pocketsand

Obvious massive asterisk that benchmarks are not always realworld use:

Apple M5, Single core Geekbench: ~4,200 (https://browser.geekbench.com/macs/macbook-pro-14-inch-2025)

Apple M5 Pro (15-core, lower core version), multicore: ~26,000 (https://browser.geekbench.com/v6/cpu/18535781)

compare this to the top of the line Intel Panther Lake chips, which have comparable battery life. I cherry picked a 16" Dell XPS machine, which has the best thermal headroom, for its best score: https://browser.geekbench.com/v6/cpu/18390748

Single core: ~2,900 Multicore (16-cores): 16,900

Geekbench is bursty, so we can look at more sustained test, Cinebench 2024:

https://nanoreview.net/en/laptop-compare/dell-xps-16-2026-vs...

Single core:

XPS 16 (2026): 111 MacBook Pro 16 (M5, 2026): 203

Multicore: XPS 16 (2026): 613 MacBook Pro 16 (M5, 2026): 2065

on GPU, https://spylab.ai/seo/v5/C54a/

"Despite the XPS 16's discrete Nvidia RTX 4070 laptop GPU, the MacBook Pro M4 Pro's unified memory architecture outpaced the Dell in 4K video export and machine learning inference benchmarks by 22 percent on average."

For GPU, this is not comprehensive. It depends heavily on whether it needs raw grunt, where Nvidia discrete chips will win. When the processes uses the NPUs on Apple's chips, it will often win. They trade blows.

Efficiency I think is close to a wash on the latest machine, but before Panther Lake, Apple's win handily. My Framework 13 on AMD would last me about 2 hours doing regular work; my Macbook Pro doing the same workload would last over 10 hours. Thank goodness Intel caught up here.

I do scientific computing where Apple has some disadvantages. Matrix math heavy things lose out to discrete GPUs, as do -- I'm told -- things depending on 512-bit extensions (e.g., AVX).

Until last week, prices on Framework/Dell vs. Apple were similar. I think Apple is probably 10-20% expesnive at this point, but adjusting for performance, Apple still comes out ahead.

Apple's displays used to have a huge advantage. Now that you can get OLED displays on performant, efficient Panther Lake machines, this is far less of an Apple advantage.

The upshot is that the new Panther Lake machines caught up considerably to Apple, but they're still about 20-30% slower (sometimes more) in most workloads, and IMO the build quality is still not quite as good. I think many of them actually have better displays. Battery life is comparable on better equipped PCs. IMO once you can work an entire day and a little more unplugged, you're good to go.

It's not hard to find this data and evaluate it objectively.

A strange retort. "Yeah well, your CHILD'S COMPUTER is better, but that's for CHILDREN. We're talking adult computers here!"

This comment is untethered from reality. Apple Silicon at present beats every single laptop on the market at CPU-bound workloads using a fraction of the power draw. Exceptions exist but usually those cases are break even or close calls.

It trails GPU workflows on the high end but wins on the low end. It still wins on efficiency.

It falls over on storage and RAM prices (well, for about 6 months it was competitive here).

I say this as someone who over the last year has done the majority of my competition on PC hardware running Linux.

You may be looking at this as a status game but it has clouded your vision. It is implausible that mass market products with mass adoption find their success solely on status. If believing that makes you feel superior, well, enjoy the rush.

Work for me or I will publicly shame you, try to damage your reputation, and make your employees potentially question your loyalty.

Threatening people if they don’t take a job with you is in the most literal way a challenge to your autonomy as a human being.

For me, the fact he tried to compel the WPE CEO to work for him or else he would expose that she was in negotiations with him is the most unhinged thing I’ve ever heard in a hiring process. Quite literally an affront to freedom.

My most charitable guess at what is going on is severe mental illness.

Having an ad partner that is doing bad things on its own direction is different from explicitly paying a company to do a specific bad thing.

Far be it from me to defend Meta on privacy —- I won’t —- but we should at least characterize their numerous evils honestly.

For what it’s worth, in most cases, an organization that overly focuses on process is superior to one with a naive individualism that throws employees under the bus by default.

If you took the "no it won't" side of every argument about "how in X number of years, AI is sure to Y", you'd be way ahead.

In any event, raw parameter/weight count to me seems like a very primitive way to judge "complexity" in comparison to the human brain. Looked at most ways, our brains are for more efficient at doing the incredible things they do than LLMs. Consider how little language young children are exposed to in comparison to LLMs given their abilities to figure out how to produce language.

