czep

The economic question is whether the average company will have the time or talent to roll their own models instead of eating the cost increases. The firms in question are exactly the same that have already decimated their teams. Can they so quickly pivot to self-hosted models if their AI workloads suddenly cost them 10x more? I bet most will simply start shoveling themselves deeper.

It is really beautiful to see distilled to the core algorithm. Despite AK’s claim that everything else is simply optimization, I am sure there’s a bit more to training a useful frontier model! Still, an excellent teaching tool and nice way to spend an afternoon walking through the code.

While true, it's still a user-hostile move. You kinda have to meet your customers where they are. If people are clicking ads without knowing it, that's a serious design problem. Yes, people should learn to read, but the risk of placing too much burden on users is that all it takes is one ambitious product manager to push an A/B test that generates huge revenue wins while enshittifying the product for everyone else.

You need to know that your data and conversations are protected and never sold to advertisers.

we plan to test ads at the bottom of answers in ChatGPT when there’s a relevant sponsored product or service based on your current conversation.

There is a severe disjoint between these two statements: the advertiser now knows what your conversation was about! This gives a lot of leverage to ad campaigns to design the targeting criteria very specifically crafted to identify the exact behavioral and interest segments they want.

If you’re referring to body shop consulting agencies this may be true, but IME as an IC consultant in DS/ML, my rates are well above Staff+ at FAANG, and nobody has ever tried to leetcode interview me. Yes, I have to do a lot of smooth talking, but performance is extremely transparent: if I don’t deliver, I don’t get paid. Honestly I doubt I could pass a leetcode style interview, and I’m glad I don’t have to do that anymore.

You're right, of course, there are suitable alternatives that work well. There are some things that will just be easier to manage such as Parental controls, backups, and Family features. Additionally, I've lost basic tech support skills on iOS. Part of my motivation for originally switching to android was to gain familiarity with the ecosystem, but now I've fallen behind on the iOS world and can barely use my kid's ipad.

I lean towards your viewpoint as well. Their assumptions (axioms, postulates?) are highly controversial, while the actual principles seem quite sound to me.

The only issue I can see is with #5. I would argue for decision making, you absolutely need a single metric, otherwise the process collapses into bickering over which measure is more important at the time (often for political or interpersonal reasons). The point is a bit vague on what exactly is being evaluated (product quality, which means what?). For launching products or running A/B tests, aim for a single metric as your decision framework. If you must have more than one, then be explicit about the tradeoffs in a flowchart: e.g., "if X is > 0, we launch. If x <= 0, but y > 2%, we launch, otherwise no launch".

There are too many insightful and thoughtful points in your comment for a single upvote to repay, so I must thank you with a reply. Becoming so demoralized with the present media and political landscape, it is comforting to find that there still remain semblances of reason for which I may again briefly restore some measure of sanity. Thank you.

Think of the people you've helped get their jobs done over the years. That's worth more than author credit in the colophon of a man page.

Bluff? Putin wants to live out his golden years in unabated luxury after anointing a loyal successor. If he nukes the entire world, he will live in squalor and misery while being hunted by the entire human race. He won't do it. Me hopes. Me also hopes someone in his inner circle gets a nice clean view of the back of his head and does the world a favor.

Any recommendations for best practices in "modern C++"? I have to resurrect an old codebase and would like to refactor it using more modern idioms, but its been a loooong time since I wrote any decent c++.

Application developers are your users.

No. Equating internal teams with paying customers is the very attitude that is causing these problems. Encouraging teams to think about their "internal customers" leads those customers to become entitled. We work together in the same company, our relationship is not the same as with actual external paying customers. I can't tell a paying customer that they're being unreasonable or lazy or unrealistic. We absolutely should be able to have that conversation with other internal teams when appropriate.

The post is describing the situation that has evolved as a result of QA being phased out. Telling Ops to suck up that extra work because "Dev are your users" is exactly why the post was written.

It's actually a very low risk investment. The HOA is suing the city, blaming the construction of the Transbay terminal next door. They have access to some serious attorneys, so if the building falls down they'll still profit.

It levels the playing field a bit, in the sense that code is more objective than an interviewer's notes

Except that writing code in a Google doc on the phone doesn't in any way resemble the real "playing field". I grant that it's level in that everyone faces the same constraints, but to write code in a gdoc with red squiggly lines underneath every keyword, automatic capitalization after a dot, and variable width font? How does that give you any kind of useful assessment? It's like handing a butter knife and some twine to a doctor and saying "here, show me how you'd stitch up this wound".

Thanks for the tips! One mantra I've tried when starting at a new job is "for the first 3 months say yes to everything, for the next 3 months say no to everything." The idea is you first immerse yourself in everything, to find out what works and what doesn't. Then you dedicate time to fix the broken processes so that hopefully when you hit 6 months your team is better positioned to be more efficient. Obviously you can't be too rigid, but it seemed to work for me when I had buy in. Curious if you think that approach sounds good.

This is so eerily familiar I swear I've had many of these exact conversations word for word. The only way this doesn't turn into a complete nightmare of a cluster is if the exec team "gets it". If so, you just might stand a chance at building a data team that gels with the rest of the org.

But if the exec team simply hired you for window-dressing, expect to be treated like a scapegoat and a punching bag. Any mistakes will be your fault. Any wins will be to the credit of the business. The Director of Product will ask to "embed" dedicated DS headcount and you won't have any real power to shape the roadmap. If the exec team doesn't give you equal footingf with Product (or Marketing, Finance, and Eng for that matter) then this will rapidly become a soul-sucking job. However, if E-team does give you the authority to call Product's bullshit, and tell Finance to stuff it, and not take direction from Eng leads, then you actually might be able to accomplish something really cool.

