It's not as elegant as PRQL, because of course it's bolted onto the existing SQL syntax, rather than a redesign from scratch. But it has a big name behind it, and it's actually running in prod in Google Cloud... so it might have more momentum.
Indeed we're doing fewer new features (and haven't posted to the posts page in a long time, as you noticed).
But it's still maintained, folks are still using it, if anyone finds bugs in simple-to-moderate queries then we'll fix them.
LLMs probably took a bit of the wind out of our sails for making this "the new standard". But I still think it's a really nice language and interface; if the world changed again such that it became more widely useful, I'd jump to spending lots of time on it again.
Thank you for building and maintaining PRQL! I'm surprised to hear that growth stalled due to LLMs.
I just found out about PRQL yesterday! I was looking for a query language that is more token efficient and easier to reason about for LLMs than SQL.
PRQL looks amazing for data analytics agents. Our first few test are quite promising.
I also really appreciate the python bindings. We don't give our agent direct access to the database, we only provide the schema information. The python api makes it super easy to convert a query into an AST, which lets us do some basic offline validation of table names, etc.
I'm sad to hear that about LLMs. I sometimes wonder if the software world is going to be "locked into" our existing languages, because it's what the LLMs can work with.
But also maybe that will change — LLMs can learn new languages faster than people, and can _write_ new languages much faster than people. So wide confidence bounds for the future!
Matching SQL in features is very hard, especially if you also want to make it more sane and more powerful at the same time while also wanting to be able the generate valid SQL from your syntax. So I am not surprised that it stalled out.
Sure, but they never intended to support everything you can do in SQL. For example, they say on the Roadmap page that they're only going to support SELECTs -- there won't be a PRQL way to do an INSERT, UPDATE, etc.
When it comes to SQL it's the select that's by far the majority of the work though, the hard work with mutating operations is on the database implementation, not really the syntax or query plan
I've only listened to the French revolution, but it was absolutely electrifying. I would listen on my car ride home -- one day I burst through the door to my house and yelled "CHARLOTTE CORDAY MURDERED MARAT!!!" at my wife like it was breaking news.
I would also say (d) make it clear to the Soviet Union that, if they invaded West Germany, it would not just be an intra-European war; it would be a world war.
> If you need granular control over every tiny aspect of your container orchestration — network policies, pod scheduling, resource quotas, multi-tenant isolation, custom admission controllers, autoscaling on custom metrics — Kubernetes gives you knobs for all of it.
> The problem is that 99% of teams don't need any of those knobs.
I keep hoping for a Docker Swarm revival. It's the right size for small-to-medium-size deployments with normal requirements.
Every enterprise team (at least who are in B2B business) needs this. The number of security clearances (zero trust boundary), security compliance is must. May be in B2C space where you might not need that depending upon how secure you wanna be based on what data you hold
Yeah I was trying to give the post a serious consider, but the author just flatly dismissed network policies as not needed, suggesting that we just make new overlay networks for every set of containers that need to communicate. This post really doesn't resonate with me, even though I am on a small team using k8s in a small company.
ECS Fargate is basically this on AWS. It’s just not cloud agnostic. But Swarm itself while being cloud agnostic is a proprietary product as well, so you still get the lock in, just at a different layer
IIUC, this was a finding when they ran the Polis experiments in Taiwan: when you map the arguments of the different sides, there are actually large areas of agreement. In other words, the median person who disagrees with you is a "potential common ground" guy, not a "planet Zargon" guy.
What I don't understand about Polis though is who is creating these less biased polls full of unbiased positions that people can vote on? It takes a lot of intelligence and wisdom to even formulate a question that isn't tainted by layers and layers of political innuendo. You can't just put something like "Do you believe in the rights of the unborn child?" into a system like this and expect quality outcomes.
I guess the theory is that you put the entire spectrum of positions on the line which allows fully biased positions on each end to exist. Then biased people on both ends will vote on slightly less and less biased positions that they still agree with and you'll see the true shared positions. But I still think that if you don't have a perfectly equal number of positions to vote on for each side you'll end up with the same problem we already have in society, people are being given biased questions not necessarily by strength but by amount. Therefore they will subconsciously and consciously conclude that the world wants them to be more towards the position that had more questions presented.
