Elon Musk Declares AI a 'Supersonic Tsunami' — And Admits He Was Wrong About Anthropic

Monday's Post: 'AI Is a Supersonic Tsunami'

On August 3, 2026, Elon Musk posted on X that AI progress has become a "supersonic tsunami." He shared a chart plotting monthly AI progress from July 2023 through July 2026. The curve remained relatively flat for years, then snapped sharply upward starting in spring 2026. The chart labels November 2024 as "it's so over" before the steep breakout. Musk predicted that AGI could surpass individual human intelligence by year-end 2026.

Critics were quick to note that the chart had no labeled y-axis, making its exact metric — spending, compute, benchmark scores, or something else — unclear. Regardless, the message was unmistakable: Musk sees AI as an accelerating, unstoppable force.

Grok 4.5: A Benchmark Win, But Not a Crown

Musk's xAI released Grok 4.5 publicly on July 8, 2026. The model's strongest independent achievement came on the Artificial Analysis agent benchmark, where it topped the ranking on cost and speed — an evaluation of agentic performance and efficiency. It also scored #1 of 28 models on the τ³-Banking benchmark for agentic tool use (33%, ahead of GPT-5.5 at 31%).

However, on overall composite intelligence, the picture is more measured. Grok 4.5 scores 54 on the Artificial Analysis Intelligence Index, placing it 4th of 168 models — behind Claude Fable 5 (60), Claude Opus 4.8 (56), and GPT-5.5 (55). Its real selling point is efficiency and price: at $2/$6 per million input/output tokens, it costs roughly a fifth of its premium rivals and uses 3–4x fewer tokens per coding task.

Where Grok 4.5 genuinely excels is in agentic coding benchmarks. It scores 83.3% on Terminal-Bench 2.1 (beating Claude Opus 4.8's 78.9%) and 64.7% on SWE-bench Pro (beating GPT-5.5's 58.6% but trailing Opus 4.8's 69.2%). On DeepSWE 1.0, it scores 62.0%, behind Fable 5 (66.1%) and GPT-5.5 (64.3%). Independent analysts describe Grok 4.5 as a strong specialist rather than a generalist leader — winning on value and speed, not raw intelligence scores.

Musk's Stunning Reversal on Anthropic

In what may be the most surprising turn of 2026, Musk publicly reversed his long-standing dismissal of Anthropic. On July 10, 2026, Musk wrote on X: "I was clearly wrong about Anthropic. They are obviously currently the leader in AI. No company has released a model as good as Mythos/Fable."

This came after years of public criticism. As recently as September 2025, Musk had written that "winning was never in the set of possible outcomes for Anthropic." He had called its models "evil" and "woke" in earlier posts. The reversal extended to a pledge not to weaponize SpaceXAI's compute access against his rival, saying it was "not my style" and pointing to Tesla's history of open-sourcing patents.

The timing is notable: just two days after Grok 4.5's public release, Musk conceded that Anthropic had pulled ahead.

Where Experts Disagree: Hinton vs. Goertzel vs. Musk

Geoffrey Hinton — the "godfather of deep learning" — has dramatically revised his own AGI timeline. He previously estimated 30–50 years, but as of mid-2026 his range is 2028–2043, with a narrower central estimate of 2028–2035 and a "good chance in 10 years or less." He also puts a 10–20% probability on AI-caused human extinction within decades. Hinton has said AI may be "reasoning better than us" within five years.

Ben Goertzel — often called the "father of AGI" — has held a consistently earlier timeline. He predicted human-level AGI within 5–7 years back in 2019 (targeting roughly 2024–2026). More recently, he expects human-level AGI by 2027–2029, with superintelligence following "only a few years" after that. He has said he "won't be surprised" if it arrives as early as 2027.

The key contrast with Musk: Musk now claims AGI could surpass individual human intelligence by end of 2026 — at the very earliest end of the spectrum. Hinton is more measured but has pulled his timeline dramatically forward (2028–2035). Goertzel's timeline (2027–2029) sits in between, closer to Musk than Hinton. All three agree AGI is near, but the spread ranges from months (Musk) to over a decade (Hinton's outer bound).

The Bottom Line

Musk's "supersonic tsunami" chart and Grok 4.5's benchmark results both reflect a field that is genuinely accelerating. Yet the same data that Musk uses to declare imminent AGI also shows his own company's model ranking fourth in composite intelligence — and led him to concede that his fiercest rival is now the leader. The AGI timeline debate remains unresolved, but the range of credible estimates has narrowed from "decades away" to a window of months to a few years.