binoct

Making incredibly complex and opaque pricing structures is not necessary for the providers to charge for and make a profit on their service. And being technically difficult is a lazy excuse. Cloud platforms have to solve many, much more difficult challenges to offer their services at all, they just don’t want to invest the time in more customer friendly billing because they expect it will result in reduced revenues.

I know your post isn’t explicitly defending the platforms, but the arguments they use feel transparently flimsy.

What does it mean to not be good at being a car?

It’s a terrific car if you want a great handling convertible at a low price point. As you said, it’s fun.

It’s not good at moving a lot of people or things, or going really fast, or lasting for 500k miles, or being as cheap as possible. Other cars are good at some of those things, but what is the Miata missing that is otherwise required to be a ‘good’ car?

> it just misses a base load component right now.

I’m sympathetic to the intent of your argument, but man, that’s a hallava gotcha there. Other than nuclear there isn’t a compelling story for a base load source of green energy. And all the new-gen nuclear tech is a long shot for safe, cost efficient, rapid scaling right now.

It’s more that unwanted disclosure of sensitive business numbers is a product-class limitation at this time. While undoubtedly some models are better than others, expecting them not to leak information in high stakes output is unseasonable. Seems like you agree, since you are reading everything.

Jumping to another company’s offering because of one instance doesn’t seem like that’s going to meaningfully change your experience. I’m guessing there was more to it, but that’s how it came across in your first post.

I hope you don’t plan to cut back on reading legal documents crafted by any LLM before executing them.

I don’t see any screaming and waving arms in the parent. It’s a passionate but well reasoned and carefully presented case. If that is to be dismissed as zealotry we’ll all need to stick to writing journal papers to avoid the label.

If anything your “copyright forever” drumbeat and incorrect read of “only being trained on copyrighted works” are more hyperbolic than anything you’re responding to.

Also, the models from China are only being released open weights as competitive positioning against the globally more used competition. If the Chinese companies had a dominant market position they’d be falling prey to the same forces that have caused OpenAI and Anthropic to behave the way they do.

The weights are in the meat

Thanks, that was interesting to learn, I’d never thought about how odd the phrase sounds as-is.

There’s a fascinating complexity to what constitutes constructive feedback, criticism, or dismissal. And to when it’s okay to provide one, the other, or none at all.

Here here. This was a joy to wake up to and wish I hadn’t stumbled into the comments.

A new bug appears, it’s in an encryption layer. You solve this by deciding to disable the encryption layer because user experience is better without the errors. You write it up as a recruitment piece for your engineering team.

There may be some good answers and lessons, but they didn’t make it into the article. Saying it’s on a cloud provider’s private network so encryption between your nodes isn’t necessary is a bold choice. Also, what happened to the root cause? Why did it start failing a week ago? Was a downgrade of the offending code not possible?

Not all bug investigations are worth really digging into. Sometimes the right call is to find any fix and move on. But all the nuance, judgement, implications, and lessons learned failed to make it into this post. And they are what make reading incident reports interesting for most engineers.

As pointed out in the article, naproxen is an NSAID like Ibuprofen, though slightly more COX1 selective. It likely has a somewhat lower risk of serious renal and cardiovascular events, but higher risk of GI bleeds. There are some studies that show little to no increase cardiovascular risk, but most do show some or even comparable to ibuprofen.

Convenience vs ibuprofen is a thing given the longer half life, but it still generally comes with similar risks. If you are taking anything for more than just an occasional headache, definitely discuss with a doctor, COX2 selectives like celecoxib may be a better risk profile and even more convenient.

(COX1 and COX2 selectivity loosely separate which systems get the brunt of the side effects)

Neither WebRTC or WebGL are remotely ‘useless’. Very fair though to say that you would prefer to have them disabled and/or whitelisted for certain sites.

Calling the industry largely a scam is pretty strong. Of course for the small minority of technically competent people with interest and time it will always be cheaper to do something yourself rather than pay a business to accomplish the same thing. But most people cannot/do not want to do it themselves, and regulations are there at least partially to help protect them against the house-fire-waiting-to-happen untrained handyman.

Sure a bunch of businesses opportunistically up-charge, some I'm sure are predatory, and there are obviously efficiencies to improve, but overall scam it is not.

Not really. The majority of data center water withdrawal (total water input) is consumed ("lost" to evaporation etc...) with a minority of it discharged (returned in liquid form). I believe it's on the order of 3/4ths consumed, but that varies a lot by local climate and cooling technology.

There's lots of promising lower-consumption cooling options, but seems like we are not yet seeing that in a large fraction of data centers globally.

Best comment by far. The post suffers from making good points about the lack of rigor and narrative nature of the book, but then does exactly the same thing to claim the opposite conclusion the book makes.

Thanks, you have helped at least me think a bit differently about this. I still believe primary energy is a valid way to look at the problem, but see more clearly how easily it can lead an uninformed audience to a bad conclusion.

And on heat pumps - it’s sad to reflect that even if we replaced all heating, it’s still only a couple % of the total rejected heat. There are few easy wins in this game, just many different ways we need to chip away at it.

Ah, seems like I was missing some context where the fossil fuel and anti-renewable folks have been using the term in arguments against trying to change.

I’m not sure of Smil’s politics but to be fair, there’s nothing in that quote that is inherently misleading. I can see through how others could spin it, and I’ll be more careful knowing the term has some politics behind it now. To me his argument in the article is that it’s not practical to expect a transition in a 25-year timescale, not that it’s impossible or not worth working on.

Heat pumps are a good example where the practice has been a lot harder than we might hope. Sure COP > 4 for heating is great, but the units are very expensive today, and in most of the US and Europe with sub-zero winter temps operate with much worse efficiencies, making them significantly more expensive to operate. I’m sure with effort those issues will improve, and major policy shifts can help mitigate some of the costs. But especially without a strong will today those changes are practically too far off for the 2050 target.

Primary energy sources are what they are, both your comment and the linked article seem to imply discussing them should lead to a deserved punch in the face. Can you help me understand why?

As far as I can tell, your link argues that if we overcome all the practical challenges (politics, resources, financing, technical innovation) and go all-electric for global energy, we only need ~1/3 as much input energy potential as we use today for the same useful work. That’s useful, but the hard part lies in those practical challenges. And the primary sources of global human energy use are a long way away from that goal.

So should we strive to get there? Sure. Should we be tactical about how? Yes. And the link seems to argue that as well. But is it reasonable to hit our 2050 goals based on the current global fossil fuel usage? Not really. So I’m really missing how this refutes Smil’s article, and why “primary energy” is such a stupid thing.

Details of this situation aside, the take "doing x is pointless because lots of other people are doing it" is a bit rough.

So there’s another force at work here that to me answers the question in a different way. Agents also massively decrease the difficulty of coming into someone else’s messy code base and being productive.

Want to make a quick change or fix? The agent will likely figure out a way to do it in minutes rather the than hours it would take me to do so.

Want to get a good understanding of the architecture and code layout? Working with an agent for search and summary cuts my time down by an order of magnitude.

So while agree there’s a lot more “what the heck is this ugly pile of if else statements doing?” And “why are there three modules handling transforms?”, there is a corresponding drop in cost to adding features and paying down tech debt. Finding the right balance is a bit different in the agentic coding world, but it’s a different mindset and set of practices to develop.

Thanks!

What does Germany use to manage microorganism growth in it's water distribution system? As I understand cloramine/chlorine is used to keep the small amounts of microorganisms that will always be present in water and pipes from growing into a problem while it travels/sits in the distribution system.

As I imagine it, Einstein would no be happy with fixing a couple bugs and making a state machine. Einstein would add a new unit test framework and implement a linear optimizer written with only lambdas to solve the problem and recommend replacing the web server with it as well. This is tongue in cheek but gets the idea across.

So this is a really good example of small sample size intuition being a big challenge. Fatalities happen on the order of billion miles driven - obviously people don’t come to that. Take a few thousand miles of positive experience sets a statistical floor on accident rates, but that is orders of magnitude away from how safe (or unsafe, depending on how you look at it) human drivers are on average. FSD and other, less capable L2 systems are amazing at paying attention in situations where humans fail, but also tend to have major limitations in places humans will largely do great most of the time. Your experience, as positive as it has been, doesn’t support the assertion that fatalities would decrease.

Launching in other cities with new problems gives experience dealing with new problems, and the meta-learnings transfer to better processes for adapting to new issues. But yeah, ice and snow are definitely major new environmental factors for New York (and DC, and many other places we are starting to see more serious testing).

Autonomous vehicles can and do take into account surface conditions, there’s not really any reason not to. There are pretty good generative models of the physics of vehicles with different surface conditions, and I imagine part of the data collection they are doing is to help build statistical of vehicle performance based on sensed conditions.

One of my favorite things to question about autonomous driving is the goalposts. What do you mean the “stated goal of full self driving”, which is unachievable? Any vehicle, anywhere in the world, in any conditions? That seems an absurd goal that ignores the very real value in having vehicles that do not require drivers and are safer than humans but are limited to certain regions.

Absolutely driving is cultural (all things people do are cultural) but given 10’s of millions of miles driven by Waymo, clearly it has managed the cultural factor in the places they have been deployed. Modern autonomous driving is about how people drive far more than the rules of the road, even on the highly regulated streets of western countries. Absolutely the constraints of driving in Chennai are different, but what is fundamentally different? What leads to an impossible leap in processing power to operate there?

Glad it worked for you. Not everyone with sleep apnea is an overweight alcoholic. Plenty of drugs help with diseases involving muscle tone, why is that so surprising?

Those who can tolerate the general CPAP experience, have bed partners who tolerate it, and don’t experience detrimental effects should absolutely use it when there are no comparable solutions. However there are lots of perfectly legitimate reasons why not everyone can, and having alternatives (which do also come with side effects) to consider is amazing for the community overall. It’s legitimately great that it sounds like CPAP treatment has been effective for you (as it has for me, mostly), but your comments end up sounding quite dismissive of the challenges faced by other patients.

Careful skepticism of new treatments is always warranted, but even if it only helps 5% of patients with OSA in absolute terms that’s a huge population impact.

The comments here are really interesting to read since there are so many strongly stated different definitions. It’s obvious “steaming” and “batch” have different implications and even meanings in different contexts. Depending on what the type of work being done and what system it’s being done with, batch and streaming can be interpreted differently, so it feels like really a semantic argument going on lacking specificity. It’s important to have common and clear terminology, and across the industry these words (like so many in computer science) are not always as clear as we might assume. Part of what makes naming things so difficult.

It does seem to me that push vs pull are slightly more standardized in usage, which might be what the author is getting at. But even then depending on what level of abstraction in the system you are concerned with the concepts can flip.

Scientific research doesn’t have to directly impact personal decision making to be useful and interesting. This study provides a data point suggesting a link between having tattoos and skin cancer - it’s certainly interesting to the medical field to better understand what increases the risk of cancer. A lot of research into smoking cigarettes and cancer also didn’t have much impact on people who decided to smoke, but it was also valuable knowledge.