For me the weirdest is Luna. It costs the same as GLM to solve a task with it, but it just calls the same failing tool over and over again until we cut it out.
Now if you look at where GLM stands, or even DeepSeek v4 Flash, things get really interesting for what they provide.
US labs completely miss a cheap model that can solve problems for 95% of the people. Gemini 3.5 Flash could've been it if it didn't burn so many tokens.
I've been testing Sol/Terra/Luna now since yesterday, running complex evals on all of them and I feel a bit... mixed on how they perform.
The eval is an agent that runs a set of tools and a prompt we can tune separately for different models. The OpenAI version of the prompt was specifically tuned based on their guide[0]. Then we let Opus to run another agent that acts as a user, trying to solve a problem (anonymized and taken from production). The problem is complex and we don't expect it to be solved by these agents, but we measure how the agents operate when faced with a vague problem:
- Opus 4.8 and GLM 5.2 both identified a constraint sooner and stopped so the user can fix an issue first that the agent cannot solve.
- Sol tried hard to solve the issue with different tools, burning tokens, until finally reached to the same conclusion with Opus and GLM. It was two times more expensive compared to Opus and six times more expensive to GLM for this task.
- Terra went even further and started calling tools that would not solve the issue, burning tokens and failing.
- Luna repeated the same failing tool call until it hit the round limit, and burned more money than Opus.
I'm kind of puzzled with the new GPT. Like, yes Sol is OK for programming, but I was expecting to get a cheap agentic model for non-programming tasks, one that can detect if things go awry and correct. Terra is too expensive and Luna not really fit for the task. Sonnet 5 is a bit better but more expensive than Opus 4.8, which is still the best in my evals. GLM 5.2 is extremely good if you can define the task and the tools clearly for it, and costs pennies!
I use it all the time through Fireworks. The normal version when I pay it myself and the fast one when company pays. It's really fast and I never get rate limited with my daily use.
For me it's only about getting into the local scene more now, to learn the language. I kind of started to finally understand the German inefficiency, the distributed nature of the country and their culture. Now when we bought our apartment, the need to be part of the culture is bigger than ever. And at the same time I'm so busy for the past 10 years already, there's more work than ever in my life.
Just need to say stop one day, take a year off and go to language school.
P.S. I don't think it's the same company: the other devs are either in Paris or in Scandinavia.
Hey. I'm with you there. My German also kind of sucks, but I've had a very successful 15 years in Berlin. The best part is how easy it is to pick jobs from the neighboring countries, like France. You pay taxes here, you commute maybe once a month to Paris and enjoy the prices and quietness of Berlin. We are lucky with my partner, and bought finally our own apartment.
My partner, an American, is fluent with the language so it helps. My plan is to make a good amount of savings, take a year or so of sabbatical and finally learn the language. Until that, we go with bar Deutsch.
Depends. I'm more into finding certain masters. And some of the albums are DSD tape transfers. DSD if that was the original recording format, if it was mixed and PCM was needed, DXD flac.
It is not only that. It's the spacing, how the bass sounds, separation of instruments. There's so many interesting headphones in the midrange to try out. Compare the Hifiman HE1000se to Heddphone 2 GT, or to Focal Clear MG and you'll understand.
Also with HEDD you get a handcrafted device made in Berlin. And if you go with nicer cables, they are very beautifully done and feel great. There is no difference in sound of course. Some people like jewelry, I can get similar enjoyment from beautiful audio equipment and cables.
Wait, what? I do download everything I listen. And Roon is quite popular in the music communities. How else you can make sure you have that correct mastering of your favorite album?
I might be something from the middle. Yes, I did spend a hefty 5000 euros to my headphone setup. And yes it sounds absolutely magical and every day I'm happy listening to music with it.
But I also have a large multi-terabyte music collection, I follow new music, go to concerts, go to parties, talk about music with my friends in signal group chats.
It's a hobby, and when you get a bit older and start having some savings, if you love music treating yourself with a better system is not that crazy.
We've all been through that no? In the beginning you can do a ton of stuff without reading code. But the LLMs miss all the good abstractions, they just push and push unmaintainable code until at some point you start having more bugs and then you NEED that LLM to fix the codebase you don't understand anymore.
There are guardrails you can and must add to protect your team if you take the vibe approach: a good type system, a good database with clearly written business model and a good data model to drive your business. Make it loud and clear when something breaks with your tooling.
But... I'd definitely not vibe everything after a certain point. Reading and fixing code is also a lot of fun.
Btw. This whole debacle made me to stop installing any Android updates. I've done my best to avoid installing even the security updates, so my diabetes apps continue working in the future.
I really need to take the time and go with Graphene OS in this device. My bank N26 kind of still allows it, but they made it harder and harder to use with certain custom checks. Looks like in the future I need a separate banking phone and my daily driver.
The device works right now how I want it. I don't want anything to change.
I already contacted the DMA authorities and complained how this has an effect on German diabetes communities and they replied that I am not the first one who approaches them on this and they are already investigating it.
Yes. It is very expensive now. I'm still so so happy I decided last summer to bite the bullet and pre-ordered the Framework Desktop 128GB model.
I paid 2424 euros in total for this machine. And it can easily run the models discussed in the comments and in the article. It's tiny, and runs CachyOS like a champ. Over 4000 euros less than the price you listed.
Yep. I've tried to use the models to build large things for me. You can't trust the code it produces. Even if it works there are parts that are hot garbage, and will bite you later on. I've found out that having an editor open, asking it to implement things until a certain point, manually fixing some of the worst things it generates, then asking it to expand from there is much better than just prompting a thing and pushing to production.
And hey, don't get me wrong, you can get pretty far with just prompting. But the subtle misses and (I'm looking at you GPT) the overengineered 20k line PRs to do a simple thing are going to cost you a lot if you're not vigilant.