If the brain doesn't work like an LLM, you can expand the size and "complexity" of these models to the moon and they won't outperform the brain. Current models can write impressively well, but they can barely do math. It's clear they don't reason as we do.

You would think so, but people like Sam Altman have suggested that they can use AI-generated data to train their own models. See here:

https://www.nytimes.com/2024/04/06/technology/tech-giants-ha...

Yeah, sure. But the marginal cost is zero, whereas a Slack subscription for every person in our org will cost about 1 million dollars a year. And it doesn’t integrate as well with every other piece of functional but mediocre software.

The person approving the $1 million dollar budget item doesn’t really care that Teams isn’t “free” in the sense that there is no free lunch, and while they perhaps have moral qualms of antitrust, that’s outside their purview. We’re locked into Office suite and right now there is no extra charge for Teams.

I would disagree. I work in healthcare and we’ve always used SQL Server. While I wouldn’t pick it, it’s been reliable and integrates with auth.

No one “loves” Teams, but honestly it serves its purpose for us at no cost.

No one loves OneDrive but it works.

I think people underestimate how much work it would take to integrate services, train people, and meet compliance requirements when using a handful of the best in class products instead of MS Suite.

My experience with DataSpell has not been great. Granted, my workflow leans toward R, and it DataSpell has a Python-first approach, but the app was basically completely broken to even load R, and StackOverflow was full of relatively old posts of people with the same problem. If they really cared about that app that would never happen.

I just do a lot of my R editing in PyCharm now and flip between terminals and RStudio. I was hoping DataSpell could unify that, but it's not ready.

The new RStudio IDE is promising, however.

As I'm sure you know, there are a lot of variations on how quantiles are calculated in various software. The 25th percentile, e.g., doesn't always line up with a value in the dataset, so sometimes nearest rank methods are used, otherwise a linearly interpolated data point, where interpolation is done in various ways.

In any event, none of these methods assume normality, or rely on CDFs of a normal curve.

If they did, every box plot would be symmetric.

The fact some people think that boxplots are constructed in such a way is a pretty good reason to take the author's article seriously as for how boxplots are confusing.

I do stats and data viz for a living and the article seemed perfectly reasonable to me.

He isn’t dogmatic.

He makes reasonable arguments.

I’m confused by these hopelessly uncharitable readings of the article.

Why? They’re non-parametric and make zero assumptions of normality.

He paid 54bn to have the run of a web site where he creates his own reality, surrounded by sycophants.

It cost a third of that for Boston to do the Big Dig.

To each his own.

Academic science is excluded from market mechanisms by legal fiat.

This to me just seems untrue. What is your basis for this claim? There is plenty of research privately funded by corporations, some of which is very influential. Often this work is published by university researchers. Ask any university researcher about the numerous compliance courses we all have to take about funding and conflict of interest.

It is true that the biggest funders (NSF, NIH) are not market-focused, but for good reason. The market does not prioritize the public good. I know first hand -- my son has a rare disease (1 out of 20,000 people). There are many drug companies putting vast resources into drug development in the hopes of a huge payoff. In reality this benefits a small number of people (I remain grateful for how improvements have helped us). I'm grateful our major scientific funding bodies are not swayed entirely by market influences because it would lead to us focusing on a narrow set of scientific problems which would ultimately limit the way it helps the public good.

Im any event, I work in biomedical research. I think your diagnosis (incentives, process) is correct, but the way you discuss the attitudes and motives of researchers is wrong-headed.

You say:

Their reputation is tanking but they just don't seem to care and why should they? They'll get grants from the NIH anyway because they're all as bad as each other, and nobody in politics is talking about a total defunding of the sector yet.

You're talking about hundreds of thousands of researchers as if they're all psychotic citation fanatics with no care for truth. That is not reality. I think the kind of psychotic, data-manipulating researcher who would put people's health and lives at risk for citations -- or fabricate data sets out of thin air -- are vanishingly small. We can point to a handful of them -- the author of this paper, and the Daniel Ariely's and Francesca Gino's of the world -- but there are tens of thousands of people in every field working on research in good faith, with utmost care. The vast, vast majority never have any scandal, never get caught up in data manipulation, and so on.

No field I know of out-right tolerates fraud (and I follow all the retraction stories fairly closely). I think the closest we get to "toleration" is researchers dealing with scientific problems who more or less say "we're not going to publicly flay you but behind your back we're all going to know what you did and your future is limited when it comes to big grants, prestigious invitations, and so on." People who are credibly accused of fraud become pariahs and often targets of scorn not only within the research community but in the press and wider community.

