Message très juste sur l’évolution de ChatGPT vers des interfaces plus naturelles et multimodales. On sent bien l’ambition de réduire la distance entre l’intention et l’action.
De mon côté, en utilisant l’IA au quotidien sur des problèmes très concrets (techniques, humains, pédagogiques), je me rends compte qu’une autre couche devient clé en parallèle de l’UI : la calibration de l’utilisateur.
Avant d’accélérer, aider à comprendre comment une personne pense, apprend, résiste à la friction, ou se met en difficulté. Sinon, on risque parfois de supprimer trop vite une friction qui était justement structurante pour l’apprentissage et l’autonomie.
Les interfaces riches rapprochent l’IA de notre manière naturelle de créer.
La compréhension fine de l’utilisateur permet, elle, de rester juste dans l’aide apportée.
Les deux combinées ouvrent, je pense, quelque chose de vraiment puissant et responsable.
So true. Every different type of user task would benefit so much more if it’s offered through an intuitive interface. Creating visuals needs a more visual interface. For example my dad just couldn’t get the hang of what to type to get what he wanted.
Similarly a shopping experience would need a different experience, it’s okay to tell a chatbot that you need a specific phone at a specific price, whole other ball game to buy a Christmas present for your toddler niece.
This nails the future of AI. Multimodal, interactive tools bridge intention and output, making AI genuinely actionable while keeping context intact.
I talk about the latest AI trends and insights. If you’re interested in practical AI strategies for creators and building ideas visually, check out my Substack. I’m sure you’ll find it very relevant and relatable.
Congratulations Fidji, you're doing an exceptional job. My only regret is that business users don't have access to many of the tools we consider essential: Image, Atlas, Group Chats, Pulse...
Exciting to see more visuals and entity taps in ChatGPT to explore inline, instead of having to constantly look them up on other platforms and learn more about them.
Congratulations on the launch! I’m already seeing notable improvements, specifically while generating brand elements for a new site I’m vibe coding. And the shift toward a dedicated visual space really makes a difference in the workflow.
Our surveys of 120,557 consumers fielded between March and December 2025 shows a great competition: ChatGPT: 17% of users create/edit images, 11% create/edit video. Gemini: 20% create/edit images and 11% create/edit video.
It's been a tight race, with ChatGPT currently leading on creation/video editing NPS +11 to Google NPS +10. Since the launch of Nano Banana, for create/edit images Google jumped ahead NPS +14, to ChatGPT NPS +10. UX is clearly an opportunity to improve the experience and output quality.
Your data efforts clearly align with a report I'm soon to publish on the transformative value created by better consumer access to their islands of confidential data. Despite model improvements, paid subscription growth is still modest, survey details show this as a significant unlock for AI value. (Fidji, it will be great to get you an early draft for your thoughts!)
Message très juste sur l’évolution de ChatGPT vers des interfaces plus naturelles et multimodales. On sent bien l’ambition de réduire la distance entre l’intention et l’action.
De mon côté, en utilisant l’IA au quotidien sur des problèmes très concrets (techniques, humains, pédagogiques), je me rends compte qu’une autre couche devient clé en parallèle de l’UI : la calibration de l’utilisateur.
Avant d’accélérer, aider à comprendre comment une personne pense, apprend, résiste à la friction, ou se met en difficulté. Sinon, on risque parfois de supprimer trop vite une friction qui était justement structurante pour l’apprentissage et l’autonomie.
Les interfaces riches rapprochent l’IA de notre manière naturelle de créer.
La compréhension fine de l’utilisateur permet, elle, de rester juste dans l’aide apportée.
Les deux combinées ouvrent, je pense, quelque chose de vraiment puissant et responsable.
So true. Every different type of user task would benefit so much more if it’s offered through an intuitive interface. Creating visuals needs a more visual interface. For example my dad just couldn’t get the hang of what to type to get what he wanted.
Similarly a shopping experience would need a different experience, it’s okay to tell a chatbot that you need a specific phone at a specific price, whole other ball game to buy a Christmas present for your toddler niece.
This nails the future of AI. Multimodal, interactive tools bridge intention and output, making AI genuinely actionable while keeping context intact.
I talk about the latest AI trends and insights. If you’re interested in practical AI strategies for creators and building ideas visually, check out my Substack. I’m sure you’ll find it very relevant and relatable.
Congratulations Fidji, you're doing an exceptional job. My only regret is that business users don't have access to many of the tools we consider essential: Image, Atlas, Group Chats, Pulse...
Exciting to see more visuals and entity taps in ChatGPT to explore inline, instead of having to constantly look them up on other platforms and learn more about them.
Congratulations on the launch! I’m already seeing notable improvements, specifically while generating brand elements for a new site I’m vibe coding. And the shift toward a dedicated visual space really makes a difference in the workflow.
Our surveys of 120,557 consumers fielded between March and December 2025 shows a great competition: ChatGPT: 17% of users create/edit images, 11% create/edit video. Gemini: 20% create/edit images and 11% create/edit video.
It's been a tight race, with ChatGPT currently leading on creation/video editing NPS +11 to Google NPS +10. Since the launch of Nano Banana, for create/edit images Google jumped ahead NPS +14, to ChatGPT NPS +10. UX is clearly an opportunity to improve the experience and output quality.
Your data efforts clearly align with a report I'm soon to publish on the transformative value created by better consumer access to their islands of confidential data. Despite model improvements, paid subscription growth is still modest, survey details show this as a significant unlock for AI value. (Fidji, it will be great to get you an early draft for your thoughts!)
I can’t wait for simple use cases like: “I’m looking for a Christmas movie”, and getting posters as the answer.
Understanding user preferences is easy for ChatGPT. What’s more complex is where the movie is available (for the user asking the question).
Very insightful
Text is always the Non Lovable MVP. Images is when products become real!