3 ms·

Also people don't talk enough about (or are bad at separating themselves) the model vs. the client tool - e.g. from your comment maybe using codex/Claude Code/a

by OJFord 1y ago

Also people don't talk enough about (or are bad at separating themselves) the model vs. the client tool - e.g. from your comment maybe using codex/Claude Code/aider with Gemini API would be better, best even, but people rarely make that comparison or separation, it's always 'Claude Code with Claude vs. codex with GPT-x' etc.

CaptainOfCoit 1y ago
To be fair, most of the times, the tools works best with the models trained with those tools in mind, and vice-versa. Not to mention not all models/inference works the same way so you can't really replicate the same experience. For example, new Harmony format means you can now inject messages while GPT-OSS is running inference, but obviously Claude Code don't support that because their models don't support that.
- kaycey2022 1y ago
  >most of the times, the tools works best with the models trained with those tools in mind This is a garbage state of affairs though
  - CaptainOfCoit 1y ago
    What do you expect? People building software using other models than they themselves develop? Or people training the models train them for software that isn't the software they develop themselves? It's like saying official car repair shops should repair any type of car, not just their brand. That's just not how the real world works.
    - OJFord 1y ago
      Or they just don't even do it? Michelin tyres aren't best fitted by a Michelin shop, those don't exist. You go to KwikFit or whatever you think's best, and get Michelin or Continental or Pirelli or whatever you think's best fitted.
      - antonvs 1y ago
        Terrible analogy, because the issues involved in fitting car tires of different brands are in no way comparable to the differences between LLM behavior across models.
kaycey2022 1y ago
Yeah. The client tool does make a difference. For example opencode, if I am correct, just spins up its own language servers and then feeds the language server errors back into the model, resulting in a much better agentic coding experience. I don't think they are doing anything much more complex than that. Unfortunately, nearly all the foundation model companies are just wasting their efforts on the clients, which are kind of ass, instead of focusing on the model. Google would be much better off if they ditch their dogshit cli, and allow us to have the generous quota login off any client.

by OJFord 1y ago

CaptainOfCoit 1y ago

To be fair, most of the times, the tools works best with the models trained with those tools in mind, and vice-versa. Not to mention not all models/inference works the same way so you can't really replicate the same experience. For example, new Harmony format means you can now inject messages while GPT-OSS is running inference, but obviously Claude Code don't support that because their models don't support that.

kaycey2022 1y ago
>most of the times, the tools works best with the models trained with those tools in mind This is a garbage state of affairs though
- CaptainOfCoit 1y ago
  What do you expect? People building software using other models than they themselves develop? Or people training the models train them for software that isn't the software they develop themselves? It's like saying official car repair shops should repair any type of car, not just their brand. That's just not how the real world works.
  - OJFord 1y ago
    Or they just don't even do it? Michelin tyres aren't best fitted by a Michelin shop, those don't exist. You go to KwikFit or whatever you think's best, and get Michelin or Continental or Pirelli or whatever you think's best fitted.
    - antonvs 1y ago
      Terrible analogy, because the issues involved in fitting car tires of different brands are in no way comparable to the differences between LLM behavior across models.

kaycey2022 1y ago

Yeah. The client tool does make a difference. For example opencode, if I am correct, just spins up its own language servers and then feeds the language server errors back into the model, resulting in a much better agentic coding experience. I don't think they are doing anything much more complex than that. Unfortunately, nearly all the foundation model companies are just wasting their efforts on the clients, which are kind of ass, instead of focusing on the model. Google would be much better off if they ditch their dogshit cli, and allow us to have the generous quota login off any client.