Musk Asks Users to 'Tweak' His New AI: Elon Musk Releases an AI That Will Get Smarter Every Week

Elon Musk has announced the start of public beta testing for the updated Grok language model. This is version 4.2, which he says features a fundamentally new training algorithm—the neural network can now retrain itself extremely fast right during active use.

Musk posted the announcement on his official X account, explaining that the release candidate is now available to everyone. However, the new version doesn't automatically kick in. To try it out, users need to go into the settings of the web version or the app and switch the model manually. For now, Grok 4.2 has "public beta" status, so stability and response quality may vary.

The update's main "killer feature," and the one they're really pushing, is its ability to learn quickly. Unlike previous versions that required lengthy retraining on servers, Grok 4.2 will be improved on a weekly basis. Musk stated that the final stable release of 4.2 will be "roughly an order of magnitude smarter" than the base Grok 4 model, both in terms of speed and reasoning quality.

The developers are clearly counting on feedback from the community. Musk urged users to leave criticism so the xAI team can make fixes quickly. Judging by his post, patches and improvements will be rolled out almost daily, and at the end of each week, they promise to publish a changelog.

As a reminder, the original Grok debuted in late 2023 and was positioned as a edgier, less politically correct alternative to ChatGPT. The model has been evolving since then, and in February, xAI released an intermediate update that they are now refining through public testing. Musk himself has previously hinted that future versions of the AI will feature even deeper integration with the X social platform.

What do you think of the idea of an AI updating weekly, "on the fly"? Do you trust these kinds of experimental versions, or would you rather wait for the stable release? Let us know in the comments!