Checking the Receipts: April

There’s a genre of tech journalism that Karl Bode calls “CEO said a thing journalism”: a CEO makes a prediction, the press reports it as news, and nobody comes back twelve months later to check whether it was true. This is me coming back. I dug up what tech leaders said about AI in April over the last four years and held it against what actually happened. If this is fun, maybe it’ll become a monthly series. 🍷 aged like wine, 🥛 aged like milk, 🤷 jury’s still out.

2022

DALL-E 2 launched in April; Stable Diffusion followed in August; ChatGPT in November. But in April 2022, the AI explosion was still seven months away.

Sam Altman: “We can imagine an AI office worker” (April 2022) 🍷

“This is another example of what I think is going to be a new computer interface trend: you say what you want in natural language or with contextual clues, and the computer does it. […] We can imagine an ‘AI office worker’ that takes requests in natural language like a human does.”

— Blog post on the day DALL-E 2 was announced, April 6, 2022

Four years later: This is essentially a description of what ChatGPT, Copilot, and Claude Code became. Seven months before ChatGPT, the shape of the future was already visible if you knew where to look. (Or were CEO of a frontier lab with access to internal models.)

2023

The year everyone lost their minds. In one week in April, Sundar Pichai sold AI on 60 Minutes, Elon Musk warned about doom while incorporating his own AI company, and Geoffrey Hinton quit Google because he was scared. The entire AI discourse in one week.

Sundar Pichai on 60 Minutes (April 16) 🥛/🍷

“This is going to impact every product across every company. And so that’s, that’s why I think it’s a very, very profound technology.”

(“Every product, every company” — when? A prediction that can’t be falsified can’t embarrass you. Watch for this pattern.)

In the same interview, asked about the black box problem:

“There is an aspect of this which we call — all of us in the field call it as a ‘black box.’ You don’t fully understand. And you can’t quite tell why it said this, or why it got wrong.”

— CBS 60 Minutes transcript, April 16, 2023

Three years later: Both milk and wine. AI has impacted many products, that’s real. But “every company”? MIT Technology Review’s “Great AI Hype Correction” found that 95% of businesses that tried AI reported zero measurable value. His black box admission, on the other hand, aged like fine wine; interpretability is still an unsolved problem, though progress has been made.

Elon Musk on Tucker Carlson (April 17) 🥛

“AI is more dangerous than, say, mismanaged aircraft design or production maintenance or bad car production, in the sense that it has the potential — however small one may regard that probability, but it is non-trivial — it has the potential of civilization destruction.”

Also announced: plans to build “TruthGPT,” a “maximum truth-seeking AI that tries to understand the nature of the universe.” This was weeks after signing an open letter calling for a six-month AI moratorium. He incorporated xAI the same month.

— CNN, April 17, 2023

Three years later: The existential risk claim is arguably still live, and honestly not something I’d bet against. What we can actually pin him on is TruthGPT, which never materialized (the 🥛 is for that specifically; the civilization destruction concern is a 🤷 at best). And the contradiction: doom warning on national TV while incorporating xAI the same month.

Geoffrey Hinton quits Google (late April) 🍷

“I thought it was 30 to 50 years away. Obviously, I no longer think that.”

Revised estimate: “5 to 20 years, but without much confidence.”

— MIT Technology Review, May 2, 2023

Three years later: No AGI. The wide range and explicit uncertainty look better than every confident timeline from Musk, Altman, or Amodei. (Arguably the vaguest prediction on this scorecard. But maybe honest uncertainty IS the right call. Nobel Prize in Physics 2024.)

Emad Mostaque at Goldman Sachs 🥛

“There will be no programmers in five years.”

— Decrypt, reporting on Goldman Sachs Disruptive Technology Symposium, April 2023

Three years later: We’re three years in; millions of programmers exist, and their tools are getting better, not their jobs disappearing. Jury’s technically still out for two more years, but the trend is clear. Mostaque was forced out as Stability AI CEO in March 2024 after an investor mutiny, a staff exodus, allegations of misrepresenting his background, and a co-founder suing him for fraud. He didn’t make it to the halfway point of his own prediction.

2024

The year of escalation. The hedging from 2023 was gone; now it was confident timelines and bigger numbers.

Elon Musk to Norway’s sovereign wealth fund (April 8) 🥛

“My guess is that we’ll have AI that is smarter than any one human probably around the end of next year. By 2029, AI is probably smarter than all humans combined.”

— Fortune, April 9, 2024

Two years later: “End of next year” was 2025. It didn’t happen. The goalpost has been moving annually: 2020 he said 2025, 2024 he said 2025 again, when 2025 passed he pushed to 2026, and by March 2026 he was claiming Tesla would “make AGI.” The prediction moves forward like clockwork; the confidence never wavers.

Dario Amodei on The Ezra Klein Show (April 12) 🥛

“No, no, I told you, I’m a believer in exponentials.”

On training costs: models in training were approaching $1 billion, and by 2025-2026 costs would trend toward $5 or $10 billion.

— The Ezra Klein Show, April 12, 2024

Two years later: The cost prediction was roughly right. The “exponentials forever” framing less so; by February 2026, Amodei himself acknowledged being “near the end of the exponential.” From believer to eulogist in under two years.

2025

The hangover.

Sam Altman on X (April 4) 🥛

“We are going to be able to make GPT-5 much better than we originally thought. We also found it harder than we thought it was going to be to smoothly integrate everything.”

— TechCrunch, April 4, 2025

One year later: GPT-5 launched in August 2025. IEEE Spectrum called it a “rocky launch.” Altman himself admitted OpenAI “totally screwed up.”

Google DeepMind 145-page paper (April 2-4) 🤷

“We are highly uncertain about the timelines until powerful AI systems are developed, but crucially, we find it plausible that they will be developed by 2030.”

The same paper warned AGI-level systems could “permanently destroy humanity.”

— Fortune, April 4, 2025

One year later: The hedged language is more responsible than most CEO predictions. The weird thing is that a corporation casually publishing a paper about how their product might destroy humanity doesn’t even register as remarkable anymore. It’s become a rhetorical device: “this is so powerful it might end the world, which is why you should invest.” Check back in four years.

Pew Research (April 3) 🍷

83% of Americans are not excited about AI in daily life. AI experts are 3-5x more optimistic than the public.

— Pew Research Center, April 3, 2025

One year later: The gap between insider optimism and public indifference is the elephant in every AI boardroom. This may be the most important data point on this entire scorecard.

The pattern

The people selling AI overpromised; the people who hedged were closer to right. Confidence is easy to sell; calibrated uncertainty is harder to monetize. When a CEO makes a prediction, ask who benefits.

See you in May.

All quotes verified against the linked sources. Full research archive available on request.