Can you sport the difference ?

New version cooking

In a nutshell, after some hours using the OpenRouter Quasar Alpha mystery model doing Python programming I can say it's the fastest model I have used so far, showing the accuracy similar to Claude Sonnet 3.7 and Gemini Pro 2.5.

If you want to test it by yourself, check:

github.com/joaompinto/a...

The best free AI experimental case for this weekend:

- github.com/RooVetGit/Ro... (Opensource for vscode)
- openrouter.ai/openrouter/q... (NEW AI model from Stealth provider, OpenSource? Comercial? China? US? EU? ...? )

In some schools in Switzerland the kids are challenged to interact (via supervised chat) with another subject with the goal of identifying whether they are interacting with a human being or an AI. Very positive and impressive adjustment on the education system.

Published Janito 0.13.0, the main new features being --trust and --continue.

Janito keeps rocking the solar system :)

Anthropic shipped a new "web search" feature for their Claude consumer apps today, here are my notes - it's frustrating that they don't share details on whether the underlying index is their own or run by a partner simonwillison.net/2025/Mar/20/...

Janito 0.12.0 is out.

github.com/joaompinto/j...

Dear @windsurfai.bsky.social ,
one of the main features from the Sonnet models it's their fine tuning to the text editor tool , please adapt to it, you do not charge the failed calls but you a) waste our (valuable) time and b) waste your own budget being forced to increase the end price.

Janito 0.12.0 getting in shape

Starting to use Janito for code vibing.

Janito 0.5.0 - very happy with the diff view

With Janito 0.3.0 I have discovered that being presented with different implementation options for a given request, the chance of getting a good result are much higher.

Trying to build a product based on LLMs outputs and achieving a good level of reliability is hard.