Hacker Newsnew | past | comments | ask | show | jobs | submit | impure's commentslogin

I recently made an AI Agent and surprisingly coding with DeepSeek V4 Flash is quite cheap. It probably has to do with the aggressive prompt caching. I'm using OpenRouter with Novita AI as the preferred provider.

Deepseek v4 via deepseek themselves is significantly cheaper.

Because (1) Huawei collab and (2) vLLM etc dont implement half of the inference optimisations deepseek proposed in their paper.


Same here, deepseek v4 flash on opencode go. It's cheap, fats and good enough to follow my instructions

I’m using zen because I have a Claude subscription and just like dabbling with the other models and I was shocked at how little flash cost but it was noticeably not at the level I’d like my model to be.

For me MiniMax 3 has really hit the sweet spot of being very cheap, though more than flash, but I’d also very capable.


Not to be that guy, but the correct term is Open Weight LLM. And I’d argue it already has. Many open models are already very competitive with closed models at a fraction of the cost.

Labs can and do open source more than the weights

https://allenai.org/olmo


No, the new method of inflation that Kevin Warsh likes is the trimmed mean which is 2.8%.

First of all the May jobs report was mostly in temporary workers possibly due to the World Cup. Second as already noted the jobs are mostly in healthcare. Third job openings does not equal employment. These numbers have been diverging for a while likely due to people holding multiple jobs. Also I believe the evidence suggests the job crisis is due to WFH and not AI https://news.ycombinator.com/item?id=48326721

They’d have to keep this up for 12 months which is not guaranteed as you have to factor in hardware depreciation and other costs. Also they’d have to issue more shares to get over the liquidity requirement. And looting may be too strong of a word. They are nudging your returns down a little bit. Although SpaceX could go up too, crazier things have happened.

Yeah, while the world is mostly clickbait at this point, "magically makes them eligible for inclusion in the S&P500" is probably some of the best (i.e. worst).

Good thing they're not dropping the profitability requirements. Ed Zitron would be proud.

They have to be rebalanced every quarter regardless. And I'm not sure how many people would actually sell due to the inclusion of a single company. They're very loud about it, but no evidence that this is causing a significant amount of selling.

Because it hasn't happened yet, and now, won't.

So by that metric the very loud people succeeded: these new IPOs will enter the index under the established rules and time-frames.


Steve Jobs famously said, "Everybody in this country should learn how to program a computer, because it teaches you how to think." Not sure if it is true, but programming definitely has applicability to other things. Programming is just a list of instructions if you think about it.

AI coding tools are great but they are restricted in what they can do. They still result in incredibly hacky solutions. They still litter your code with security vulnerabilities. And they still fall into loops of trying to do things the hard way, possibly failing, and burning your API tokens. I've been hearing about these problems since people started coding with AI and nothing has improved. You will always need a programmer, especially if you plan on maintaining and updating your code.

Plus AI is a wonderful learning tool. You don't have to scour the internet for a solution to your problem only to find someone who asked the same question, resolved it, and never explained how. You can just ask your AI tool.

I am very skeptical about the AI replacing programmers hype. Sure, there are a lot of AI-coded apps but there were always a lot of apps. Discovery on the app store has been a problem for a while now and I don't really see AI changing things much. It might even result in more programmers as programmers need to come in and fix all this AI slop code.


Oh no, not Rsync. I guess that's one good thing about MacOS shipping with an ancient version of rsync. Oh, wait, they ship openrsync now, but the command is still called rsync.


I had an emoji cut in half problem in Dart. I was a bit surprised because I thought substring operations worked on characters. It only caused an invalid Unicode symbol though so not too bad.


> I thought substring operations worked on characters

"character" turns out to be too vague an idea to correspond to some specific fact about the software. If you co-worker says his Uncle is "conservative" does he mean like "Believes Right To Work laws are a good idea" conservative or "Believes Joe Biden is a Communist" conservative ?

https://en.wikipedia.org/wiki/Character_(symbol) gives you some idea about this rabbit hole. Suffice to say, no, you can't have operations on "characters" until you've nailed down exactly what it was you meant by that.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: