Human Terms

AI’s biggest names put dates on their promises. Did they deliver?

AI leaders make big predictions with a date attached, and almost nobody goes back to check. Here are seven, set against what actually happened by November 10, 2026: two kept, two late, two broken, and one still running.

Two are below. Get all seven, plus a card for your phone, by subscribing to Human Terms: one free email every Tuesday that checks the week’s AI news for you.

  1. A million robotaxis by 2020 · Broken

    “Next year for sure, we’ll have over a million robotaxis on the road.” Elon Musk, Tesla, April 22, 2019. He added that some places wouldn’t approve it. TechCrunch

    What happened. Tesla says its robotaxi service is live in seven US metro areas, with a safety driver still on board in the Bay Area. In Texas, its biggest market, 811 vehicles were cleared to drive themselves as of October 21, 2026. That’s less than 0.1% of a million, six years after the deadline. Tesla Q2 update · CNBC, Oct 21

  2. AI writing 90% of the code within six months · Late

    Dario Amodei, Anthropic’s chief executive, said on March 10, 2025 that AI would be “writing 90 percent of the code” in “three to six months,” and in twelve months maybe “essentially all of the code.” CFR transcript

    What happened. In October 2025, seven months later, an outside researcher estimated that about half the code at Anthropic was written by AI, with some teams near 90%. In February 2026 Amodei said it “happened at Anthropic,” by his own count. Nobody has a reliable count for the rest of the industry. Redwood Research · Dwarkesh Podcast

  3. Human-level AI by the end of 2026 · Too early to call

    In late August 2026, Sam Altman told TIME that OpenAI was “not quite yet” there, but that by the end of the year it would have an internal system he would call AGI, meaning AI that matches people at most work. TIME

    What happened. On September 3 OpenAI released GPT-6 Astra, and its president Greg Brockman said “it’s not unreasonable to feel that we are now in the AGI era.” ARC Prize, the nonprofit that runs one of the best-known tests, said of the same model: “we are not claiming that it is AGI.” Nobody independent decides what counts as AGI, and the deadline is December 31. Fortune · ARC Prize · Issue #9

  4. GPT-5 “within months” · Kept

    On February 12, 2025, Sam Altman laid out OpenAI’s plan: GPT-4.5 within weeks, then GPT-5 within months. TechCrunch

    What happened. GPT-4.5 launched February 27, 15 days later. GPT-5 launched August 7, 2025, just under six months later, free for every ChatGPT user. “Months” is a loose deadline, but he met it. TechCrunch · Fortune

  5. AI-designed drugs in human trials by the end of 2025 · Broken

    “Some AI-designed drugs in clinical trials by the end of the year… That’s the plan.” Demis Hassabis, head of Google DeepMind and its drug company Isomorphic Labs, at Davos, January 21, 2025. The Register

    What happened. In January 2026 Fortune reported that Isomorphic “has yet to move a drug into the critical clinical trial phase.” By February the goal had moved to human trials by the end of 2026. Fortune, Jan · Fortune, Feb

  6. Meta AI for a billion people in 2025 · Kept

    On January 24, 2025, Mark Zuckerberg said Meta AI should become the “leading assistant serving more than 1 billion people” that year. CNBC

    What happened. On May 28, 2025, Zuckerberg said Meta AI had 1 billion monthly users. That count covers all of Meta’s apps, where the assistant is built in. So: kept, counted Meta’s way. CNBC

  7. Grok 4.7 “in 10 days” · Late

    On September 1, 2026, Elon Musk said xAI’s Grok 4.7 was “10 days” away, and that it should land “roughly on par with” Anthropic’s Claude Opus 5.0. Decrypt

    What happened. It shipped September 21, 20 days later, after at least five delays since late July. Decrypt · Issue #11

Save it. Download the card (an image for your phone), or bookmark this page.

Every Tuesday at 7am, Human Terms checks the week’s AI news for you, in plain English. Read past issues.