ainotis Join
My notis

Corrections

Every mistake we have corrected, with the date and what the story said before.

6
corrections
338
stories published
3921
facts checked
13 Nov
Jensen Huang resigns as Nvidia CEO; CFO Colette Kress is interim chief executive

We added what Huang and Nvidia's board said on 8 November, and the reason The New York Times reported for his exit: he was replaced after reports of an affair with a woman U.S. officials suspect of spying for China and another with an executive who reports to him. We also added that the Attorney General says the Justice Department has not determined whether the woman is a spy, and that China denies the claims.

Why we corrected it

The story rested only on Nvidia's filing and press release, which give no reason. The reason was already widely reported when we published, and without it the story was incomplete.

What it said before

Colette Kress, 59, stays chief financial officer while she serves as interim CEO, and the board will consider internal and external candidates for the permanent role. The filing and the press release give no reason for the resignation; Nvidia says it will report third-quarter fiscal 2027 results on Tuesday 17 November as previously scheduled. Neither the filing nor the press release gives a reason for the resignation. Jones is quoted saying the board expects operations, customer commitments and the product roadmap to continue without interruption. Nvidia still plans to report third-quarter fiscal 2027 results on Tuesday 17 November as previously scheduled.

9 Nov
October: critical PHP flaws need patching; Gemini 4 Argon reached defenders before developers

The cover line of the October issue now reads: A cheaper Haiku; Google’s ‘strongest’ to defenders first.

Why we corrected it

The old line stated Google’s description of its own model as our own. Google says Gemini 4 Carbon is “our strongest model”.

What it said before

Cover line of the October issue: A cheaper Haiku; Google’s strongest to defenders first

31 Oct
Anthropic files S-1 for Nasdaq IPO, reporting $16.3bn half-year revenue

The number of underwriters in the body changed from 18 to 22. The claim it rests on was corrected to 22 too.

Why we corrected it

The S-1 lists 22 underwriters; we said 18.

What it said before

The prospectus lists 18 underwriters; Goldman Sachs & Co. LLC and Morgan Stanley & Co. LLC are their representatives.

13 Oct
Reflection AI's Laskin says open models will take most token demand

A key point now gives the period as Laskin put it, maybe six months, and the time stamps where he says it, 25:19 to 25:29.

Why we corrected it

The key point said roughly six months, which is firmer than his words: he said it was majority closed maybe six months ago. The time stamp also pointed to the middle of his sentence rather than its start.

What it said before

He says gateway token use flipped from about 70/30 closed to 70/30 open over roughly six months (25:27), and expects most token demand to go to open models (26:03).

13 Oct
SemiAnalysis's Patel: a trillion in AI capex needs $250 billion of revenue a year

The sentence now says that the $2 trillion Patel gave for next year is AI-related spending, with its supply chain included: data centres and chips, and the rest of the supply chain behind them.

Why we corrected it

The story said capex across the US, which reads as all capital spending in the country. In the interview he was answering a question about the AI buildout, and counted its supply chain. Our summary and key points already said this; the body did not.

What it said before

Patel said capex across the US next year will be on the order of $2 trillion, which he put at five or 6% of GDP.

3 Oct
OpenAI cuts GPT-6 Sol and Luna API prices against GPT-5.6 promotional rates

The first key point said GPT-6 Sol was down from $4 and $20. Those were GPT-5.6 Sol's promotional prices, not an earlier GPT-6 Sol price. It now says so.

Why we corrected it

The key point dropped the condition our checked claim carries: the cut is measured against GPT-5.6 promotional pricing.

What it said before

GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, down from $4 and $20.

Found a mistake? Write to hello@ainotis.com. A correction is dated and kept here for good.

Corrections

Every mistake we have corrected, with the date and what the story said before.