Mobile attention is collapsing.

In 2018, mobile ads held attention for 3.4 seconds on average.
Today, it’s just 2.2 seconds.

That’s a 35% drop in only 7 years. And a massive challenge for marketers.

The State of Advertising 2025 shows what’s happening and how to adapt.

Get science-backed insights from a year of neuroscience research and top industry trends from 300+ marketing leaders. For free.

👉 Get the free report

For decades, medicine has advanced by learning how to read and modify DNA. But despite breakthroughs in gene therapy, biologics, and vaccines, most treatments are still built through slow, bespoke, trial-and-error processes.

A new research paper from Basecamp Research proposes a radical shift: treating therapeutic design as a programmable engineering problem — powered by foundation models trained on the full diversity of life on Earth.

At the center of this approach is EDEN, a family of massive AI models trained not on human-curated biology alone, but on billions of years of evolution encoded in environmental DNA.

EDEN (Environmentally-Derived Evolutionary Network) is a new class of metagenomic foundation models designed to understand, generate, and manipulate biological sequences at scale.

Key highlights:

Crucially, the training data is enriched for:

This gives EDEN exposure to evolutionary solutions biology has already invented — many of which never appear in traditional biomedical datasets.

Modern drug development is still largely handcrafted. EDEN proposes something different:

Instead of building one model per problem, EDEN treats biology as a generalizable design space — similar to how large language models operate across tasks.

The paper validates EDEN by challenging one model architecture to design therapeutic solutions across three entirely different biological scales.

EDEN was tasked with designing new recombinase enzymes capable of inserting large DNA payloads into specific locations in the human genome — using only 30 base pairs of target DNA as input.

Results:

This opens a path to precise, large-payload gene insertion without viral vectors or double-strand DNA breaks — a major safety and scalability improvement over existing gene-editing methods.

Using the same model, EDEN generated a small library of novel antimicrobial peptides.

The results were striking:

This suggests AI-driven antibiotic discovery may finally outpace resistance — something traditional pipelines have struggled to do.

EDEN was also used to design an entire synthetic microbiome:

More than 1,500 generated species were completely outside the training dataset, yet still exhibited correct ecological behavior — demonstrating true generalization, not memorization.

This work suggests a fundamental shift:

Instead of asking:

The question becomes:

EDEN points toward a future where:

By combining deep evolutionary data with foundation-model scaling laws, this approach could turn therapeutic design into a true engineering discipline.

EDEN isn’t just another AI-for-biology paper — it’s a blueprint for what comes next.

If this approach continues to scale, we may look back on today’s drug development methods the way we now view hand-written assembly code: powerful, but painfully inefficient.

The future of medicine may not be discovered.
It may be designed.