The article is vague, I'm having a hard time seeing if they're laying down a critique of git's underlying data representation and storage, a critique of git's UX or both.

If the main issue is agent-to-git overhead, until proven otherwise I see it as more of an agentic and/or UI issue than a source-versioning one.

Mistral doesn't look like it's benching at all. They're just as well funded as a lot of Chinese labs doing much more interesting work R&D-wise. Tailoring products for compliance doesn't cut it IMO, but I'm not in their shoes.

> When I was in Germany a few years ago, I was browsing jobs available in France and most software development jobs were under $55K USD. I don't know how you're supposed to survive on that. You can probably make more money as a waiter.

Salaries are calculated differently in France, Germany, US etc. For example working in Germany you're likely to fork a huge amount monthly to your health insurance. It actually makes it difficult to negotiate comp across borders.

$55k USD brut in France is around 2.8k EUR / month after taxes in France, which counts as the upper bracket of Middle Class in most of the country outside of Paris. Paltry when you're used to tech comps but decent for the vast majority of people here.

On the whole I agree that it's not competitive regionally or globally, and as for myself I'd leave for Switzerland or Australia if I had the opportunity.

I will break don't my past salary in france to give people a better idea.

2200€ in my pocket each month with health care and taxes paid.

My yearly gross salary 39_000€. My yearly super-gross salary (what the company actually pay) 54_000€.

Actually the company paid less due to CIR (https://entreprendre.service-public.gouv.fr/vosdroits/F23533) I don't the number but at absolute minimum the company could have paid 31_320€.

I live in the south east of france not in a big city and I have 13 years of xp.

And to compare to the cost of life :

  - 200g of bread is 1€.
  - tomato 4€/kg
  - beef 27-40€/kg
  - 100m2 individual house 350_000€
  - 60m2 flat 150_000€

In Australia, house would be at least 50% more, apartment would be maybe 70% more. Beef is cheaper (about half price) and tastier (from my experience). Bread is probably half price and way tastier in France though.

Actually, after living in Europe and coming back to Australia, the quality and price of meat is what surprised me most. A lot of times in Europe, I would order a burger and the patty would basically taste like sausage meat; tough and flavorless. In Australia, beef is really good and cheap.

Banking regulation varies from country to country, where I'm from a regular personal account isn't usually allowed to directly transfer more than a few tens of thousands of dollars per month. Anything beyond that is handled by the bank itself. Business accounts work differently.

I'm certain receiving a lump six figure sum on my personal account would trigger a review, if only because the taxman wants to know about it.

Backend-wise, it's just different sets of flags. The infrastructure is the same.

Depending on your context professional diagnosis might be expensive and out of reach and more importantly not incentivized. It's a space where self-diagnosis is more often than not the only path to diagnosis and understanding.

I'm myself bipolar (diagnosed) and very likely autistic (undiagnosed) and having a base understanding on the why and how my brain works has been a significant improvement.

Try DS through their own API if that's feasible, AFAIK they're much cheaper than through OS due to cache hit rates.

I'm using hermes and keep close control over open router provider to DSv4 flash 0731

Use only the deepseek provider, you can configure openrouter to do that(but they create some obstacles, go to configurations and allow all providers)

Right now i'm testing deepseek harness, don't wait to test it. The plugin architecture and self awareness of workflows really make you think about "what is a tool vs what is a project".

It has been an amazing experience

They just raised their prices sadly.

Is there actually a difference in cache rates between OR and official API? I have a preset set up on OR so that I only send traffic to deepseek. The preset is important otherwise you will send traffic to different providers but if you weren't doing this already then what can I say, water is wet, of course cache rates will be awful. I get about 70% cache hit rate with the preset which is appropriate for what I'm doing. I haven't used the official API though.

Zenmux say the cache hit rate is 98% for the deepseek flash API. I don't know why, but performance is definitely worse using openrouter.

https://zenmux.ai/deepseek/deepseek-v4-flash

Could be compression or metadata in their pipeline adding noise and making requests less generic.

they will train / retain your prompts though, so to me it's not a valid option

> As soon as things get in range for motivated amateurs, progress skyrockets.

That's how the AI boom came around to start with, before CUDA it was an academic topic.

Going directly though DeepSeek's API and hitting all day long with light/medium tasks I'm at $5/month. That's with pretty vanilla ohmypi.

"Nobody ever got fired for buying AWS."

Nowadays storage and bandwidth is cheap as long as you're not going through a hyperscaler and it doesn't require compute (ie. "dead" data). I'm working mostly on GCP so I'm not sure what are the rates now, but at least AWS were known for billing insane numbers on egress.

As long as you're smart on caching on the edges and deduplication/overlaying with your content (which structured data types like models, containers and repos typically are fit for) you can get remarkably far for less than you'd think.

The economics of serving models seem to be way more dodgy.