Elon Musk: Optimus 3 Is Coming, Recursive Self-Improvement Is Already Here, and the Singularity | #239

Summary

• Musk says AI is already in “the hard takeoff,” with humans progressively leaving the recursive-improvement loop. Every successive model is built to a large degree by its predecessor; full automation “may be there by the end of this year, but not later than next year,” though it is not yet human-free.
• Grok 4.20 is “really, really good” and, by some metrics, the best at predicting things—“arguably the best metric for intelligence”—but xAI is “currently behind on coding.” Musk says they should probably catch and exceed competitors by the middle of this year, while cautioning that AI progress follows overlapping S-curves rather than one uninterrupted exponential.
• Absent World War III and assuming current trends continue, Musk calls a greater-than-10X expansion of the economy in roughly 10 years a “fairly comfortable prediction.” Within that decade he also expects a lunar base, people on Mars and a mass driver on the Moon.
• Musk expects Optimus 3 production to start this summer, initially very slowly, before reaching high volume around next summer. Optimus 4’s design should be complete next year; Tesla is planning a roughly 10 million-square-foot Optimus factory and annual robot-design upgrades.
• Automation does not translate into planned Tesla layoffs: Musk expects headcount to rise while output per employee becomes “incredibly high.” Tesla has roughly 150,000 employees—about two-thirds in factories in one form or another—and perhaps 1–2 million people across suppliers, making labor productivity the nearer-term operating lever.
• Musk assigns an approximately 80%-or-higher probability to a great outcome in which AI and robots make goods and services grow far faster than the money supply, producing deflation and universal high income. He referenced an interim period of possible civil unrest over 2–5 years and COVID-like checks before that transition. Eventually money may cease to matter: machine economies would value “power and mass, wattage and tonnage,” while smart humanoids could give everyone better medical care than the richest person on Earth.

Deep dive

1. Recursive self-improvement is approaching a human-free loop

• Musk’s threshold for full recursive self-improvement is removing humans entirely. Every successive model is already built to a large degree by the one before it, humans are “gradually getting less and less in the loop,” and full automation “may be there by the end of this year, but not later than next year.”
• Asked whether that transition produces a hard takeoff, Musk answered categorically: “We’re in the hard takeoff. Right now.” He now goes to sleep and wakes to successive major AI breakthroughs.
• Grok 4.20 is “really, really good”; Musk says Grok is the best at prediction by some metrics, which is “arguably the best metric for intelligence.” His candid gap: xAI is “currently behind on coding,” with a target to catch and exceed competitors around midyear; he said they should probably get there by the middle of this year.
• His forecasting caveat is structural: AI advances through overlapping S-curves—breakthrough, exponential growth, linear progress, then logarithmic returns. Asked whether modern institutions can keep up, he replied that the singularity is hard to predict because it is hard to know what happens inside it.

2. Intelligence ultimately becomes an energy-and-space industry

• Diamandis sought timing and bandwidth for SpaceX/xAI data centers, but Musk invoked SpaceX’s quiet period and declined to disclose anything that could create problems.
• Musk’s scale model starts with energy: an economy consuming one million times Earth’s current electricity would still harness only roughly one millionth of the Sun’s output. Such intelligence, he argued, would be almost impossible for humans to appreciate and would “solve everything you can possibly think of.”
• Earth intercepts only about half a billionth of the Sun’s energy, which Musk presented as a limit on Earth-only intelligence. He therefore expects solar-system intelligence to exceed Earthbound intelligence by many orders of magnitude.

3. A 10X economy comes with lunar and Martian infrastructure

• Musk predicts the economy will exceed 10 times its current size in roughly 10 years, provided current trends continue and World War III does not “put a kink” in the outlook. He describes that conditional forecast as “fairly comfortable.”
• In the same decade, he expects a Moon base, people on Mars and a lunar mass driver—bringing Gerard K. O’Neill’s vision, in Diamandis’s framing, closer to reality.

4. Optimus scales slowly before productivity turns incredibly high

• Optimus 3 is in its final completion stages and, Musk claims, will be “by far the most advanced robot in the world.” He expects production this summer, a slow initial manufacturing S-curve and high volume around next summer; Optimus 4’s design should be completed next year.
• Tesla’s planned Optimus facility is roughly 10 million square feet and will use a new factory design, though Musk added that it is still not different from other factories. He aims to release and improve the robot design annually.
• Diamandis’s robots-building-robots premise met an important qualification: Tesla still relies heavily on people. Musk cited roughly 100,000 people building things or managing those who build them, roughly 150,000 total employees, about two-thirds connected to factories, and perhaps 1–2 million supplier workers. Tesla plans to increase headcount, not cut it, while making output per human “incredibly high.”

5. Abundance drives deflation, but the outcome is not guaranteed

• Musk puts the probability of a great future at roughly 80%, “maybe more,” while warning against complacency because the range of possible outcomes includes bad ones. His revised posture is optimism paired with trying to make events go well.
• Discussing the path to universal high income, Diamandis recalled their earlier 2–5-year civil-unrest timeframe and the possibility of COVID-like checks in the interim, before demonetization and deflation. Musk’s mechanism is that AI and robots make goods and services grow far faster than the money supply, effectively requiring money to be issued to people and producing deflation.
• At 1,000 times today’s economy, Musk thinks people have probably already saturated anything they can express that they want. Money eventually “will stop being relevant.” He imagines an Ian Banks Culture-like future in which AI ignores human currency and instead values physical inputs: “power and mass, wattage and tonnage.”
• Musk makes the abundance claim tangible through medicine: dexterous, intelligent humanoids could give everyone better care than the richest person on Earth. His example was personal—three neck operations because the first two were wrong—and his requested proof point is simple: “Can AI please solve back pain?”