There are a lot of Aldi in France (also Lidl, also Action (Netherlands)...) Not sure of your point.

DeepSeek without OpenRouter is wayyyy cheaper

Is it? OpenRouter shows DeepInfra being cheaper than DeepSeek directly

https://openrouter.ai/deepseek/deepseek-v4-flash-20260731#pr...

I agree with parent. OpenRouter might be cheaper list-price, but i have been using 10$ on DS platform since April/May, still have 2$ left. Using OpenRouter i depleted the same dollar-amount in a 1-2 weeks with same usage pattern. No idea why.

Not used it but from perusing forums, consensus is OpenRouter has terrible/broken caching for Deepseek wildly inflating token use. The cache metrics are visible so you can verify for yourself (apparently, it's what others have done).

What does your usage look like?

DeepInfra's Cache Read is 6.5 times more expensive than DeepSeek.

And in many scenarios, cache hits are 99% of tokens, so the price difference in caching really adds up quickly.

Interesting. My anecdotal experience is that it is.

You are American aren't you ?

It does ?

> not because pizzerias in Berlin or Warsaw

Well kinda actually. In countries with good culinary, even basic konbini food can be better than more "serious" restaurant in bad culinary countries.

I think you could make a relatively uncontroversial argument that pizzerias in Germany or Poland on average are pretty bad, but if you made that claim about Berlin or Warsaw _specifically_, I wouldn't take you seriously.

Your conclusion is really a false dichotomy. The square of something very very small (close to zero) is negligeable : thus subortibal hop (Concorde was flying at 18km altitute for example).

I don't know much of PHP, but reaching "Rust/Go" performance is not simply because they have AOT compiler, but because the semantic of the language itself allows efficient compilation.

Many attempts has been made for an AOT compiler for Python and Ruby, but they don't reach those performance because of this.

This is not what "by definition" means.

A - your code

B - a heavy runtime that is greater than 0

C - system libraries

By definition

A + C < A + B + C

Again, this is not by definition. This is by deduction.

What about https://github.com/tw93/mole ?

Mole looks great — I hadn't seen it before. Go is a solid choice for this kind of tool.

  The main differences from onClean:

  - Mole is free/open-source (MIT), onClean is $5.99 one-time. If free matters most, Mole wins there.
  - onClean's scanning engine is written in Zig with direct POSIX syscalls — different tradeoff (smaller
  binary, no GC pauses during large scans).
  - onClean has first-class support for developer-specific caches (Xcode DerivedData, nested node_modules,
   Cargo, CocoaPods, Docker) which tend to be the biggest space hogs for devs.
  - We also have a menu bar helper for scheduled cleaning and a recommendations system for stale
  downloads/screenshots.

  Honestly, the more tools in this space the better — CleanMyMac charging $40/yr for cache deletion has
  gone on long enough. Thanks for sharing the link.