Moved to Garage, it's actually pretty easy to run and use.
Would be even nicer if the official Docker image would support initializing a default bucket and access key from env variables instead of having to exec into the container and follow https://garagehq.deuxfleurs.fr/documentation/quick-start/ but that's not a dealbreaker.
Note: I only needed the single-node install, it was either this or SeaweedFS. Also used MinIO and Zenko in the past, but even the latter seems pretty much dead.
As lines of code become executable line noise, I swear that we need better approaches to developing software - either enforce better test coverage across the board, develop and use languages where it’s exceedingly hard to end up with improper states, or sandbox the frick out of runtimes and permissions.
Just as an example, I should easily be able to give each program an allowlist of network endpoints they’re allowed to use for inbound and outgoing traffic and sandbox them to specific directories and control resource access EASILY. Docker at least gets some of those right, but most desktop OSes feel like the Wild West even when compared to the permissions model of iOS.
> And that message would be "We have a product so valuable/useful that not even their weak ideals and moral obligations could keep them away!"
Who knows, maybe within those 30 days you find that other offerings are good enough for your needs - I've largely moved over to Anthropic's Max subscription for all my needs, I don't even need Cerebras Coder anymore because Opus 4.6 is just so good.
In an imaginary world, this would be a precursor to Anthropic coming to EU in a greater capacity and teaming up with Mistral, eventually leading to similar innovation and progress that DeepSeek forced upon the West, benefitting everyone in the long run. They seem to have the morals for it and the respect for human rights and life given their recent announcement (after some backtracking), unlike OpenAI. Sadly, that's not the real world.
> Amazing how you can just deflect any criticism of LLMs here by going “but humans suck too!” And the misanthropic HN userbase eats it up every time.
I think it's rather people trying to keep grounded and suggest that it's not just the hallucination machine that's bad, but also that many doctors in real life also suck - in part because of the domain being complex, but also due to a plethora of human reasons, such as not listening to your patients properly or disregarding their experiences and being dismissive (seems to happen to women more for some reason), or sometimes just being overworked.
> You simply would not be able to replace a real doctor with an LLM and get desirable results.
I don't think people should be replaced with LLMs, but we should benchmark the relative performance of various approaches:
A) the performance of doctors alone, no LLMs
B) the performance of LLMs alone, no human in the loop
C) the performance of doctors, using LLMs
Problem is that historical cases where humans resolved the issue and not the ones where the patient died (or suffered in general as a consequence of the wrong calls being made) would be pre-selecting for the stuff that humans might be good at, and sometimes wouldn't even properly be known due to some of those being straight up malpractice on the behalf of humans, whereas benchmarking just LLMs against stuff like that wouldn't give enough visibility in the failings of humans either.
Ideally you'd assess the weaknesses and utility of both at a meaningfully large scale, in search of blind spots and systemic issues, the problem being that benchmarking that in a vacuum without involving real cases might prove to be difficult and doing that on real cases would be unethical and a non-starter. And you'd also get issues with finding the truly shitty doctors to include in the sample set, sometimes even ones with good intentions but really overworked (other times because their results would suggest they shouldn't be practicing healthcare), otherwise you're skewing towards only the competent ones which is a misrepresentation of reality.
The fact that someone would say stuff like "Doctors are more like machines." implies failure before we even get to basic medical competency. People willingly misdirect themselves and risk getting horrible advice because humans will not give better advice and the sycophantic machine is just nicer.
> I think it's rather people trying to keep grounded and suggest that it's not just the hallucination machine that's bad, but also that many doctors in real life also suck
No, you see this line or argumentation on every post critical of LLM's deficiencies. "Humans also produce bad code", "Humans also make mistakes" etc etc.
> No, you see this line or argumentation on every post critical of LLM's deficiencies. "Humans also produce bad code", "Humans also make mistakes" etc etc.
So your reading of this is that it's a deflection of the shortcomings?
My reading of it is that both humans and LLMs suck at all sorts of tasks, often in slightly different ways.
One being bad at something doesn't immediately make the other good if it also sucks - it might, however, suggest that there are issues with the task itself (e.g. in regards to code: no proper tests and harnesses of various scripts that push whoever is writing new code in the direction of being correct and successful).
Hey that seems pretty cool! No doubt it's gonna be a way to either collect more info of successful devs or maybe just upsell stuff after those 6 months are over, but it's something!
I went for their 100 USD paid tier and it's honestly been immensely useful (Claude Code with the desktop UI with multiple parallel tasks), I've done more and with better quality in the past few weeks than others do in a month - maybe I just got lucky with the domain but it really is a force multiplier and I'm working on like 4 projects in parallel at work and am crushing it, being overworked aside.
Finally I also have enough capacity for various side projects and utility tools/scripts, or at least I will until I burn out, but that's not really the fault of the tool, rather the amount of work.
Being able to throw the latest Opus model at every problem is also really, really nice. Way better than any of the slop before.
Feels like they’re leaving a lot of money on the table and inviting existential peril by not bending the knee to the current Great Leader.
It does feel like what anyone sane should do (especially given the contradictions being pointed out and the fact that the technology isn’t even there yet) but when you metaphorically have Landa at your door asking for milk, I’m not sure it’s smart.
I feel like what most corpos would do, would be to just roll along with it.
In isolation, I think it's cute and silly - something to write about in a blog, have a chuckle about, and to have a nice sort of gimmick/ceremony within the company. Maybe a few data points towards studying or keeping track of how the model writing style changes over time. Nothing wrong with that.
> delusions of people who ramble about model consciousness
On one hand, it's interesting how the technology has advanced to where it essentially passes the Turing Test, often just because of how much people choose to anthromorphize it. Sadly, putting that in context, yeah, that's a bit unfortunate too, given how some of those interactions become unhealthy.
I haven't seen people often refer to the stack in the site as a "microservices architecture": databases, key-value stores, message queues, object stores and other specialized solutions are typically viewed in a category of their own, orthogonal to what the application architecture is, unless you'd have for example 3 different PostgreSQL instances for the same project.
From what I've seen, even if you'd have PostgreSQL, Redis, Garage (S3) and RabbitMQ, that might still be considered a monolith if you just have a singular deployment unit for your app (even if scaled to multiple instances). It'd still be a distributed system, sure, much the same way how an app that's connected to a singular DB is (just fewer points of failure), you'd just be separating the concerns across specialized solutions.
Aside from that, you can get pretty far with just PostgreSQL alone, but sometimes it's nice to have the more specialized solutions because the libraries and APIs that they have are optimized for working within a particular domain - for example, object storage easily letting you handle file uploads, downloads, Range requests, presigned URLs and so on, so you don't have to build anything yourself or think that much about what would be the best way to store a bunch of binary data in a RDBMS.
On the flip side, if you can somehow avoid the pain of trying to put everything in PostgreSQL, the idea of having a single source of truth for all of the concerns in your app is pretty darn nice! Might as well take a step further and get rid of the front-end/back-end altogether and make the DB serve web requests (or use an embedded DB within your app process), but that gets dangerously close to the mess that was Oracle Forms or that is Oracle ADF.
Would be even nicer if the official Docker image would support initializing a default bucket and access key from env variables instead of having to exec into the container and follow https://garagehq.deuxfleurs.fr/documentation/quick-start/ but that's not a dealbreaker.
Note: I only needed the single-node install, it was either this or SeaweedFS. Also used MinIO and Zenko in the past, but even the latter seems pretty much dead.
reply