HN user

schaefer

1,602 karma

Hi friends,

I enjoy Rock Climbing, Go, Software Engineering, and Buddhism.

dilapidated homepage: https://slow.camp

Posts2
Comments537
View on HN

I flip flop between qwen 3.6 27b and qwen 3.6 35b 4b active.

But there’s also the quantization of DeepSeek v4 flash called dwarfstar

From TFA:

In this last Bikeshed in acmqueue, I will ponder the far future of free and open source software (FOSS), hoping to upset so many readers that...

During the past couple of decades, rampant neoliberalism and “globalism” allowed...

And I’m out. I guess congratulations to the author. Mission accomplished.

But I’m disappointed that the article took a turn towards partisan politics.

in the first example:

```

10: rjk::duck<Container> c{std::vector<int>{1, 2, 3}};

11: c.size(); // 3

12:

13: c = std::string{"hello"};

```

Does the assignment on line 13 call the destrucor for the vector of ints created on line 10?

So an effective way is to replace the time you spend in front of a screen... I removed Instagram, YouTube...

Really great advice. Last week I configured my wifi router at home to block youtube entirely. I literally feel like a different person in just a week. I have so much more free time and I am so much less anxious.

Avoid even audiobooks.

Controversial. I suppose I used to not have an opinion at all on this topic, until I saw an interview with Salman Rushdie after the failed attempt on his life. He said he since the attack, and loosing his right eye, he reads with enormous fonts on an ipad, or yes, he even listens to audio books now.

If audiobooks are good enough for a seven time nominee of the Booker prize, who am I to quibble?

Claude Fable 5 1 month ago

I don't think anyone has a firm grasp on actual inference costs.

There are huge numbers of users (myself included) that do have an exact idea of what inference costs are - on open models. We can buy tokens from 3rd parties that have no motivation to subsidize our use. That's to say, there's a fair marketplace[1] and we're hanging out there.

If you want to say "I don't think anyone has a firm grasp on actual inference costs on these proprietary/closed models", then I could agree with that.

[1] https://openrouter.ai/rankings#leaderboard

(re)built are not what I'd describe as "terrific".

It sounds like you are a strong candidate to try out the new improvements mentioned in this devlog and see what benefits you can get for yourself.

having moved from Linux ... to freeBSD default install makes my laptop last about twice or thrice as long.

I’ve literally never heard this from anyone before, and I have to admit, I’m curious enough to try it for myself.

The last time I tried FreeBSD was 2001.

Okay, I'm curious.

From the FAQ, I see:

Can I point it at a self-hosted / private DeepSeek endpoint?

Yes. Since 0.30 we accept non-standard key prefixes for self-hosted DeepSeek endpoints. Just point `baseUrl` at your internal address — the loop, cache strategy, and tool protocol are unchanged.

But my question is: If I use Reasonix to talk to a deepseek endpoint through openrouter, am I still getting the cache-hit benifits of this agent harness?

I don't have an LLM-positive culture at work. I'm on a bit of an island. Or under a rock.

Anyhow, I'm pulling myself up by my own bootstraps.

For me a 5% overhead is fine... if it gives me better visibility of this rapidly moving field.

C is horribly and unfixably broken. ... Let's move on.

I love that you are both making this argument and that you have a link to a boutique C compiler written in assembly on your home page.

While I'm commenting on your home page - I recognize that photo as being Red Rock. Possibly pine creek? but can I ask which route specifically?

What's the best way to use it with Pi, OpenRouter?

I can't claim it's "the best"...

But the Pi.dev and OpenRouter combo is what I'm doing at home, and I love it. Setup was easy, I can use /model to switch between any of the openrouter models and whatever I'm hosting locally via VLLM.

For me, the value in local inference is getting your hands dirty and goofing around. That is to say, learning.

So we shouldn’t be comparing it to the cost of open router api access at all, we should be comparing it to the cost of a 4 credit university course.