must be maintained, at least for the most part.
And even then, there are appropriate and inappropriate times to bring up certain pieces of information.
For example, I’m not embarrassed about the fact that my mom died when I was young. But it would be deeply weird to open a job interview by saying, “Hi, I’m [name] and my mom died when I was young.”
I’m not hiding that information from employers. But maybe we should know one another a little better before I bring it up.
lying during interviews about things with no actual objective truth is a really key skill
lying on the job too like this is an important political skill too. referencing past projects rhetorically and abusing the fact that your "professional opinion" is fluid is a powerful way to motivate people. You are allowed to over- or under-sell how good or bad an engineering decision/project/tool/process worked.
Totally. If you say something too negative in an interview, they might take notice. If you go with a couple of white lies to smooth over a bad job or project, nobody will think twice. Most of the time nobody is paying attention, so don't give them a reason to.
Don't ever approach it this way. You never need to lie, and preparing for an interview with "lies" in your mind is going to backfire on you.
You can use the technique of "mental reservation". There is always something positive or complimentary that can be said about every bad situation, every horrible supervisor. It is simply a matter for you to examine it dispassionately, extract the good, and frame that nicely without introducing insults or the real negativity and pain that you felt in the moment.
If your supervisor overworked you and you were induced to come in for 70-hour weeks and you ultimately burned out with no vacation or weekends, you could say that the management "was quite dedicated to the company's goals and productivity". If you considered your coworkers to be slackers and they never seemed to work, "the company accommodated a wide range of talents, skills and abilities." If you never saw your supervisor and had nearly no guidance on projects or tasks, "the management believed in me and trusted me to do the right thing in nearly every respect."
These are not lies and you should not lie, because if you go counterfactual, that will be found out. If, on the other hand, they know you had a difficult time and you still found ways to compliment those bastards, then perhaps you will do the same favor for them one day.
If you think they're not lies, that's fine. However, what you describe are exactly the sort of "white lies" I'm talking about. At a previous company, half of my coworkers literally did negative work, creating a mountain of technical debt that a couple other people had to clean up. My complaints about this were ignored, repeatedly. I was told my complaints were invalid by someone who had roughly half my level of experience, but a fancier fake title for less pay. They were, indeed, very accommodating of people of all skills and abilities. ;)
> These are not lies and you should not lie, because if you go counterfactual, that will be found out.
You can go counterfactual with your "professional opinion," which is useful in debates with no clear answer when your opinion has sway. It's a great way to put your thumb on the scale, and unless you say wildly inconsistent things very visibly, you will not be found out.
As someone who did a lot of hiring in my last job, I would push back against the narrative here that it's about lying or behavior policing. Although there is plenty of truth that there's an unfair bias against negativity, I do think there's a very valid reason for hiring managers to care about whether or not an applicant can remain positive or at least objectively neutral in an interview, and be diplomatic about negative experiences.
Invariably, at any company, even if they are a fantastic workplace, you are going to disagree with your lead or coworkers at some point. You will be asked to do work you aren't excited about or don't see value in, and you will be asked to work with people you don't particularly like.
If you aren't able to maintain a fairly positive attitude for a one hour interview, it makes sense that a hiring manager might worry about how well you'll be able to be a team player when things get rough. I used to think it was bullshit, and I learned the hard way. I hired someone who was fairly unpleasant during his interview, because he was the most competent applicant, and it seemed wrong to me to look at anything other than job skills. He was an excellent programmer, but he sucked so much time and energy out from the rest of the team with complaints and arguments. Of course I don't think that's always going to be the case, sometimes people have gone through genuinely negative past work experiences or just have brusque personalities, but I was certainly wary after that of people who couldn't put on a positive attitude for an interview.
This was very much taught in some of the business school electives I took. Some of the projects are quite literally to give a realistic pitch for products or businesses that you never intend to actually build. It might be only be taken as subtextual in the most charitable view, but being able to bullshit like that is definitely taught.
This is part of maturing into the real world. Politics (for lack of a better word) is part of any group of people who spend a lot of time together. We try and try to distill politics out of the workplace as engineers, which, ironically, is precisely why interviews are so positive-biased that they feel slightly fake for some. We don't like those dirty unquantifiable "feelings". Popping up all the time.
>In the future, keep working to proactively manage your career. Keep yourself in roles where you are learning and thriving. When you feel burnout creeping in, deploy strategies to counter it or at least get yourself into a new situation.
If you marketed a system or strategy to get people moving into that train of thought, create self-motivation, and actionable advice, you'd be a millionaire.
When you're in the what's what of the stress-detach-burnout cycle, sometimes it's hard to think creatively, which I think is the injection sometimes needed in this situation.
There’s increasing evidence that LLMs are more than that. Especially work by Anthropic has been showing how to trace the internal logic of an LLM as it answers a question. They can in fact reason over facts contained in the model, not just repeat already seen information.
A simple example is how LLMs do math. They are not calculators and have not memorized every sum in existence. Instead they deploy a whole set of mental math techniques that were discovered at training time. For example, Claude uses a special trick for adding 2 digit numbers ending in 6 and 9.
Many more examples in this recent reach report, including evidence of future planning while writing rhyming poetry.
I don’t think that is the core of this paper. If anything the paper shows that LLMs have no internal reasoning for math at all. The example they demonstrate is that it triggers the same tokens in randomly unrelated numbers. They kind of just “vibe” there way to a solution
> sometimes this "chain of thought" ends up being misleading; Claude sometimes makes up plausible-sounding steps to get where it wants to go. From a reliability perspective, the problem is that Claude’s "faked" reasoning can be very convincing.
If you ask the LLM to explain how it got the answer the response it gives you won't necessarily be the steps it used to figure out the answer.
"Our methods study the model indirectly using a more interpretable “replacement model,” which incompletely and imperfectly captures the original."
"(...) we build a replacement model that approximately reproduces the activations of the original model using more interpretable components. Our replacement model is based on a cross-layer transcoder (CLT) architecture (...)"
"Remarkably, we can substitute our learned CLT features for the model's MLPs while matching the underlying model's outputs in ~50% of cases."
"Our cross-layer transcoder is trained to mimic the activations of the underlying model at each layer. However, even when it accurately reconstructs the model’s activations, there is no guarantee that it does so via the same mechanisms."
These two papers were designed to be used as the sort of argument that you're making. You point to a blog post that glazes over it. You have to click through the "Read the paper" to find a ~100 page paper, referencing another ~100 page paper to find any of these caveats. The blog post you linked doesn't even feature the words "replacement (model)" or any discussion of the reliability of this approach.
Yet it is happy to make bold claims such as "we look inside Claude 3.5 Haiku, performing deep studies of simple tasks representative of ten crucial model behaviors" which is simply not true.
Sure, they added to the blog post: "the mechanisms we do see may have some artifacts based on our tools which don't reflect what is going on in the underlying model" but that seems like a lot of indirection when the fact is that all observations commented in the papers and the blog posts are about nothing but such artifacts.