The most serious issue IMO is not outright but poor norms around scientific practice, leading to p-hacking, harking, and other "forking paths" problems. Calling that type of behavior "fraudulent" is perhaps justifiable under some ways of thinking, but I think the word fraud mischaracterizes what is going on. There are, in fact, many serious efforts to root out this type of behavior and put in transparency rules to open up research to scrutiny, including among funders like the NIH.

A family member is paid very highly. He has switched jobs almost ever year for over a decade, including at top tier tech firms. I'm not sure why he does this but I have my suspicions. What I do know is that he is never seems happy with his work. And I also doubt how much impact he can be having with such short stints.

I have never quit a job without a major life event forcing me to (spouse getting a job, moving, etc.). I've had great experiences everywhere and I'm probably underpaid for it. I have no regrets. I'd prefer my lifestyle and job satisfaction to money.

2854684

Not an elite 6 number UIN, but at least on the bottom side of 7 digits.

Strange that people would be upset over an optional feature.

I like the idea and may enable it.

I did not say they're destroying the chargers. You are nevertheless underselling the potential implications IMO. These chargers are finicky. People's primary complaint with Electrify America is how hit or miss they are in actually working. People have the same complaint about Tesla, but to a lesser degree. That combined with Tesla's wider presence have made people correctly laud the quality of the market. Expansion aside, these chargers will need to be maintained, and sacking the entire division doesn't leave me hopeful they will be.

Moreover, EVs are going to just keep growing in number. Every brand now has a pretty solid EV for sale. Keeping the same number of chargers isn't helping anyone.

It's their right, but my main point is that this just further convinces me Tesla isn't serious about keeping up with EVs. Pumping money into a useless truck, chasing a dozen other fanciful projects, abandoning their world class charger network. The future is -- every other manufacturer -- it would seem.

I understand the competitive advantage has changed with the NACS deal, but the last major Tesla brand advantage I as a lay-observer saw was their clearly superior charging network and connector. People seemed to unanimously agree their connector was better, that opening it up was good, and that Tesla charges are more reliable and more available than competitors'. Why you'd immolate that brand equity is beyond me.

I hosed an iBook G4 repairing a hard drive back in the day. Oops

What was enlightening for me was watching my wife, who is a physician, work in electronic medical record software like EPIC. It's ugly, it's dense, the colors are high contrast.

It also fits everything they need to see in one place and makes it easy to make changes. I do not think there would be a way to make it look "modern" without making it harder to use.

I've always wondered if the "hard productive way" as understood by conventional wisdom (using VIM, using a language like APL) is actually more productive.

For this topic, I have in mind two exercises:

First, get people to nominate "experts" in a language, and assign them a straightforward task, and see who is actually faster using their native environments with code completion, language servers, etc. by systematically running these trials.

Second, get people untutored in each language of similar experience, randomly assign them to a language, give them a not short but not long period of time, then do a similar exercise as above.

The first has problems with assuring there is skill balance between the groups, but would I think tell you about how productive one can be in each language.

The second would have good causal inference properties, but probably not tell you about the downstream effects of learning something hard with great rewards (e.g., vim or emacs).

I'd bet there are marginal gains to APL (or vim) but probably nothing that matters in long-term productivity.

However, I do think people underrate how fun it can be to "get good" at things like vim or APL, and I think that has really positive knock-on effects.

In fact, the CLT is remarkably robust to distributional assumptions. Examples where it breaks down (e.g., non-finite variance) are comparatively rare, even if there are "many" of them.

As with all things statistical, judgment is required.

I saw a post shared on some platform recently that some screens in Amazon still use this technique.

I remember slicing and dicing up the rounded corners in photoshop.

For me, the life changer was SSI. You mean I can just include files and not copy menus all over the place?

I also remember being enamored by image maps. Mapping out the polygons without software was beyond the confidence I had as an 11 year old.

And a self-hosted CGI-backed counter--the holy grail.

z-scores just require knowing a standard deviation and a mean. you only need an assumption of normality if you want to do things like assign p-values. of course, that is mostly how they are used.

My issue here is that Imbens and Gelman are respected because they are good statisticians but also clear thinkers who have proved themselves dedicated to doing good, careful work. If you have major issues with work their name is on, I would think their reputation would at least lead reasonable people to contact them first before writing something which leads to hurt feelings and numerous comments and corrections and more posts on each end.

This whole affair is a case in point. Several authors ended up responding with very reasonable responses, which the author then acknowledged. Then he writes another post, which is more measured and insightful.

The original post would have in fact been a far better one if the author had first sought those responses and just addressed them all together at once. His points about metascience would not then be dragged down by back and forth in the comments that clearly got personal.