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

One of the most interesting features of sodium batteries is that they still perform good in cold temperatures.

And high temperatures, too. Meaning they don't require cooling nor heating, basically matching the per kg capacity of ready modules with LFP while being significantly safer and less complex.

So does Opencode Go ($10/mo: https://opencode.ai/go) for DeepSeek v4 Flash and MiMo 2.5.

That looks pretty nice. How does it compare cost-wise to just using OpenRouter?

The Go plan essentially gives you $50 of inference for $10 per month ($5 for the first month).

$60/mo currently: https://opencode.ai/docs/go/#usage-limits

Their limits are staggered: 5h (max $12), weekly ($30), monthly ($60).


My mistake. You are correct.

Having proof is not always enough. I had a bad experience with a shady used car salesman and my review was removed by them. I appealed, nothing happened. I deleted the review and put up a new concise one with proof. It wasn't published. The appeal was eventually successful but my new review still not published. I appealed via arbitration, which was decided in my favor but the review is still not published, as their decision is not legally binding. So Google rather pays hundreds of Euros in arbitration fees than publish my honest review. In the arbitration process they said that because I deleted my first review there is nothing they could do, which to me is nonsense. Overall the Google Maps review situation is very frustrating. At least they now show how many reviews have been removed in the past year for defamation for every Google Maps listing.

There are US providers for DeepSeek v4, MiMo 2.5 and GLM 5.1.

But those US providers AREN'T CHEAP like the Chinese ones are (for the big, actually useful ones, like 1.6T+ models)

Does the location help though, if the company isn't trusted? I can't even visit the webpages of these companies from my enterprise network

I'm speaking of third-party providers. They just host those open models themselves on their hardware.

And even if so, I'll try to get rid of any US affiliations within my workplace, so US providers are not an option either.

There are also EU providers for those models, e. g. Tensorix.

> As a bit of an aside, I have been toying with the idea of adding some sort of second pass/security auditing/scaling offering to my consultancy for people vibe coding projects which wind up generating interest. (Not sure what the fuck else I'm going to do!) I have a few non-technical friends who have found themselves in this situation and there's a real need for it.

There might be a need for it but as a consultant your daily rate should be way above what a small vibe coder is willing to pay.

> As signs of how desperate I'm getting, I recently signed up for Task Rabbit and am seriously considering applying for a job at Tractor Supply.

I hope you'll find a way to keep going. Signing up for gig work is a race to the bottom though and not something I'd recommend. May I ask how you've arrived at this point?


> There might be a need for it but as a consultant your daily rate should be way above what a small vibe coder is willing to pay.

Yes, I agree and that's part of the tension. I called attention to projects which are actually generating interest because, no, not everybody needs that sort of treatment. People who have real (i.e. not friends/family) users, are storing PII, accepting payments, etc. probably do -- after some threshold has been crossed, though.

> I hope you'll find a way to keep going. Signing up for gig work is a race to the bottom though and not something I'd recommend. May I ask how you've arrived at this point?

Thank you. I sincerely appreciate that. I know Task Rabbit and co. are a race to the bottom but ... I've been burning through savings and need funds to start coming in. There's also an opportunity cost, as that's time I can't spend networking, applying for jobs, writing blog posts, etc. but, again, I can't sustain the burn and don't want to take on more debt. "Gig work" seems simpler than getting hired and put on a schedule somewhere and needing to quit when (hopefully!) I can land some more "real" work.

As for how I got here: Historically, my spouse has had the "safe" job, along with benefits, etc. and I've been the one to freelance/run the consultancy with the healthier rates. (I'm a bit ... neurospicy and can't handle traditional employment, anyways. This also complicates outreach, lead gen, marketing, etc.) For many years, I had a good mix of new/repeat clients through word-of-mouth and was turning down hours. My schedule was also flexible enough that I could be the one deal with getting the kids to/from where they needed to be. Fast-forward a few years and a lot of the work I was doing has either been halted (e.g. due to arts/science funding in the US), brought in-house (e.g. experiential/creative tech) or commoditized by Claude, ChatGPT, etc. (e.g. building CMSs, web services, IAC, AI/ML-ops supporting training/deployment of bespoke LLM/VLMs). I'm now stuck being the one with the flexible schedule, so 9-5 would be really challenging (on a few levels) and my resume isn't getting "through the front door" anywhere, anyways -- despite my experience being strong and varied. So, I've been burning savings while taking on what work I can drum up but it's slowed to basically nothing over the course of the last year.


+1, it's good enough for what I need to do as a DevOps engineer.

After reviewing the video tapes the police concluded that the women knew that they were handling poison - they kept their hands away from their body and immediately washed them after the attack.

Someone could have told them it was anything else that you wouldn't want on your body. Like, fart spray or whatever. A prank. That behavior doesn't really tell you anything conclusive, but I guess they just let anyone be a cop these days.

I've been using DeepSeek v4 Flash with OpenCode for the whole week to refactor a Terraform code base I inherited and it worked surprisingly well.


In that scenario others host the model, not DeepSeek themselves, so they indeed charge their own prices.


Which one-off tasks need 10k lines of code?


Would depend on what AI and prompt you use ultimately. Ask it to add tests (functional, E2E and unit, maybe invent a new type too), packaging, modular code and/or whatever, and you get to 10K relatively quickly with some of the more verbose LLMs out there.

Personally it's probably the biggest struggle, trying to rein in the "spray and pray" approach LLMs typically like to take, and reducing the "patch on top of patch" syndrome too.


One off web app for scrubbing through some data, that, once done, will never be run again?


Java programs


Enterprise programs*


Calculate the engine power of a 2015 VW polo when travelling 70 mph on a flat road behind a box truck. Draw a chart of drag Vs follow distance. How significant is humidity on the result?


European or African Polo?


You're not supposed to post that you just like a comment, but this was best comment on HN in ages.



I don't know th- AAAAHHHH


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

Search: