Old dystopias were too neat. Even the evil systems had a plan. Reality is messier and more absurd. Maybe aliens are the only thing left that's still safely fictional.
A Chinese example of this for me is Jin Yong, probably the most influential wuxia (Chinese martial-arts fiction) novelist in the Chinese-speaking world. His books are popular fiction in the most literal sense: enormously entertaining, widely read, and for many people first encountered simply as page turners.
As a kid, that was how I read them too.
Only after I started working did I realize how perceptive some of the character writing was. Murong Fu, for example, is genuinely capable, but his ambition is even greater. Everyone around him, friends, followers, even people who love him, is ultimately someone he is willing to sacrifice for his goal. I have met people who seem to reproduce almost the whole arc: the ambition, the triumphs, the disappointments, the pain, the struggle, and sometimes even the madness. Suddenly Murong Fu no longer feels like an exaggerated fictional character.
The book never changes. I had finally acquired the experience to see what was already there. That feels like thickness to me.
I mostly use local models when the data has personal information. Earlier this year, I felt the coding quality was still not as good as Claude Code.
One thing that works for me is to ask the local model to make some fake data with the same format, let Claude Code work on the fake data, and then bring the code back and run it locally on the real data.
This way the real data never leaves my machine, but I can still use a stronger model for most of the coding.
Qwen3.8-27B has been the turning point for me. It's not as strong as the absolute frontier, but it's the first time I feel local coding models are actually functionally useable as daily drivers.
Man I am having a hell of a time trying to optimize 3.8 over 3.6. I don’t have a particularly powerful setup but I can usually push 20-30tok/s on 3.6 and I can barely get to 10 on 3.8. Both unsloth same VRAM/RAM distribution more or less. My 3.6 is still producing consistently better results and faster
Maybe you're holding it wrong because it's the same architecture between the models, assuming you're using the dense 27B model in both cases. And 3.8 is a significant improvement on 3.6.
That's interesting since both models are dense. I wonder if this is more of an optimization issue with 3.8 rather than something inherent to the architecture.
I have definitely use that to great success, and I do think it colors some of my memory here. I need to check if the 3.6 27B I was using previously was also a dense model. Good suggestion appreciate it
I did try some finance analysis earlier this year. I was using a DGX Spark, so I could run some relatively large models, but the results were pretty mixed at the time. I honestly can't remember which models I used anymore.
I'm having a really hard time doing on twin DGX spark what I could do on my quad 3090 rig (which is a scaled down version of what I was using before, the power requirements and the noise were really an issue but I loved the speed and the amount of VRAM). The results tend to be inconsistent, there is lots of looping, far more tokens generated for the same job and lower quality output. I suspect there is some kind of regression in the B12X kernels or something to that effect because none of that should happen, the exact same model on both machines gives wildly different results. Probably this will sort itself out over time. If I may ask, what model / software combo were you using?
I was only using a single DGX Spark, and this was earlier in the year, so I was running some pretty aggressively quantized models — probably in the 1–3 bit range.
My main issue at the time was that my financial data had lots of messy notes, comments, and irregular annotations. The quantized models often failed to process all of that context consistently and would miss things. So I ended up generating a fake dataset with the same structure, asking Claude Code to work out the analysis on that, and then bringing the result back to the local model for the final pass.
I was mainly using llama.cpp at the time, before B12X support was integrated into vLLM, so I think I wasn't using it then.
Thank you, interesting info! I think the Sparks are an interesting platform, the power consumption / memory bandwidth / memory amount trade-off is completely different from the regular cards and it will take a while for the software to really take advantage of them.
I saw this announcement when it came out and completely forgot about it. That explains why Claude Code felt so surprisingly generous these past few months.
I once spent two days on a pr and got an obviously AI generated review that contradicted what we agreed one before. So I had AI respond to it.
The next day he asked if I'd used AI. I used the same justification he'd used for his review.
Fight magic with magic. He never reviewed my pr that way again.
The NVIDIA+AMD support is the part I find really interesting. I know OpenMP and SYCL can already target multiple GPU vendors, but doing this while keeping Rust's safety model seems pretty compelling. I'm curious how portable the performance is in practice.
I learned this. I had what I thought was a pretty good lock, but the chain was cut very clean. The cut was almost perfect flat, so I assume they used a power tool.
I think solitude is great. The hard part is finding the right people to talk to. I'm not sure if this is a Silicon Valley thing, but I've met some people who use a lot of jargon in a way that feels partly about status. At first I thought they just had a lot of ideas. Then they started talking about things in my own field, and I realized they were getting some pretty basic concepts wrong.
My guess is that some of this comes from listening to too many podcasts. You can pick up a lot of jargon and big ideas pretty quickly, and some podcasts are better at giving you that "oh, now I get it" feeling than actually helping you understand the ideas.
I don't live in Silicon Valley but dealing with same issue, finding people brainstorm or discuss ideas is very hard these days. A lot of trend following, listening to same podcast and reading same books, emphasis being on non fiction, using terms like neuroplasticity, neurodivergence etc. Then I end up talking with with Claude, which can be useful sometimes but cant replace human creativity.
> but I've met some people who use a lot of jargon in a way that feels partly about status.
You mean basically every HN commenter nowadays?
Big words. Big terms. Heavy phrasing. All to say a bunch of nothing. True knowledge lies in simplicity. True wisdom lies in humility. Both of which are sorely missing around here.
I’ve been running into this too. It’s especially frustrating when you ask Claude to explain one of its own terms or summaries, and instead of just defining it plainly, it sometimes goes through several rounds of tool calls before giving you a usable explanation. I really don't think such time/tokens should be wasted.
100%. It is crazy that the default response to everything is act then explain. It starts writing code or running commands and I’m just like my dude wtf are you trying to do, can you just clue me in first.