Many (most?) issues don't fit on a single dimension. Using your example, people hold positions that include "Absolutely!", "Yes, but also the rights of the mother.", "Yes, but I won't impose my beliefs on others.", "No, but I don't think people who feel otherwise should be forced to pay for abortions through taxes.", and many others.
In addition to the problem with biased questions you note, there are often built in assumptions that make yes or no responses impossible.
I find that the median person who disagrees with me, actually agrees with me, but I accidentally triggered their social media PTSD and they flagged me as an enemy because I didnt slavishly polish their preferred set of boots.
That too is a major problem, in theory you could be posing fine questions but they are already politically or socially tainted so it's game over before it even started, you will get zero actual new thought from the person you asked.
Trouble is, the "large areas of agreement" can be pretty superficial. You can probably find broad agreement across the entire political spectrum about "cutting government waste", but it turns out that who is tasked with doing this and the low level details of what gets cut matter a lot more than the basic principle.
> The thing nobody's talking about is that none of this AI automation works without Backstage doing the boring work of tracking who owns what, how things build, and what depends on what.
Do you mean that Backstage has the metadata like what services call which other services, and AI needs that to make changes safely? Sounds helpful to both AI and human developers ;-)
I mean, presumably some future Democratic administration will reinstate the rule. But with this precedent set, this might become a switch that turns on and off every time the political winds change. When Republicans are in power, the US will do nothing at all to fight climate change. When Democrats are in power, they will belatedly try to undo the damage.
And of course there will be knock-on effects from other countries. Why should (for example) Mexico do anything at all to fight global warming, when the US (which is much richer, and a much larger polluter) declines to help?
China is building so fast it won’t matter. They destroy 1M barrels a day in global oil consumption for every 24 months they build EVs at current production rates (which continue to increase), for example. ~25% of global light vehicle sales are EVs as of 2025, ~50% in China (the largest auto global auto market). The world is approaching 1TW/year of global solar PV deployment. Solar and storage are the cheapest form of generation, and will only continue to decline in cost.
Consider it a case study in governance failure. The US' failure is China's opportunity, and they appear to be taking it.
I hope you are right but China needs to block the US' influence on international policy. Its strange to see countries adopt policies in a direction the US has obviously chosen to screw everyone else instead of going so far into free trade the US can't stay relevant.
Certainly, diplomacy is the art of saying "good dog" until you make it to the rock [to throw at the dog] as the saying goes. Counterparties are still headed towards the rock. Once they have sufficiently decoupled from the US, their policy and destiny are in their hands. Easier to say no to a bully when you're beyond their reach and the harm they can cause is immaterial.
It sucks that Congress don't do their job of making reasonable laws. I hate that the executive and judicial branches have to do so much work that should be done by Congress.
> I hate that the executive and judicial branches have to do so much work that should be done by Congress.
In recent years the Supreme Court has turned against the use of regulatory agency rule making authority to stretch the meaning of older statutes and accomplish what Congress is too gridlocked to do. Most notably was the 2022 decision striking down Obama-era EPA power plant carbon emission limits (https://en.wikipedia.org/wiki/West_Virginia_v._EPA), but there are many other decisions in a similar vein (e.g. overturning Chevron), and more coming down the pipe (see https://www.scotusblog.com/2024/10/supreme-court-allows-epa-...).
Between SCOTUS decisions limiting how older statutes can be reinterpreted to encompass global warming on the one hand, and fundamental economic incentives on the other (the West Virginia decision didn't result in a rush back to coal), this move by the Trump administration is unlikely to change the course of things, except to perhaps spur Congress to involve itself more heavily one way or the other.
Rather than handwringing, the left needs to finally accept that relying on lawsuits and aggressive Federal regulatory agencies, rather than the ballot boxes (plural--not just the presidential election), to enact their social and environmental policies is no longer viable. But it's going to be a difficult change because the Democrats sacrificed a ton of grass roots support (real, substantive support, as opposed to professional class and social media popularity contests) as they came to rely almost exclusively on imaginative legalistic and technocratic solutions, an evolution that started decades before the courts took their sharp conservative turn.
