Grok the Laggard: Just Doge It, Elon
September 2, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I have a vague recollection that Elon Musk allegedly said that DOGE is “what democracy is all about.” Really, Mr. Musk? Why not apply DOGE-think to your Grok project. I think you purchased a company to provide some innovation. The story “Not ‘Used to Losing’: Apparent Elon Musk Pep Talk to His Cursor Staff Paints Sad Picture” includes some factoids I found interesting; for example:
- The DOGE himself wants Grok to be “dominant in the AI coding world.”
- He wants to “catch up” to Anthropic and OpenAI.
- The full self driving visionary “was not used to losing.”
- SpaceX and Tesla are dominant companies. Note: Tesla is not in the number one spot. Tesla sells fewer cars than Subaru. SpaceX has one big customer: Elon Musk’s Starlink. Another note: Repeated launches are helpful because the Starlink satellites do fall from orbit, thus replacements are necessary.
But the killer comment in the article, in my opinion, is this statement:
Musk also supposedly said AI models are sure to become too advanced for humanity to control…
Does this mean that if he creates the Number One AI, only he a non-human or the most superior humanoid on earth other than Pavel Durov, can be trusted with the technology?
Thanks, Midjourney. Good enough.
I am a dinobaby with a tough hide. I am not sufficiently durable to do much hard thinking about life under the control of the DOGE master himself. People who are on the pick up and hold list of the French judiciary make me uncomfortable. I prefer the type of individuals who are in my continuing education lectures at secure event sponsored by the US government. A scoff law makes me want to move back to Washington, DC, and get a job working in a secure facility in Georgetown. (Most people here in rural Kentucky know that those types of offices were available in that part of town without a Metro stop no less.)
The write up included a passage that did not resonate with me. Take a look:
Earlier this year, Grok, Elon Musk’s flagship AI model, was in bad shape—even if you set aside the “MechaHitler” and non-consensual deepfake stuff. In March, Musk admitted as much, writing that Grok’s parent company, xAI “was not built right first time around,” and was “being rebuilt from the foundations up.” Perhaps the clearest internal evidence that something was deeply wrong with Grok was that, according to Bloomberg, not even xAI staffers who liked to vibe code used it. They frequently used Claude, Bloomberg wrote. Musk apparently praised Claude’s parent company Anthropic at his August Cursor pep talk.
Our testing of AI models follows a simple pattern: If it fails, try a model until one works. The passage suggests that “non-consensual deepfake stuff” works. Something called “MechaHitler” works. However more serious work, like programming so X.com mostly functions as a user expects and outputting code for Teslas is not Grok’s strength. Note: I think this explains why M. Brousse at the French judiciary wants to speak with the DOGE master himself. The DOGE master elected to ignore M. Brousse’s request. A jaunt to Paris for a weekend before the masses return from vacation might not be a good idea for the DOGE master to entertain. When one’s own employees do not use the firm’s AI systems, isn’t that a signal that the software is less useful that a spyware infested from an obscure Chinese lab in Beijing?
I liked the wrap up to the Engadget article: “Gizmodo reached out to SpaceX for comment. We will update if we hear back.” Will the DOGE master respond, or will he ignore requests for comments. He dismissed the quite bright and engaging M. Brousse, however. That’s not polite but what can one expect from a person who appears in public on a speaking platform wielding a chain saw? Those tech bros at BAIT (big AI tech) outfits are classy.
Stephen E Arnold, September 2, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
