HN user
shawn.fm
I like Rust and I'm an advocate of Rust, but this really isn't true (at least it hasn't been and I doubt much has significantly changed).
The syntax complexity and the ecosystem haven't been ideal for LLM development. And there have been publications on findings of LLM efficacy with different languages. Rust is most often towards the lower end of efficiency/correctness when benchmarked.
Approximately 5 years ago
This is a good take. I was interested in accomplishing my goals and had an interest in both Rust and Zig. Going in, Rust was already proven to meet my needs and I was exploring Zig. Everything being centered around anti-Rust and “better than Rust” without meeting my needs made it a non-starter, it got in the way of discussing the languages themselves.
I don’t like the Rust culture. There’s no better way to put it.
This is just so weird to me, because I would say the same about Zig.
I tried to get into Zig even chatted with Loris Cro when he was streaming. I was looking to explore what my Rust project could look like in Zig but there were features simply missing that I couldn't do without. The entire interaction was mostly about how bad Rust is and how I could just do something different in Zig (completely misunderstanding my ask, with little interest to explore my actual requirements).
I remember watching HN and seeing every time there was something Rust related trending, there was ALWAYS a post made shortly after trying to hype Zig and this went on for like 4 years.
I'm not a Rust contributor and I don't care for some of the challenges that come with Rust, but I love what it accomplished and I find it does it very well. Back then I found the Rust community had interest and respect for Zig, so the discourse was very much one sided.
Has the Zig team/community warmed up towards other systems-ish languages?
This is speaking from recollection engaging with programming communities years back - I found the Rust community to be interested and respectful of Zig but it didn’t seem to be well reciprocated, if anything much the opposite.
Would you mind giving these a try and let me know how they work for you? I’d imagine you would get better results and the latter will fit on a single GPU.
https://huggingface.co/easiest-ai-shawn/Qwen3.6-27B-ExCal-EX...
https://huggingface.co/easiest-ai-shawn/Qwen3.6-27B-ExCal-Mi...
Do be sure to use dflash and/or mtp for the draft:
Hey so... mind sharing findings? I have a QS8 :)
What is the target demographic? The specs seem... nice. Nothing particularly special compared to the likes of Lucid, etc.
The design though, it seems very... uninspired? It has hints of throwback in the design, but imo it does not have the look of luxury or sports car.
If you’re eyeing Prusa, that’ll probably be ideal. ...but it does look like Sovol is teasing an INDX alternative (I have a Sovol SV08, it’s a “good value” tinkerer printer based on the Voron)
And if you really want “open”, there’s isn’t much better than a Voron in that aspect.
I was a bit confused, because I've only known of the old models. Apparently there's been a refresh and they're no longer a few hundred dollars. FWIW they were not using one of the new models, they were using an old model which would have been maybe a few hundred on the used market at the time.
It's kinda wild to look and see even the old models listing for double the price they used to.
One coffee shop near me (since closed) had a Group 3 Slayer paired with a Super Jolly (but they also didn't know how to pull a decent shot).
For those unfamiliar, Slayer is (imo the best) one of the top $$$ machines and pairing it with a budget grinder is a classic sign the owner doesn't know a thing about coffee. Often the grinder is more influential than the espresso machine.
And how I mention "Group 3" that means it has three brewing heads. They were using a ~$20-30k espresso machine paired with a run of the mill budget grinder.
So there’s a good number of comments, I’ll try to address them:
- Yes, these issues persist with WSL2.
- WSL2 allows mounting between the system/subsystem, but there is considerable overhead.
- Using WSL for remote workspaces from the host is very much a mixed bag.
- Attempting to use WSL entirely with graphical applications has very limited/poor support.
- If you wish for VM acceleration, you have to use Hyper-V, not all toolchains work with Hyper-V and this heavily restricts the host machine.
- If you wish to do anything that crosses the subsystem and the host, line delimiters and platform detection are very error prone.
- If you accidentally misconfigure WSL2 (which is quite easy to do) the WSL userspace can have substantial access to the host files, often beyond what may be initially apparent.
- Of compatibility issues, non-standard socket implementations have caused a lot of incompatibilities with software for me.
This is great and all... Except I’ve long given up on WSL. I really tried to make WSL work for many things, only to find the entire experience (reliability, performance, and beyond) was simply better in every way without Windows.
No matter what you do, there will always be some weird platform detection or line termination that pops up somewhere. And if it isn’t that, it’s degraded performance or kernel-level incompatibility.
It’s hard to elaborate just how wild this model might be if it performs as claimed. The claims are this can perform close to Sonnet 4.5 for assisted coding (SWE bench) while using only 3B active parameters. This is obscenely small for the claimed performance.
