Sponsored by
Hello readers,
Welcome to the AI For All newsletter! Today, we’re talking about how AI is learning to listen for endangered wildlife, model selection vs. harness strategy, and more!
AI in Action: AI is learning to listen for what's disappearing
Researchers across Europe are teaching AI systems to identify wildlife by sound alone. The BioacAI project, a €2.4 million EU-funded doctoral network led by the Naturalis Biodiversity Center in Leiden, is building AI tools that can pick individual species out of raw audio recordings from forests, wetlands, and cities. Nine research institutions and eight additional partners — including the UK's Bat Conservation Trust and France's Natural History Museum — are working together through 2027 to close a gap that has kept bioacoustic AI stuck in prototype mode: almost nobody has the combined expertise in acoustics, ecology, and machine learning to build something field-ready.
Passive acoustic monitors left in the field for months can generate hundreds of gigabytes of data per device, and the Bat Conservation Trust estimates its recordings alone would take 20 to 30 years of human listening to process. BioacAI is training AI on deep embeddings, a technique that maps animal calls into a spatial layout so similar sounds cluster together, letting the system flag likely species matches and surface anything it doesn't recognize for a researcher to check by hand.
Bats are the hardest test case: their navigation calls shift with their surroundings, making species nearly impossible to separate by ear, even a trained one. So the team is instead teaching the AI to recognize bats' social chirps, which vary less and could cut that decades-long backlog down to a fraction of the time. If it works at scale, the payoff extends past bats — birds and insects are both in steep decline, and better acoustic data could reveal new habitats and undetected populations, feeding directly into the EU's 2030 biodiversity targets.
🔥 Rapid Fire
Commentary: Don’t Look Up
NVIDIA’s $500B circular “deal” is just a memorandum of understanding
The MOU may not actually be $500B
The only two customers for this compute can’t afford it
NVIDIA’s $100B MOU with OpenAI went nowhere
Analysis: The Hater’s Guide to NVIDIA (Part 2)
Private credit is under increasing stress with default rates rising
Oracle plans more layoffs to keep the money furnace going
Understanding the surge in CoreWeave’s stock
Addendum: Axios fails to mention that CoreWeave’s revenue backlog is mostly OpenAI and Anthropic, indicating a lack of demand, yet the media reports “rising demand for AI compute”
Further reading: The AI Demand Bubble
AI data center turbines are suffering early deaths
OpenAI continues to lose key executives 🤔
Microsoft kills off unsuccessful AI features, merges Copilot apps
Roku’s AI slop channel is even worse than expected
Find Out What's Holding Your Revenue Back
Your business didn't plateau overnight. Growth slowed one bottleneck at a time.
But more hours at the office, more hires, and more marketing won't remove a hidden limitation. They just make it harder to spot.
The Revenue Ceiling Report reveals the four most common growth ceilings so you can identify what's quietly costing your business revenue and what to do next.
BELAY’s U.S.-based Assistants help build the systems and support that turn sustainable growth into reality.
📖 What We’re Reading
Enterprise artificial intelligence (AI) conversations tend to start with the wrong question: Which model should the company use? While that question matters, it is no longer the one that determines whether an AI initiative can survive production. As agents move beyond demos into customer records, financial workflows, internal knowledge bases, and service processes, the more important issue is the harness engineering system built around the model.
Experience in building production AI systems reveals a simple conclusion: reliability is increasingly a harness-design problem, not just a model-selection challenge. The model supplies reasoning capability. The harness turns that capability into a safe, useful, measurable business outcome.
