Beans and Custard: AI Image Generation and My Perspective on AI in 2026

Beans and Custard: AI Image Generation and My Perspective on AI in 2026

Today, my five-year-old daughter asked me a question that caught me off guard: “What would beans and custard taste like?” I told her I wasn’t sure it would be that great, but her curiosity sparked my imagination. I decided to take her question one step further by using Google’s Gemini image agent to generate a picture of that very combination—just to see her reaction. Watching her reaction was a small but profound reminder of how AI can ignite creativity and curiosity.

Generated with Nano-Banana Pro, and used Photoshop's AI Expand functions too.

Over the past few years, I’ve experimented with a handful of AI agents. Like many, I started with ChatGPT, which quickly became a staple for brainstorming and writing tasks. For image creation, Gemini has become my go-to, especially with its latest Nano-Banana engine (these AI names never fail to amuse me). At work, I rely on CoPilot Business Pro, and at home, both Claude Pro and Claude Code are part of my daily toolkit. The sheer power and convenience of these tools is both exhilarating and - if I’m honest - a little scary. I embrace AI more than most, but I’m also aware of its limitations and risks.

Oh, I also use Adobe Photoshop and Illustrator, which also have some amazing AI generation capabilities. I use the ‘generative fill’ (removes unwanted objects and replaces them seamlessly) and ‘generative expand’ (extends your canvas in any direction, letting AI imagine what's just outside the frame) functions regularly now when designing brochures or enhancing images for websites, etc.

Navigating AI’s Double-Edged Sword

Despite my enthusiasm, I share the concerns of many who are hesitant to fully trust AI. The temptation to copy-paste chunks of code—including sensitive data like API keys or passwords—is real, and it’s a genuine security risk. On the flip side, letting tools like Claude Code understand live endpoints can unlock powerful automation and insight. Yet, the conversation about job security looms large. Robotics and AI are advancing rapidly, and while we’re okay for now, it’s clear that many non-laborious and labour-intensive jobs are at risk of automation. Warehouse pickers, factory workers, agricultural labourers, cleaners, fast-food kitchen staff, delivery drivers, and textile workers are just some roles most likely to be replaced soon.

Conversely, roles that require emotional intelligence, complex physical skills, creativity, or leadership are less likely to disappear. Healthcare professionals, skilled tradespeople, creative directors, educators, social workers, and rescue workers are relatively safe—for now. These jobs combine unpredictability, human connection, and expertise that AI still struggles to replicate. The BBC recently examined this divide, highlighting that jobs with a strong human element are the most resilient against automation (source).

AI as a Creative and Educational Partner

Back to the beans and custard moment - Gemini's image generation genuinely impressed me, and I'm convinced it's currently ahead of the pack when it comes to bringing weird and wonderful prompts to life (the engine names never stop amusing me either).

I've also been exploring AI as a creative tool with my daughter, always with supervision. Recently, she drew a picture in biro on plain A4 paper, and I fed it into Gemini, asking it to generate a realistic version of her creation - the result was brilliant. See here:

We also visited Wendover Woods, which has a fantastic trail featuring Julia Donaldson characters — the Gruffalo, Room on a Broom, Stick Man — as well as a Wallace and Gromit trail (though fair warning: it's quite steep, so perhaps one for when the little ones are a bit older). Along the Gruffalo trail, I snapped photos at various points and later used AI to enhance them — even with fairly vague prompts, it understood exactly what I was going for, and the results were stunning.

And Stickman...

Practical AI in Everyday Life

For my own projects, Claude Code has become invaluable for debugging, suggesting alternative solutions, and even planning new websites. I’ve started giving it access to certain database structures, and its ability to help tighten up security is impressive. Still, there’s always a lingering worry about accidentally sharing sensitive data. While AI platforms are designed to protect privacy, the future is uncertain. Companies must also consider how employees use these tools—can we truly ringfence AI use to prevent information leaks?

Not everyone shares my comfort with AI. At work, I once shared AI-generated suggestions to improve a colleague’s writing, only for it to be taken personally and escalated to HR. This was a wake-up call: AI isn’t for everyone, and introducing its input can be sensitive. AI’s strengths - offering endless suggestions for improvement - can also become a source of friction if not handled with care.

AI writing tools are creating mixed results in customer service, too. While they help departments respond more efficiently, they also sometimes generate more work, as customers become savvier and expect higher standards. It’s a classic catch-22 scenario, and there are always pros and cons in this evolving landscape.

Security Risks and the Future

AI’s power isn’t limited to productivity. It’s also being used by malicious actors to identify loopholes in websites and systems, just as anti-virus software is always playing catch-up with new threats. As this Medium article points out, the exponential growth of AI mirrors the cautionary tales of films like iRobot. As we approach 2035—the year in which iRobot is set—the lines between science fiction and reality are starting to blur. AI’s demand for resources and its capacity for both good and harm continue to grow.

Beans and Custard - A Metaphor for AI

Coming full circle to my daughter’s original question, after generating the image, it’s certainly put me off trying the real thing! But in a way, “beans and custard” is a perfect metaphor for AI in 2026: a strange, unexpected combination that can be delightful, unsettling, or both - depending on how we choose to engage with it.

AI is here to stay. We can resist, adapt, or embrace it - but ultimately, it’s our creativity, curiosity, and humanity that will determine whether AI is a force for good or something to fear.

Written by David Ferdinando