Generally, with the regular in-IDE agents you have the ability to easily intervene, correct and live-check. Considering the high fail rate of agents (depending on software complexity of course), that's required if you want to get anything done and not be slowed down by it.
Otherwise you'd always have to context switch, consider which git state it's actually working from, etc. - rather than just letting the code directly before you change in your IDE.
It's significantly lower cognitive load and has a higher success rate, in my experience.
But, of course: Highly depends on the software being written and the general code infrastructure.
There is no reason such a state would have to set things up this way.
As an example: you probably know that germany has socialized healthcare. It is, however, not implemented as a single-payer model. Instead there are tons of different insurances competing with each other, while having a highly regulated floor of what they MUST offer.
Is the model perfect? Hell no, it has tons of issues - though overall it's pretty solid. My point is just that social policies and "no internal competition ever" does absolutely not have to go hand in hand. There is a massive middle ground.
See: social democracy as a concept and in its current implementation.
Germany has a dual healthcare system with both public (GKV) and private health insurance (PKV) options. About 10% of residents use private health insurance.
Problem here being that those terms aren't used as defined in regular discourse. Language changes and casual use differs from academic use.
When on an american-centric board anybody writes about "communism", I assume they refer to anything from marxism to stalinism to socialism to democratic socialism to social democracy up to anything non-hyper-capitalist. Not great, but sadly something to be taken into consideration.
Especially when looked at in context - parent was criticizing the EU initiative by essentially claiming something like that leads to a kind of monoculture like in a planned economy reminiscent of "communism", here probably meaning stalinism, from what I assume is a radical libertarian position. Which tells me the person is likely american, implying a rather ... minimal awareness of the nuance here.
Please, look at the actual comment chain and it should be rather trivial to make out what everybody is talking about. Does your comment really add value here?
The fundamental reason why terms are confused is because malicious
actors (bad faith) intentionally set up to at best muddy the waters
and at worst associate things they fight against with the worst
adjectives. That doesn’t reflect on anyone present right here, just a
process that took decades.
I don’t care if someone “doesn’t know” about the nuance when they
breathlessly throw back with One Cereal For Everyone Decided By The
State. Come on.
> Please, look at the actual comment chain and it should be rather trivial to make out what everybody is talking about. Does your comment really add value here?
I can understand that you think my reply is pedantic noise. That’s
simply because we have different goals and things that we intend to
communicate. I’m content with setting the record straight. You
apparently want to calmly explain the difference between apparent
Stalinism and Bernie Sanders-style Socialism.
I think I am able to make out what people are talking about. But you
can’t seem to, right in this context, imagine that we all have different
goals ourselves about what we wish to get out of commenting here.
>But inference costs are dropping dramatically over time,
Please prove this statement, so far there is no indication that this is actually true - the opposite seems to be the case. Here are some actual numbers [0] (and whether you like Ed or not, his sources have so far always been extremely reliable.)
There is a reason the AI companies don't ever talk about their inference costs. They boast with everything they can find, but inference... not.
I believe OP's point is that for a given model quality, inference cost decreases dramatically over time. The article you linked talks about effective total inference costs which seem to be increasing.
Those are not contradictory: a company's inference costs can increase due to deploying more models (Sora), deploying larger models, doing more reasoning, and an increase in demand.
However, if we look purely at how much it costs to run inference on a fixed amount of requests for a fixed model quality, I am quite convinced that the inference costs are decreasing dramatically. Here's a model from late 2025 (see Model performance section) [1] with benchmarks comparing a 72B parameter model (Qwen2.5) from early 2025 to the late 2025 8B Qwen3 model.
The 9x smaller model outperforms the larger one from earlier the same year on 27 of the 40 benchmarks they were evaluated on, which is just astounding.
Anecdotally, I find you can tell if someone worked at a big AI provider or a small AI startup by proposing an AI project like this:
" First we'll train a custom trillion parameter LLM for HTML generation. Then we'll use it to render our homepage to our 10 million daily visitors. "
The startup people will be like "this is a bad idea because you don't have enough GPUs for training that LLM" and the AI lab folks will be like "How do you intend to scale inference if you're not Google?"
Recently people built a super-lightweigt alternative, named copyparty[0]. To me that looks like it does everything people tend to need without all the bloat.
I think "people" deserves clarification: Almost the entire thing was written by a single person and with a _seriously_ impressive feature set. The launch video is well worth a quick watch: https://www.youtube.com/watch?v=15_-hgsX2V0&pp=ygUJY29weXBhc...
I don't say this to diminish anyone else's contribution or criticize the software, just to call out the absolutely herculean feat this one person accomplished.
