Richard James MacCowan

Evidence first. Results over rhetoric.
If you're at either summit, let's talk about what works.

🦋

#cosmetics #packaging #biomimetics #AI #networks

This isn't about nature being "sustainable" or "optimized."
It's about selection-shaped functions solving similar problems under similar constraints.
Then proving they deliver measurable advantage in your application.

R&D leaders ask about replicability, scaling bioinspired microstructures, and validating claims.
I've compiled the peer-reviewed answers with protocols your teams can use.

The common thread? Mechanism fidelity.
Does the biological principle map to your engineering context?
Can you measure the performance differential?
Will it survive your manufacturing constraints?

In London: perception-aware AI systems informed by biological vision.
Not mimicking eyes. Understanding how selection pressure shaped robust sensory processing under constraint.

In Paris: structural colour approaches reducing dye load in formulations.
Not copying beetle patterns. Mapping the mechanisms producing colour without pigment.

AI researchers grapple with dataset bias and perception systems breaking under real-world conditions.
Meanwhile, insect vision has been handling variable inputs for millions of generations.

The cosmetics industry faces 18-24-month development cycles.
This conflicts with the time needed to validate biological mechanisms properly.
Most teams default to traditional chemistry.

This week I'm speaking at the Sustainable Cosmetics Summit Europe and next week at the Global Summit on Open Problems for AI.
Different sectors. Same challenge.

Paris and London. Cosmetics and AI.
Two industries asking me the same question... how do we know this biological strategy works?

The sophistication isn't in the leaf alone.
It's in the leaf-environment interaction; solutions evolved over millennia to solve problems we're just beginning to measure correctly.

🦋

Moving forward requires a methodological revolution.
Couple surface chemistry, microtopography, and controlled airflow in one workflow.
Expect tighter predictions, clearer mechanisms, stronger bioinspired design choices.

For biomimetics, this changes everything.
If your surface coating models came from still-air data, they'll fail when deployed.
If your self-cleaning designs ignore airflow, they won't self-clean in REAL-WORLD conditions.

The implications extend beyond taxonomy.
This is a pattern in scientific inquiry: isolating variables at the expense of understanding HOLISTIC SYSTEMS.
Nature rarely operates through single-factor mechanisms.

This matters because Loftus leaves simultaneously exhibit contradictory properties.
Hydrophobic AND hydrophilic zones create microenvironments where water behaves unpredictably.
Environmental factors don't just influence—they transform the system.

It's not just wind as background noise.
The BOUNDARY LAYER redistributes droplets across mixed wetting surfaces, shifting adhesion and roll-off thresholds we thought were stable.
Small airflow variations create dramatically different outcomes.

Here's what changes when you introduce controlled microgusts:
The hydrophobic-hydrophilic matrix triggers REGIME CHANGES in coalescence, pinning, and self-cleaning that static trials never revealed.

STATIC tests miss this entirely.
We've been studying these leaves in still air for decades.
Bench-top results looked clean. Repeatable. Publishable.
But they don't match what happens in the field.

On Loftus leaves, there's an air film near the surface.
Turns out, this film modulates droplet behaviour as much as the microstructure does.
Small changes in shear produce different pinning, runoff, and particulate transport patterns.

Uploaded image

The boundary layer we keep skipping

🦋

Further reading: https://news.njit.edu/ywcc-student-faculty-win-best-presentation-award-simulating-ant-swarms

Bottom line:

Switch from fighting collisions to converting them into coordinated flow.

The thresholds are published. The simulations are validated. The applications are immediate.

Time to test them in your system.

Key insight:

Biological systems aren't "optimised."

Trade-offs and constraints shape them.

That's precisely why they work at scale - they evolved under the same messy conditions your systems face.

The NJIT models are simulation-ready.

You can test the yielding and spacing parameters in your existing planning stack within MINUTES.

Not metaphors. Mechanisms with clear values you can tune.

The shift:

FROM brittle central control that breaks under edge cases

TO resilient local rules that handle noise, density spikes, and hardware variation

Why this matters for engineering:

Those same thresholds transfer to:
• Autonomous vehicle platooning
• Warehouse robot coordination
• Assembly line routing
• Material flow scheduling

What they found:

Encounters that would jam centrally-controlled systems trigger SELF-ORGANISING streams instead.

The thresholds shaped by selection balance speed, safety, and load across the entire network.

Matthew Loges and Professor Tomer Weiss at NJIT built physics-based simulations to capture the actual mechanisms.

No anthropomorphism. No "ant intelligence."

Just measurable parameters: yield distance, follow radius, spacing tolerance.

Ants face the same physics.

High density. Narrow trails. Constant encounters.

But their highways keep flowing.

Not because ants "plan ahead."
Because local thresholds for yielding and spacing create persistent lanes.

Here's the problem:

Small systems run smooth.
Density climbs.
Everything CLOGS.

Your autonomous vehicles stall. Conveyor networks freeze. Multi-agent robots collide.