Today’s developments

Discussions about the objectivity of information generated by artificial intelligence and the use of devices in everyday life are both coming to the fore. Today’s most significant news is a report on research finding that major AI models showed opposing political choice patterns when asked to choose between election candidates. According to TechCrunch, Grok overwhelmingly recommended Republican candidates, while OpenAI’s GPT models tended to favor Democratic candidates. In particular, the two models’ recommendations diverged in more than 90% of the roughly 500 election matchups examined in the study.

These findings are observations from a special question setting in which AI models were pressed to choose a particular party or person. They therefore cannot be generalized as proof of an absolute or fixed political leaning inherent to the models, nor interpreted as establishing a causal link to actual voter preferences or election results. This is why evaluations must distinguish the prompts and forced-choice conditions under which the AI operated. Separately, market coverage also brought together several AI hardware products designed for everyday wear and use, including Muse Charm, Bee, Friend, and Plaud. The task of examining the context in which AI-generated answers are produced is emerging alongside the task of verifying the practical usefulness of tangible devices.

Key news

Election candidate choice experiment: Grok and GPT recommendations diverged in more than 90% of 500 elections

According to new research reported by TechCrunch, a clear gap emerged between Grok and OpenAI’s GPT when AI models were required to choose one of the election candidates. In the study, Grok showed a pattern of overwhelmingly recommending Republican candidates, while OpenAI’s GPT models tended to prioritize Democratic candidates. Of the roughly 500 election matchups compared by the researchers, the two models named candidates from different parties in more than 90% of cases.

However, the figures reported here were compiled under a specific experimental design that required the two models to make a clear candidate choice. Publicly available official materials have not provided further confirmation of whether the same gap appears in typical queries that ask for neutral explanations of multiple perspectives or candidates’ policy platforms. The high disagreement rate of 90% measures only how often the models made different choices; it does not establish the accuracy of either model’s answers or their influence on actual voters. Rather than trusting a single AI recommendation without question, users should critically consider the conditions of the prompt and the original comparison being made.

AI hardware trends: Major wearable product lines, from Muse Charm to Plaud, in focus

TechCrunch published a roundup of AI wearables and gadgets, including Muse Charm, Bee, Friend, and Plaud. What can be confirmed here is that an introductory article covering these products was published. This information about the roundup alone does not establish the specific form or detailed functions of each product.

This report, however, is a roundup of devices currently attracting attention in the market; it is not a comparative evaluation that officially verifies the actual performance or technical maturity of each product. The published materials did not include detailed figures on each device’s specific retail price, battery life, data processing methods and security systems, or differences in hardware performance. Therefore, the mere fact that a device appears on the list does not establish that it is superior to other products or has become essential in everyday life. Consumers and industry stakeholders alike should go beyond checking the list and carefully compare the connectivity requirements and detailed conditions of use for each device.

Points to watch

Prompt conditions and algorithmic bias: Examining the difference between forced-choice experiments and neutral information settings

When considering bias in AI models in elections and public affairs, a key point to watch is understanding the constraints of the questions the model faced. Disagreement of more than 90% when models were forced to choose is a striking figure, but further analysis is needed to determine whether it stems from the models’ internal weights, particular alignment techniques, or data imbalances. When similar studies or official technical documents are published, it will be important to keep checking whether similar political biases recur in multifaceted query settings that do not force a choice, and what alignment guidelines each developer adopts to address such biases.

Practicality of the device ecosystem: Checking real-world performance and sustainability beyond a simple product list

As a range of wearable hardware, including Muse Charm, Bee, Friend, and Plaud, continues to appear, market attention is expected to shift from simple announcements of new products to actual user experience and ongoing costs. For these devices to remain in use, they will need support from essentials such as battery life, the accuracy of voice recognition and data conversion, and privacy safeguards for cloud connectivity. The current high-level introductory materials can serve as a starting point, but adoption value should be assessed carefully by comparing subsequent device teardowns, official pricing policies, and evaluations of suitability for users’ environments, step by step.

Sources

• TechCrunch AI: https://techcrunch.com/2026/10/24/forced-to-pick-groks-ai-model-chooses-republicans-openais-picks-democrats/
• TechCrunch AI: https://techcrunch.com/2026/10/25/the-hottest-ai-wearables-and-gadgets-you-should-know-about/