I have tried to run micro https://micro-editor.github.io/ on my phone but this is some other beast if someone is running tmux and vim on their phone
I have found that typing normally is really preferably on android and usually I didn't like having to press columns or ctrl or anything so as such since micro is really just such a great thing overall, it fit so perfectly that when I had that device, I was coding more basic python on my phone than I was on my pc
Although back then I was running alpine on UserLand and I learnt a lot trying to make that alpine vm of sorts to work with python as it basically refused to and I think I learnt a lot which I might have forgotten now but the solution was very hacky (maybe gcompat) and I liked it
I do a lot of development and sysadmin stuff on phones and tablets, to a large degree due to PentiKeyboard. It helps a lot to see the entire screen and have all the usual keyboard sends that a regular, physical keyboard has.
I'm using micro in termux on my android. The keyboard of termux is quite adapted to the CLU bindings (ctrl-f, ctrl-s,...). My main use is to take notes, though I bought a physical small bluetooth keyboard perfect for making a little bit of python scripts from time to time.
I must admit that coding in vim on a kinda big project on a smartphone is really impressive.
This is not an alternative as it only covers files. Mind what is in the article: "I like what Nextcloud offers with its feature set and how easily it replaces a bunch of services under one roof (files, calendar, contacts, notes, to-do lists, photos etc.), but ".
For us Nextcloud AIO is the best thing under the sun. It works reasonably well for our small company (about 10 ppl) and saves us from Microsoft. I'm very grateful to the developers.
Hopefully they are able to act upon such findings or rewrite it with go :-). Mmh, if Berlin (Germany) wouldn't waste so much money in ill-advised ideology-driven and long-term state-destroying actions and "NGOs" they had enough money to fund 100s of such rewrites. Alas...
Why should Germany be wasting public money on a private company who keeps shoveling more and more restrictions on their open-source-washed "community" offering, and whose "enterprise" pricing comes in at twice* the price MS365 does for fewer features, worse integration, and with added costs for hosting, storage, and maintenance?
* or same, if excluding nextcloud talk, but then missing a chat feature
It makes a lot of sense for Germany to keep some independance from foreign proprietary cloud providers (Microsoft, Google); Money very well invested imo. It helps the local industry and data stays under German sovereignity.
I find your "open-source-washed" remark deplaced and quite deragoraty. Nextcloud is, imo, totally right to (try to) monetize. They have to, they must further improve the technical backbone to stay competitive with the big boys.
At the very least their app store, which is pretty much required for OIDC, most 2FA methods, and some other features, stops working at 500 users. AFAIK you can still manually install addons, it's just the integration that's gone, though I'm not 100% sure. Same with their notification push service (which is apparently closed source?[0]), which wouldn't be as much of an issue if there were proper docs on how to stand up your own instance of that.
IIRC they also display a banner on the login screen to all users advertising the enterprise license, and start emailing enterprise ads to all admin users.
Their "fair use policy"[1] also includes some "and more" wording.
This may come as a surprise to you, but there are organizations, for example German municipalities, that have more than 500 users but can't afford to start pumping tens or hundreds of thousands per year into a file sharing service. Nextcloud themselves recognize this and offer 95%+ discounts to edu, similar to what Adobe, Microsoft, and Git[Hub,Lab] are doing.
There is no way it’s going to be completely rewritten from scratch in Go, and none of whatever Germany is or isn’t doing affects that in any way shape or form.
Actually, it's already been done by the former Nextcloud fork/predecessor. OwnCloud shared a big percentage of the Nextcloud codebase, but they decided to rewrite everything under the name OCIS (OwnCloud Infinite Scale) a couple of years ago. Recently, OwnCloud got acquired by Kiteworks and it seemed like they got in a fight with most of the staff. So big parts of the team left to start "OpenCloud", which is a fork of OCIS and is now a great competitor to Nextcloud. It's much more stable and uses less resources, but it also does a lot less than Nextcloud (namely only File sharing so far. No Apps, no Groupware.)
Thanks for sharing this, I've been wanting to look at private cloud stuff but it was all written in PHP. It looks like OpenCloud is majority Go with some php and gherkin, which is a step in the right direction.
I have OpenCloud working on my home server, and it features integration with the Collabora suite of software for office apps. Draw.io is also already supported.
They offer a Docker compose file that sets up Collabora for you, but I can't find anything info on other apps, let alone integration. Where can I see what they support?
You're right, it was my mistake. The docker compose file can set up Collabora for you and allows you to open documents from inside OpenCloud by opening the file in an embedded Collabora view. Likewise, Draw.io works in a similar fashion, opening a view to embed.diagrams.net. Underneath it's just hosting the files and offloads the operations to other apps. It's convenient, but not particularly sophisticated.
There are no "Apps". It's not a universal App platform like Nextcloud. It's just file sharing (and optionally a Radicale calender server via Environment Variable but without UI). There's optional plugins to open vendor specific files right in the browser.
