Every AI boom creates more costs. Security gaps, data restrictions, and ballooning video files are the quiet expenses underneath the hood in the AI era. Three startups, each led by a founder
Every AI boom creates more costs. Security gaps, data restrictions, and ballooning video files are the quiet expenses underneath the hood in the AI era.
Three startups, each led by a founder who spent years inside the affected industry, are turning those costs into businesses. All three build on Bittensor, a decentralized network where contributors compete to improve AI models.
Their founders told TheStreet Roundtable what they are fixing and why.
Related: Roundtable 100 analysts move Super League 71 places lower
John Yu has been programming professionally for more than 10 years and ran smart contracts serving millions of users a month.
Then one was exploited for $200,000. Over a weekend, he spent $1,000 on AI credits to see whether a model could find the same flaw. It did.
“$1,000 versus $200,000. There is a huge difference,” he said.
That experiment became Bitsec.
Traditional security audits cost at least $20,000 and take months, Bitsec scans code continuously and returns issues within days. Yu said its system found more vulnerabilities in a client’s code base than a newly released frontier model.
With AI models growing better at finding flaws, his warning to teams without AI security is blunt.
“2027 could be a bad year for you.”
Mamad Ahangari holds a PhD in genomics and has spent the past decade in genomics research. He knows the field’s central obstacle well: genetic data is locked away for privacy reasons, so AI has had little to learn from.
TheMinos generates synthetic genomic data, which allows AI models to train without touching any real person’s records. The data can be tailored to specific diseases and specific regions of the genome.
Potential users include AI labs, drug developers, sequencing companies and universities. A second phase, launching soon, is meant to speed up data generation.
Garreth Howells has spent 25 years in the video and film industry and spent parts of his career with Netflix, Amazon, Disney and the BBC.
While launching a video on demand service in Africa, he could not get high quality files to phones over weak connections.
Vidaio handles compression, upscaling, restoration and dubbing. Its benchmarks show files about 43% smaller, a figure Howells says matters enormously once you consider the size of major content archives and what distributors pay in storage and bandwidth.
“To try and get that type of R&D in a normal company, you’d have to spend millions,” he said. “we’re definitely ahead of the curve.”
Proof of concept is under way, and after showcasing the technology at a major media tech show in Amsterdam.
These companies are helping their customers create cost efficiencies. Yu, Ahangari and Howells know those costs firsthand, and leveraging a decentralized network gives them the computational power to address them.
