Light, Letterforms, and the Cost of Faking It

We've spent months talking about AI replacing production work — the wireframes, the prototypes, the first drafts. Fair enough. But today's stories pull focus to something more interesting: the stubborn, granular details that separate output that feels right from output that doesn't. A team teaching AI to handle light. A typeface engineering randomness to feel human. And two global brands learning, painfully, that audiences can smell inauthenticity from a mile away.

The thread connecting all of it? Realness is a design problem. And it's harder than it looks.

AI Is Learning to See Light

Creative Bloq published a fascinating piece on a team that has been teaching AI systems to understand the physical behaviour of light — reflections, shadows, colour temperature, the way illumination interacts with surfaces in the real world [Creative Bloq]. If you've spent any time with AI-generated imagery, you know the problem they're solving. The outputs often look almost right: glossy, detailed, compositionally competent. But something is off. The light doesn't land correctly. Shadows contradict each other. A reflection bends the wrong way.

These aren't aesthetic nitpicks. They're the visual equivalent of the uncanny valley. Our brains have spent a lifetime learning how light works — we notice when it's wrong even if we can't articulate why. For artists and filmmakers, this breakthrough matters because it moves AI-generated visuals from "impressive demo" toward "actually usable in production."

But here's what designers should take from this: the hardest problems in AI-generated content aren't about more pixels or faster rendering. They're about physical truth. The thing that makes an image feel real isn't resolution — it's whether the light behaves like light. AI is getting there. But the fact that it took focused, deliberate work to teach a system something our eyes do unconsciously should tell you how much craft knowledge is embedded in what designers take for granted.

A Font That Tries to Forget Itself

Meanwhile, in typography, a new handwriting font called Rando Sans is tackling another authenticity gap — one that's plagued handwriting typefaces forever [Creative Bloq]. The problem: when you type two of the same letter, they come out identical. Which is exactly what real handwriting never does. Your two lowercase e's are always slightly different. A digital font that repeats them perfectly announces itself as fake.

Rando Sans introduces built-in variation so repeated characters look different each time, mimicking the organic inconsistency of actual handwriting. Creative Bloq's headline calls it "Comic Sans on steroids," which undersells the sophistication here. This is a typeface that has been engineered to appear unengineered.

That tension — designing systems that produce authentic-feeling results — is one of the defining challenges in contemporary design. Whether it's AI learning to place a shadow correctly or a font randomising its own letterforms, the goal is the same: bridge the gap between computation and human perception. The craft isn't in the output. It's in understanding why the output feels wrong, and building the intelligence to fix it.

When a Brand Fakes a New Identity

The same principle applies at the brand level, and the stakes get much higher. Creative Bloq examines the dilemma companies face when a rebrand goes wrong — using Jaguar's controversial identity overhaul and HBO's brief transformation into just "Max" as case studies [Creative Bloq]. The core question: should a brand commit to a new direction that the public has rejected, or revert?

HBO ultimately restored its legacy name alongside the Max streaming brand, effectively admitting the rebrand had undervalued decades of built-up equity. Jaguar is still navigating the fallout.

For designers, the lesson isn't simply "bold rebrands are risky" — we covered creative conviction in an earlier briefing. The newer, sharper insight here is about irreversibility costs. A rebrand isn't just a new logo. It's a new promise. And if that promise doesn't ring true — if it feels like a costume rather than a genuine evolution — audiences will treat it as a betrayal. Reverting is expensive and confusing. Staying the course with something inauthentic is worse. The only real protection is doing the strategic homework before launch, not after.

Resource Worth Your Time

Smashing Magazine is now shipping Accessible UX Research by Michele Williams [Smashing Magazine], a practical guide to embedding accessibility into the research process itself — not just the outputs. If we're talking about authenticity in design, inclusive research is where it starts. You can't design for real people if your research excludes entire groups of them. Worth a look for any team serious about research that reflects actual human diversity.

The Bigger Picture

There's a pattern forming across these stories that goes beyond any single tool or trend. As AI gets faster and more capable, the bottleneck is shifting from production to perception. Generating a layout is easy. Generating a layout where the light feels real, the type feels human, and the brand feels true? That's still hard.

This is good news for designers — but only the ones who have developed the sensitivity to notice what's off. The value isn't in making things. It's in seeing things. Knowing that a shadow is falling wrong. Feeling that a letterform is too uniform. Recognising that a rebrand is strategically hollow before it ships.

Takeaway

The most interesting design work right now isn't about speed. It's about closing the gap between what AI can produce and what humans actually believe. Light that behaves like light. Letters that look like they came from a hand. Brands that feel like they mean it.

AI is getting better at faking realness. Your job is knowing the difference — and that knowledge comes from craft, not prompts.