I, for one, invite diminished environmental regulatory agencies. In so far as it concerns global warming, renewable energy, and land use (e.g. mass transit, housing, etc), they've become impediments much more than enablers of (net) environmentally friendly change. What does it matter if an agency favors one set of policies over another when it takes years if not decades for projects to make it through the thicket of red tape? For energy policy specifically, the economics favor renewables, so less regulation can only hasten the transition.
The administrative state worked extremely well before Trump dismantled it.
Do you really want Congress to use legislation to make decisions about day to day technical rules and regulations that they quite obviously have no chance of understanding?
The whole point of the administrative state is to put non-political experts in charge of hashing out the specifics of rules and regulations while Congress legislates the broad process of making those rules.
An analogy to “making Congress do it” would be like if you had to raise every pull request you wrote to the board of directors of your company and check if they were okay with it. That is insanely inefficient and your board of directors would make the wrong decision most of the time. Instead, your company’s board of directors hires competent people and sets the general goals of the company and directs everyone to work toward them, trusting them with the implementation details and putting in places systems that ensure good performance is rewarded.
In the past we just trusted presidents to operate at some bare minimum level of basic good faith that they were non-traitor citizens who actually wanted this country to succeed, rather than being completely apathetic to the future and viewing American society purely as an asset to exploit.
The idea of a president who would make the country worse on purpose, going as far as making it worse for the wealthy in addition to common people, was unheard of.
But then we elected the New York Russian mafia’s real estate guy. And now we have found out that it’s very likely that his best friend’s sex trafficking operation was potentially used to compile kompromat [1]. The probability that the Russian intelligence apparatus is directly instructing Trump to sabotage the geopolitical position of the USA is astronomical. Part of that sabotage is almost certainly the dismantling of the administrative state.
Call it a conspiracy theory if you want, but we’ve crossed this conspiracy theory bridge many times with Donald Trump and he has never really given any of us a good reason to not trust the idea that “oh, it’s actually worse than we thought…”
Bold of you to assume that any other party will ever be in power - why do you think there's been talk about taking control of the elections? It's certainly not to make it easier to depose Republicans.
> Just adding another service is easy, but making them work seamlessly together is much harder.
This argument in favor of PaaS makes a lot of sense to me. A PaaS vendor might say, "sure, you could set up your own observability layer, configure storage for it, troubleshoot that one service that can't connect to the API's endpoint, etc. Or you could just check the "observability" box on our dashboard and boom, there it is."
I'm currently at a company that really likes setting everything up from scratch. And the amount of time my team has spent building out a platform, as opposed to providing business value, is really sobering.
Yeah that same experience at our previous startup was one of the main reasons why we started Modelence. We got to a point where we had a dedicated 3-4 cloud platform engineers (out of the 20 engineers total) working full time only on things like observability, alerts, performance, cloud deployments, etc, none of which was specific to our product.
And when you take a step back and think for a moment, it just doesn't make sense that you have to run a whole separate team for something that's pretty much reusable across products.
I often wonder: if there were more parties in America, how many of these independents would register with one of them?
In other words, if there's a record-high number of independents, how much of it is "market failure" where people want more options than the usual 2 parties?
At least in Bulgaria, 60% of people don't vote. Granted that we aren't that socially active as a country, I think that any amount of parties wouldn't bring much difference. Until the system starts working again, as in convicting criminals, preventing policymakers from benefiting from "timely" investments and overall bringing the wealth gap down, I don't think that you'd be able to reach all of the people
Totally agree. But I mean: supposing that we had a voting system where minor parties were more viable and didn't play spoiler. What % of independents would register as Greens, or Libertarians, etc, in that system?
It's not as elegant as PRQL, because of course it's bolted onto the existing SQL syntax, rather than a redesign from scratch. But it has a big name behind it, and it's actually running in prod in Google Cloud... so it might have more momentum.
[0]: https://cloud.google.com/blog/products/data-analytics/simpli...
reply