Artificial intelligence can now write software, summarize meetings, and generate polished images in seconds. Yet one of OpenAI’s own leaders believes the technology still struggles with the part of creativity that matters most, which is deciding what truly feels original, meaningful, and culturally relevant.
That distinction matters far beyond artists and designers. It also affects smart home products, consumer apps, and connected devices, where great experiences often depend on subtle design choices that cannot be measured as easily as software performance or technical accuracy.
Why coding gives AI a major advantage
Andrew Ambrosino, who leads OpenAI’s Codex efforts, recently explained that programming gives artificial intelligence something creative fields often cannot. Software has objective answers because code either compiles, passes tests, or fails.
Design follows a different path. A user interface, industrial design, or visual identity rarely has one correct answer. Good taste depends on culture, emotion, timing, and audience expectations, making it much harder for artificial intelligence to learn what consistently qualifies as outstanding creative work.
Why taste remains difficult to teach
Large language models learn patterns from enormous collections of text, code, and other data. They become skilled at predicting likely outputs rather than independently understanding beauty, originality, or cultural significance through direct experience or personal judgment.
That limitation means models often produce work resembling existing examples instead of creating something genuinely unexpected. While results can appear polished, they frequently settle near the average of previous designs rather than introducing bold concepts that redefine products or influence future trends.
How Codex became a powerful coding assistant
Codex and newer coding-focused models combine advanced language understanding with software reasoning. They can generate functions, explain unfamiliar code, refactor existing projects, and even help complete longer engineering workflows using natural language instructions from developers.
Those capabilities continue improving because programming offers measurable verification. Engineers can immediately test generated code, identify mistakes, and automatically evaluate performance. Every successful verification strengthens confidence that the model solved the intended problem correctly instead of merely appearing convincing.
Little-known fact: Inside OpenAI itself, roughly 95% of engineers use Codex weekly, and those users submit about 60% more pull requests than non-users.
What modern coding agents do well
Recent Codex-focused releases also emphasize longer technical workflows instead of isolated programming suggestions. Instead of producing only a single code snippet, these systems increasingly manage connected tasks, organize projects, and help developers maintain larger software environments with greater consistency.
That shift makes them valuable productivity tools for companies building consumer technology, connected devices, and smart home ecosystems. Engineering teams can spend less time creating repetitive boilerplate code while focusing more attention on architecture, security, and customer-facing innovation.
Why creative work follows different rules
Creative professionals rarely receive the kind of objective feedback programmers enjoy. A new logo cannot simply pass a unit test. An app interface cannot automatically prove that users will emotionally connect with its visual style before widespread real-world testing begins.
Instead, successful creative work depends on human interpretation. Designers balance accessibility, cultural expectations, brand identity, and emotional appeal simultaneously. Those overlapping priorities frequently change across countries, age groups, and product categories, making universal evaluation extremely difficult for artificial intelligence alone.
What research says about AI creativity
Academic research presents a more balanced picture than either extreme position. Studies involving more than 100,000 participants indicate that advanced artificial intelligence performs surprisingly well on structured creativity assessments compared with average human participants across several measured categories.
However, those same findings consistently show that elite human creators continue outperforming artificial intelligence on richer creative challenges. Storytelling, nuanced writing, and highly original artistic work still benefit from uniquely human judgment, lived experience, and intentional creative direction.
Little-known fact: The 2026 study behind this research compared leading AI models against a dataset of 100,000 human participants using a word-association creativity test.
Why this matters for smart home products
Smart home devices compete through more than hardware specifications. Consumers often choose products because applications feel intuitive, notifications remain helpful, and physical devices blend naturally into living spaces without creating unnecessary visual distractions or confusing interactions.
Those qualities depend heavily on thoughtful design decisions. A technically capable security camera may still frustrate users if notifications interrupt family routines, menus confuse first-time owners, or onboarding instructions ignore everyday household habits and expectations.
Human testing still shapes better experiences
Artificial intelligence can rapidly suggest layouts, interface variations, and automation ideas during product development. That speed helps teams explore possibilities they might otherwise overlook, especially during early brainstorming or rapid prototype creation before expensive engineering work begins.
Even so, companies still rely on usability studies, A/B testing, and customer feedback before launching finished products. Human participants reveal confusing interactions, unexpected behaviors, and emotional reactions that automated evaluation methods cannot fully predict with consistent reliability.
AI works best as a creative partner
Ambrosino‘s perspective aligns with many technology companies using artificial intelligence internally today. Rather than replacing designers, writers, or product specialists, organizations increasingly position these systems as assistants that accelerate repetitive work without making final creative decisions independently.
That collaborative model already appears throughout software development. Engineers frequently accept useful suggestions while rejecting weaker recommendations. Creative teams follow similar patterns by generating concepts with artificial intelligence before refining messaging, visuals, and experiences through professional expertise and human discussion.
Industry leaders share similar concerns
Several executives across the technology industry have expressed comparable views. Figma Chief Executive Officer Dylan Field has argued that artificial intelligence often produces competent but predictable designs because models learn from existing distributions instead of developing authentic creative taste independently.
Creative agencies have also questioned whether automated systems can replace experienced designers responsible for defining brand identity. Successful campaigns frequently rely on cultural awareness, emotional timing, and distinctive storytelling that extend beyond statistically likely outputs generated from historical examples.
Could better feedback improve creative AI?
Researchers continue exploring methods that could narrow this gap. One promising direction involves creating measurable design evaluation systems using customer satisfaction, behavioral analytics, conversion performance, and long-term engagement as practical signals for future model improvement.
Those approaches could gradually provide stronger learning objectives than subjective preferences alone. Still, many experts believe certain aspects of creativity will remain deeply connected to human experiences, cultural understanding, and emotional interpretation that resist straightforward numerical evaluation.
What consumers should expect next?
For consumers, the practical message remains encouraging. Artificial intelligence will likely continue making smart home products easier to build, update, and personalize while reducing development time for software features behind connected household devices.
The final experiences people remember, however, will probably continue reflecting human judgment. Whether designing companion apps, choosing interface animations, or refining product personalities, successful technology companies still depend on creative professionals who understand people as well as machines understand patterns.
TL;DR
- OpenAI Codex leader Andrew Ambrosino argues that artificial intelligence excels at coding because software provides objective correctness signals, while creative fields rarely offer comparable feedback for learning good judgment.
- Modern coding models continue to improve rapidly because generated software can be tested automatically, allowing developers and systems to verify results through measurable performance.
- Research shows that advanced artificial intelligence can outperform ordinary people on structured creativity tasks, but top human creators still lead in storytelling and design.
- Smart home companies can use artificial intelligence to accelerate engineering, prototyping, and brainstorming, while relying on designers and user testing to finalize customer experiences and product identity.
- The strongest long-term strategy combines artificial intelligence with human expertise, allowing automation to increase productivity while preserving creative judgment where taste and cultural understanding remain essential.
This article was made with AI assistance and human editing.
If you liked this, you might also like:
