How Taktile Labs' lobster stunt got Wall Street talking about reliable AI

On the morning of March 10, 2026, commuters in Manhattan's Financial District walked past something that wasn't there the day before: a giant bronze lobster facing down the Charging Bull.

Within hours, the internet had questions. Posts from Y Combinator's Harj Taggar, OpenClaw’s Peter Steinberger, and Founders Fund's Mike Solana ignited a viral trend across X and LinkedIn. By the end of the day, the stunt had racked up more than 10 million views. Nobody could agree on what it meant, who was behind it, or whether the photos were even real.

It was the beginning of a very different kind of launch, as Taktile prepared to introduce the financial services industry to its new AI research institute: Taktile Labs.

In the weeks before the launch, the lobster had become a symbol of what happens when powerful AI goes unsupervised. Off-the-leash agents were making unauthorized payments, leaking credentials, and making decisions they should never have been trusted to automate.

We dropped a giant lobster on Wall Street to spark a conversation: What if these uncontrolled agents came to regulated banks?

Then, we used our viral moment to launch Taktile's new applied AI research institute, Taktile Labs. The mission: to bridge the gap between what frontier models can now do—and what regulated institutions need to trust agents with their most critical decisions.

Every day, banks, fintechs, and insurers run on a current of critical decisions. Through complex and often highly manual processes, teams analyze potential fraud, assess credit risk, and research customer histories for AML compliance. The promise of AI automating these workflows has been building for years.

But even as models have reached a new level of capability in recent months, trust is still lagging.

For regulated financial institutions, accuracy must be exceptionally high as erroneous decisions can cost millions or a license to operate. Critically, the leaders tasked with driving AI adoption are the same people whose careers are on the line if something goes wrong.

Taktile Labs is designed to bridge this gap between potential and reality. Our mission is to give financial services leaders the concrete, domain-specific research they need to deploy AI with confidence, with three key pillars setting our work apart from generic AI research labs:

Already, Taktile Labs is cultivating research that helps leaders answer essential questions like which model configurations are the most reliable in core use cases.

With our first benchmark testing the latest models in real-world banking use cases, we found that models released in December 2025 now exceed human performance in financial spreading: 96%+ vs. 89%.

This common yet complex task is a cornerstone of decisions in business loan underwriting. That AI can now reliably automate it is a yardstick for what’s to come. We believe financial services leaders need exactly this kind of concrete, domain-specific evidence to balance fast implementation with reliable performance.

Taktile Labs is guided by an external Research Council and Advisory Board—practitioners and academics working at the intersection of AI, financial services, and regulation. Their input shapes our research priorities and ensures our work stays grounded in the real challenges institutions face.

Advisory Board

Research Council

Taktile Labs will publish applied benchmarks, research reports, and practical guidance for teams deploying AI in high-stakes financial decisions.

If you're building in the space and want to collaborate on evaluations, we'd love to hear from you.