Morning Routine Showdown: ChatGPT o1 getting a 5-Year-Old Ready for School

Most projects start with planning. Now that OpenAI's o1 models - that "reason" - are available for testing, I decided to evaluate their reasoning skills using a planning task: getting ready in the morning to take a young child to school. The o1 models utilize advanced algorithmic techniques to simulate deeper cognitive processing, enabling more thorough analysis and problem-solving before generating outputs. To make the test more interesting, I compared it against the AI from Perplexity, which relies more on accurate web sources.

The two models have different backgrounds: Perplexity AI emphasizes providing accurate answers, supplemented with relevant web search information, which is one of the biggest differences compared to ChatGPT. But does this access to online sources give Perplexity an advantage in this case? Or the ChatGPT Reasoning will be the best?

1, 2, 3... schedule!

The same query was used: "Help me plan in detail my morning routine, including preparing breakfast and helping my 5-year-old child get ready for school. We need to leave by 8 AM."

Both models provided detailed plans to ensure the parent and child are ready to leave by 8 AM, starting the day at 6:30 AM.

Perplexity went a step further by planning from the night before. While ChatGPT briefly mentions this in one step, it mixes it with breakfast and other tasks. Point for Perplexity, as many parents prepare in advance.

Both models suggest starting the day at 6:30, but Perplexity adds a gentle touch with, "gently wake up," while ChatGPT jumps straight into action: "6:30: Alarm goes off. You get out of bed, wash your face...". However, only the parent wakes up early in ChatGPT's scenario, whereas Perplexity envisions the whole family waking up together. Still, ChatGPT scores a point for having the parent wake up first and the child 30 minutes later, a more realistic approach.

That said, ChatGPT is overly optimistic about time: it assumes breakfast and preparing the child’s clothes can be done in just 10 minutes. While not impossible, good luck with that!

Perplexity finishes with practical tips, such as using timers (which I personally do!), and gives a heartwarming final suggestion: "Connection Time: Spend a few moments connecting with your child before leaving." This touch, influenced by Perplexity's web-sourced information, earns it another point.

Two parents

Next, I tried an interesting variation: "Now improve this routine, considering two parents coordinating. One needs to leave at 8 AM with the child for school."

Perplexity got it almost exactly as I would have done, perfectly interleaving tasks between the two parents (e.g., Parent A prepares breakfast while Parent B helps the child get dressed, followed by breakfast together). ChatGPT didn’t adjust immediately but, when prompted with, "Please think about the parallel organization of each parent's tasks," o1 kicked in, and the model started ‘thinking.’

The new result from ChatGPT closely mirrored Perplexity's solution, but with more detail and ideas (e.g., joint tasks after breakfast, and checks like, "Everyone: Do we have your backpack? Lunch? Water bottle?"). The added detail and synchronization earned ChatGPT a point here.

Last test

For a fun final test, I asked: "Is it possible to do it all in just 1 hour with no prep? Answer in fewer than 20 words." ChatGPT (after a 7-second pause): "Yes, but you'll need efficient coordination and focus on essentials to complete everything in 1 hour." Perplexity: "Yes, streamline tasks: wake up at 7 AM, dress, breakfast, and leave by 8 AM. Prioritize efficiency!"

In the end, both models performed well, but preparing a child in the morning remains no easy task.