OCIS does only a small part of why people deploy NextCloud. I have run it, it’s great, but it’s not a replacement for the full suite nor is it trying to be.
It makes perfect sense to me that nextcloud is a good fit for a small company.
My biggest gripe with having used it for far longer than I should have was always that it expected far too much maintenance (4 month release cadence) to make sense for individual use.
Doing that kind of regular upkeep on a tool meant for a whole team of people is a far more reasonable cost-benefit analysis. Especially since it only needs one technically savvy person working behind the scenes, and is very intuitive and familiar on its front-end. Making for great savings overall.
I think what you described is basically ownCloud Infinite Scale (ocis). I haven't tested it myself but it's something I've been considering. I run normal owncloud right now over nextcloud as it avoided a few hiccups that I had.
> NOTE: full bidirectional sync, like what nextcloud and syncthing does, will never be supported! Only single-direction sync (server-to-client, or client-to-server) is possible with copyparty
I watched the video, too, and while amazing, it's the poster child for feature creep. It starts out as a file server, and at some point in the demo it's playing transcoded media and editing markdown??
It's an amazing piece of software. If only the code & the configuration was readable. It's overly reliant on 2-3 letter abbreviations, which I'm sure has a system, but I haven't yet been able to decipher.
Personally, the only thing I need is stable clients on both desktop and mobile with bidirectional sync. Copyparty seems really cool, but it explicitly does not do that.
Have you considered syncthing? There’s shiny new and super cool Sushi Train (or Sync Train by other name) app for iOS (I wish the author would make it a paid app, so much I like it!): https://github.com/pixelspark/sushitrain
Not affiliated, but a very happy user.
I mention iOS, because that was what I needed personally, as there was syncthing for Android since forever.
This comment is always so strange to me - do you really, seriously believe that the people setting up the grids never thought about dunkelflaute? And I don't mean that in an attacking way, I'm genuinely curious about your thoughts there.
Like, yes, we're aware. At least in the german south we have the opposite problem right now. We are getting negative electricity prices (you get paid for taking some) more often because we have more electricity than we can use due to solar, at least during the day. Proper power storage is being built at this very moment all over the country.
Aside from dunkelflaute, the wind is statistically stronger when solar power generation is low, so at night and when it's super cloudy. And dunkelflaute is a couple days to weeks per year. (german perspective, don't know enough about the other countries' grids)
Regarding that problem in portugal, you misunderstood something there. The big 2025 power outage wasn't caused by clouds, it was an combination of localized blackouts and a sudden power _surge_ which caused a cascading failure which couldn't be stabilized by the conventional power plants even though on paper they had the capacity. How did you get the idea it had anything to do with "cloudy" weather?
RE "... dunkelflaute is a couple days to weeks per year..."
My guess is its VERY expensive to build the needed storage so the supply reliability matches the current reliability 99.99%? ? ( in my area there has never been any unintended power outages for several years ) Which is why its never been done?
Then again maybe people will be more tolerant of the situation. I've always though smart meters could always have a "mode" to reduce everyone's max demand to a small amount ...like a few hundred watts ...too help handle extended periods of dunkelflaute
Open cycle gas turbines are the perfect low CAPEX high OPEX backup. They are what we currently use to manage the once a year winter storm.
Force them to run on decarbonized fuels like ammonia, hydrogen, synfuels or biofuels (with decarbonized inputs) when even the backup needs to be decarbonized.
People setting up the grids answer to politicians. They do what they can within the constraints given by public policy. If public policy is completely idiotic, like the one in germany, there's no much they can do other than try to duct tape whatever they can.
Relevant reading: "Replication of Quantum Factorisation Records with an
8-bit Home Computer, an Abacus, and a Dog" by Peter Gutmann and Stephan Neuhaus [0].
Shows in a humorous way how the vast majority of quantum computing "records" are utter nonsense based on simplifying the factorization so far, that it turns into a problem on the difficulty level of "factorize 9" - _before_ running the experiment.
Journalists however tend to lack the knowledge to accurately represent that, resulting in nonsensical record claims.
Because "confidence" isn't just something a conscious being can have, it is also something text can simply convey, irrespective of the author. It's just how we perceive language and about what we feel, not about what the author intends.
Something semantic then, interesting. The words chosen and forcefulness, or persuasiveness,coupled with the doubletalk of repeating the query as an intro, okay, I get that. Thanks
Otherwise you'd always have to context switch, consider which git state it's actually working from, etc. - rather than just letting the code directly before you change in your IDE.
It's significantly lower cognitive load and has a higher success rate, in my experience.
But, of course: Highly depends on the software being written and the general code infrastructure.