On November 21, Elon Musk posted on X that SFT (supervised fine-tuning) is "more like getting a haircut with a chainsaw" than fine-tuning and should be built into primary training, and that having a system prompt is "silly" because default behavior should be sensible. He said both are already true for Tesla AI. At writing time the post had about 14,600 likes, 1,241 replies and about 2.2 million views. This is an opinion, not a technical paper.
Key Takeaways
- ✓Point one: Musk argues SFT is a crude after-the-fact edit, "a haircut with a chainsaw", and should be built into primary training
- ✓Point two: he calls having a system prompt "silly", saying default behavior should be sensible
- ✓He says both are already true for Tesla AI, calls it "the most efficient AI on Earth" and says FSD V15 goes to wide release next month; these are his claims, with no data attached
- ✓Engagement at writing time: about 14,600 likes, 1,188 reposts, 1,241 replies, 391 quotes and about 2.2 million views
- ✓Pushback: one quote post notes that for an auditor the system prompt may be the only readable statement of agent policy, and moving it all into training leaves governance with no document to inspect
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
In-Depth Technical Analysis
Background
Today's mainstream LLM recipe is large-scale pretraining, then SFT (supervised fine-tuning) and preference training to shape behavior, and finally a system prompt at deployment to set role and rules. On November 21, Elon Musk criticized this pipeline directly on X, and the post quickly became one of the day's most-discussed AI topics.
What he said
Musk made two points. First, SFT is "more like getting a haircut with a chainsaw" than fine-tuning and should be built into primary training. Second, having a system prompt is "silly" because default behavior should be sensible. He said both are already true for Tesla AI, called it "the most efficient AI on Earth", and said FSD V15 goes to wide release next month. The post gives no training details, paper or data, so all of this should be read as his opinion.
Numbers that matter
There are no benchmarks here. Engagement at writing time: about 14,600 likes, 1,188 reposts, 1,241 replies, 391 quotes and about 2.2 million views.
What it means for developers
There is pushback in the thread too. One quote post notes that for a company auditor, the system prompt may be the only readable statement of agent policy, and moving that policy entirely into training leaves governance without a document to inspect. For developers, major APIs from OpenAI, Anthropic, Google and others still rely on system or developer instructions as the main way to steer behavior, so there is no reason to change practice now; writing key rules down clearly, versioning them and running evals remains the auditable, reversible choice.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.