Skip to content
Agentguiden.no
Submit your own

Let Claude Code test new local models while you sleep

Claude Code tested 18 new open models in two weeks for Siim Tolk, at night, and none took over from the model he uses. He runs a personal Hermes assistant on a home server.

Done with Claude CodeModerate

What was done

Claude Code tested 18 new open models in two weeks for Siim Tolk, at night, and none took over from the model he uses. He runs a personal Hermes assistant on a home server. He ticks which ones get downloaded, and at breakfast the verdict arrives on WhatsApp. The agent never switches models itself. He writes that some of the limits hold only as far as the agent follows its instructions.

How to do it

  1. Paste the prompt into the tool, and replace the text in brackets with your own.
  2. Read the answer through. Do not send, delete or publish anything until you have seen it yourself.
  3. Check the original before you trust a figure, a name or a date.

Try it with Claude Code

Write a test plan for whether [new model] can replace [the model I use] on [my usual tasks], with the same tests for both, and you are done when I have a table with results and speed for each. Ask before you download, install or change the setup running today.

Replace text in brackets with your own.

Done when. you are done when I have a table with results and speed for each.

Safety rule. Ask before you download, install or change the setup running today.

Read the original

Blogg via Hacker News ·

More examples

Five agents a day

A short daily email with five concrete examples of AI agents. Nothing else.