
What happened
In Verge testing, NYT Cooking's AI assistant rewrote a One-Pot Chicken and Rice With Ginger recipe around the tester's ingredients, producing a dish her husband approved, but during baking it gave the modified version's 6 tablespoons of butter instead of the original's three-quarter cup.
Why it matters
The tool, which NYT Cooking says was trained on user comments and not to comply with bad-tasting requests, can be genuinely useful for substitutions and pan swaps, but the tester still felt she had to double-check amounts and times against the original recipe.
Summaries like this, every morning.
The tester had already been using AI as a kitchen helper through Alexa Plus on an Echo smart display and the Samsung Food app on her fridge. What made her try the NYT Cooking assistant was the friction of scrolling through hundreds of reader comments at 7PM on a weeknight, even though those comments had often solved her problems. The assistant, trained on user comments, promised the same answers without the search.
In practice, the tool's value showed in small gaps. Asked to adapt a One-Pot Chicken and Rice With Ginger recipe, it suggested a shorter browning time for tenders and steaming green beans on top of the rice instead of chard. It missed the missing cilantro at first, then accepted Thai basil when asked. The resulting dish took the same amount of time as the original.
Baking exposed the limits. Converting the cake to muffins halved five eggs to three, a smart move. But because the new recipe cannot be saved separately, the original recipe and the AI version blurred together: the assistant answered a butter question with the modified amount without saying which version it was using. The tester recommends deleting your version to avoid exactly that.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Alignable's Eric Groves writes that Meta brought its Muse agent to small businesses in September, while Anthro…

At Fal.Con in Europe on Nov

Nous Research fine-tuned Alibaba's Qwen3.8-27B on 53 million tokens to create Arete-27B, which reduced its pro…

In the "Into the Unknown" benchmark, nine AI agents produced 135 submissions on 15 open research tasks from te…

Owen Queen and James Zou re-tested 50 cited LLM claims across 20 models, finding 44% expired, 34% weakened, 16…

Check Point Research found 1,608 live ARTEX consoles and artifacts from 10 misconfigured ones; one operator's…