This model is very impressive for the size and the architecture enables incredible speed.
Thus far community comments suggest the performance claims are real. However, the region gating for the company’s main website does seem odd.
Historically, the FAA had zero interest/jurisdiction in most hobbyist aircraft below 500 feet, they would not even bother entertaining such as it was not relevant to their opperations. Helicopters, planes, and quadcopters could be flown freely under general guidance, with no actual enforcing regulations. These were largely not seen as something the FAA had jurisdiction over, nor did the FAA express any interest. It was widely accepted that fields away from an airport and at low altitude were outside of controlled airspace.
Then come affordable drones and suddenly the FAA attempts to exert full regulation over the space.
To answer your question, roughly 2012 is when this started.
It's been overreach since the start of FAA claims on low altitude space.
From what I understand their jurisdiction didn't begin until 500 feet into the air.
Not only is it overreach, it's encouraging impediments on what has been largely considered private property.
Also, a less efficient option with less “tech” would be a Silverado WT (work truck), those have a lot of range and sometimes you can find them for good used prices. Those are a hit or miss for deals because the MSRP is on the higher end.
For cheap, the 2018 and later Nissan Leafs (old tech) or an Ariya (new tech, nothing notably exciting.
Maybe also check out Ioniq 5, EV6, Equinox, etc.
FWIW, my wife drives a Mach-E and I drive a Fisker Ocean. The Mach-E is very comfortable but tends to be a bit higher in price than some other options. The Fisker Ocean is.. (from what you’ve said) probably not for you.
This does read/interpret a bit odd, because the Hummer H2 doesn’t strike me as a reliable vehicle and I’ve generally heard of them to be cost sinks (completely disregarding the horrible efficiency).
Why not start off looking at the cheapest EV or PHEV that you can find without high mileage that’ll fit your daily driving habits, then give it a test drive? Consider how much monthly expenses will cost (might save ~90% on fuel) and then consider if you like the driving characteristics more.
I’m sharing a prompt/template/document created to weight context with engineering principles.
The goal is to weight initial context with principles to improve software development when using LLMs.
I didn't have the patience to click through after visiting a few pages only to find the depth lacking.
About an hour ago or so I used Opus 4.5 to give me a flat list with summaries. I tried to post it here as a comment but it was too long and I didn't bother to split it up. They all seem to be things I've heard of in one way another, but nothing that really stood out for me. Don't get me wrong, they're decent concepts and it's clear others appreciate this resource more than I.
I can't stop any megacorp from training on my code, and it's futile to try.
I do not like this take and I do hope you reconsider repeating such. This very much reads as accepting a lack of any claim to reasonable privacy and ownership, borderline on accepting what I would consider theft.
I have no doubt your intentions were good and I am a fan of open-source myself. Unfortunately, since this was written, it’s become more widely known and commonplace for large corporations to disproportionally benefit from open source.
And that wouldn’t bother so many as much if it weren’t for the fact that large corporations often do not give back. It’s become so much of an issue that OSS maintainers have switched licenses, some have shifted closed-source, and others have simply abandoned their projects.
Just last week I began rethinking usage of MIT/Apache licenses for future work. For the longest time I was hesitant about GPLv3 and almost scared to use in my personal projects, but it turns out my hesitations were fueled by...large corporations.
ik_llama is almost always faster when tuned. However, when untuned I've found them to be very similar in performance with varied results as to which will perform better.
But vLLM and Sglang tend to be faster than both of those.
There is plenty that can run within 32/64/96gb VRAM. IMO models like Phi-4 are underrated for many simple tasks. Some quantized Gemma 3 are quite good as well.
There are larger/better models as well, but those tend to really push the limits of 96gb.
FWIW when you start pushing into 128gb+, the ~500gb models really start to become attractive because at that point you’re probably wanting just a bit more out of everything.
In theory, it’s only sufficient for pipeline parallel due to limited lanes and interconnect bandwidth.
Generally, scalability on consumer GPUs falls off between 4-8 GPUs for most. Those running more GPUs are typically using a higher quantity of smaller GPUs for cost effectiveness.
At first glance, it appears to have lower negative pressure and a much higher filter surface area than most air purifiers.
AFAIK PC fans are designed to move air at low pressure while quality air purifiers are designed to handle more pressure.
Other than that, it’s a neat writeup. Would like to see a follow-up with static pressure considerations.
i think people that were promoted for building systems that turned out bad, should be demoted
Nope, in the same vein of "lording" over others, they become the expert of knowledge of bullshit. The environments that allow such behavior have already engrained reward of such behavior.
I'm convinced it'll be in the form of emergence and it won't immediately be known when it's started.
It may very well already be here, but the feedback loops are excruciatingly long and expensive.