The problem is that the average voter will not understand that distinction. They will be persuaded to vote against environmental protections because of the doubt cast on "climate change". My original point was about the framing of the debate in the public sphere. In an effort to be pedantically correct, we've handed the opposition the upper hand by allowing them to reframe the debate away from the scary word of pollution.

It used to be called "pollution". Reframing the topic as global warming or climate change was a dedicated effort to soften the language and introduce room for doubt. I mean, everyone agrees that pollution sounds bad and is bad and we should stop doing it. But by calling it climate change, suddenly there's now an avenue to challenge its legitimacy, and shift attention onto the ideological debates. Meanwhile, as everyone is distracted, industry gets a free pass to continue polluting with zero consequences.

But were they really your choices? Everyday you wake up constrained by what someone who is no longer you did yesterday. Nobody is responsible for anything. I mean that philosophically of course. Give yourself a break. I recognize your name from many threads here and believe it you have made a difference. Though I don't know you personally, your words could as well be mine, so I know whereof I speak. You are not missing anything by choosing not to run in the rat race. Retirement is overrated. I will regret everything on my death bed, but that would be true no matter how I lived. To face its closure is regret enough.

The "infinite backlog of work" is the problem. It will be used as an excuse to ding you on your next performance review. Th fact that companies are structured to ensure that there is always too much work for anyone to do is what leads to stress and burnout.

To this day I still actually have a physical reaction of nausea when hearing the numbers 162 and 163.

<Immediately clicks to Calculus>... "Spivak". Yup, list is legit.

What a wonderful summary of the benefits this change has wrought. I feel the same, but haven't maximized the opportunity as much as you have! I might head out to listen to some birds right now...

Most people don't learn this lesson till the first layoff

You hit the nail on the head. While I definitely feel for all the early-career folks out there who are genuinely missing out on the camaraderie of working together, I urge all of you not to fall into the trap of thinking that your employment relationship is anything more than mercenary.

They will cut you without a moment's hesitation. Buying into the "we're a family" propaganda makes them rich and makes you dependent on them. Use the learnings of the Pandemic to redefine the future of work on your own terms. It's time we all flipped the script. We're the ones producing value, and from now on we're going to do it on OUR terms.

Fundamentally this is about respect for the autonomy of my time. I am paid to deliver value. Sometimes that is best served by direct in-person collaboration, but not every single day. The deal I offer to prospective employers is this: I do my best not to be a blocker when I can't be physically present, and you realize that I am a whole human being who does not exist solely to pad your bank account.

Pre-pandemic, I had a CFO say to me "I don't pay for all this expensive office space so you can work from home." I had a CTO tell me "WFH sends a message to your reports that you don't want to be here." I was too obsequious at the time, not to mention afraid to lose my job, to push back, so I relented despite a punishing balance of childcare when my son was very young. But today I call bullshit on this ancient relic of Taylor's "scientific management" peddled by the HBR-reading neocons populating finance departments all across tech. A job that requires my ass to be in a seat is not a job, it's a prison.

To address some of the skepticism about when and where javascript would be appropriate in data science, would you want to fit a logistic regression model in javascript? Probably not, but to build a solver that takes model outputs and visualizes the changes in predicted probabilities based on different combinations of variables? This is definitely where javascript would make sense. Visualization, dashboards, reporting, and exploratory analysis are all ripe domains for developing rich responsive UIs. Basically, any layer where you have a data-to-human interface can be leveraged with javascript.

There is a lot of great work happening in this space already. In the R world for example, shiny makes heavy use of js to the point that you often can't tell where R code ends and javascript begins. Plotly's Dash provides bindings for R, Python, and Julia. Personally, as a data scientist, I have been excitedly learning React because it really rips the landscape wide open for all the use cases I mentioned above. It then makes sense to have libraries that give JS users a good data model and can do most of the same numerical computation that we'd be doing in other languages. Again, you probabaly don't want to do serious numerical work in js, but remember people said that about Python ten years ago too.

I love the framing of this book, because I want more data scientists to start thinking about the presentation of data and spark some bits of ingenuity to make datasets and model outputs accessible to non-data scientists. Data scientists should be the ones writing the tools that interface data with humans because of their domain knowledge. But this is a different skillset and usually the work of SW engineers. Of course engineers can also have great data intuition too, but I really do encourage data scientists to develop their front end skills, it's well worth it.

Hi HN, I started working with the ANES data to develop a tutorial series on How to Analyze Data. The dataset is super interesting, very topical, and a very high quality source of data on US political opinions. As part of exploratory analysis I generated a massive deck of crosstabs to compare likely Biden voters with likely Trump voters (focusing on the pre-election survey which fielded from mid-August to just before election day). I wanted to share for anyone interested, along with the source code to make working with this dataset easy for anyone unfamiliar with it.

The rendered report link is the url in the title, and the source code is on github here: https://github.com/czep/HTAD

I flirted with this world in the mid 90s. The workload is intentionally set well beyond what one person can reasonably handle. It is designed to force you to a level of hyper-efficiency without compromising accuracy. But the primary function is as a symbolic hazing ritual. Director told me I will do nothing else for 4 years, but if you make it there you will have 7 figures annually for the rest of your life. I said no thank you and never regretted it, but there is no shortage of people who will gladly take that deal.

no three-hundred-and-sixty degree reviews, no stack ranking, no OKRs, no skip-levels, no numerical grades

Sign me up now! This is so refreshing I would happily take the pay cut to be rid of the toxic by-products of cutthroat review culture. I seriously hope this gains traction because it's about time we rethink the kind of work life we really want to have, one that is organized to serve the people who do the work.

As long as data science remains subservient to Product, the inevitable result is cherry picking and post hoc rationalizing of some PM's intuition. To blame is the "embedded" model in which DS works within (and takes direction from) a cross functional team driven by a PM.