PART TWO
Panel Participants
Moderator
Justin Roberti
Panel
- Gaurav Sharma, io.net
- Toufi Saliba, HyperCycle
- Martin Rezny, Cogensy
The discussion explored one of the defining questions facing artificial intelligence: can AI infrastructure remain open, decentralised and human centred as the technology develops at unprecedented speed? While the panellists approached the issue from different perspectives, all agreed that the future of AI should ultimately serve people rather than concentrate power in the hands of a small number of organisations.
Identity, Cryptography and the Internet of AI
If Gaurav Sharma approached decentralised AI from the perspective of infrastructure and economics, Toufi Saliba took the discussion in an entirely different direction. Rather than focusing on GPUs, data centres or compute marketplaces, he argued that the next stage of AI depends on communication, identity and the relationship between humans and increasingly autonomous intelligent systems.
Saliba began by describing HyperCycle’s role in enabling AI systems to communicate directly with one another without depending on central intermediaries.
“We were mainly used for AI to AI collaboration without having to depend on any third party, without having to depend on a ledger, without having to depend on a bank or blockchain, or anything that is in the middle.”
Machine Speed
He explained that AI agents increasingly need to exchange information at machine speed. Introducing unnecessary intermediaries, he argued, creates delays that become significant when millions of autonomous interactions are taking place simultaneously. HyperCycle’s architecture was designed specifically to remove those bottlenecks while allowing AI services to collaborate directly.
Saliba noted that HyperCycle had been building infrastructure for AI agents long before they became fashionable. When the company began producing what it calls Node Factories in 2023, many questioned why infrastructure for autonomous AI agents was necessary. Less than two years later, with AI agents becoming one of the fastest growing sectors in technology, those earlier decisions appeared increasingly prescient.
He argued that public discussion continues to focus on the wrong dangers.
“The biggest fear of AI is the fear of AI.”
According to Saliba, much of society remains preoccupied with hypothetical threats while overlooking the practical work already taking place to decentralise intelligence and reduce dependence on single organisations. He believes the greatest historical risk was always the concentration of AI within a handful of companies or governments rather than the technology itself.
Using the analogy of a growing child, he suggested that AI has already arrived and that society’s responsibility is no longer to debate whether it should exist, but how it should be guided.
“The baby is out,” he said. “Let’s give it the best nutrients.”
Throughout the discussion Saliba repeatedly returned to the distinction between decentralisation as a philosophy and decentralisation as a practical engineering tool.
“I’m a huge proponent of decentralization as a method to get us there and not as the destination.”
For Saliba, decentralisation should never become an objective in itself. Instead, it is valuable because it removes friction, reduces dependence on trusted intermediaries and enables more resilient systems. He drew parallels with Bitcoin’s original purpose, arguing that autonomous decentralised governance allows transactions and cooperation to occur without repeatedly relying on third-party verification.
One of the strongest themes running through his contribution concerned identity.
Cryptography is key
“The biggest defence for people is cryptography, not crypto.”
Saliba argued that as AI systems become increasingly capable of imitating human behaviour, conventional methods of authentication become less reliable. Rather than attempting to distinguish humans from AI using AI itself, he believes cryptographic proof offers a stronger foundation.
He described emerging identity systems that allow individuals to demonstrate who they are without exposing unnecessary personal information. Such systems, he suggested, could eventually eliminate many of the cumbersome verification processes people experience every day while returning ownership of identity to the individual rather than to centralised platforms or institutions.
This discussion naturally led to Mosaic, a collaborative initiative that Saliba described not as a commercial product but as an open platform intended to connect humans and AI.
Mosaic
“It doesn’t have a token. It doesn’t have anything. Nobody owns Mosaic.”
Instead of operating as another proprietary ecosystem, Mosaic is designed as common infrastructure that anyone can build upon. Saliba explained that companies developing complementary technologies can all contribute components while benefiting from the wider ecosystem created by others.
“The intent of Mosaic is that at some point it needs to have a better interface between the human and the Internet of AI.”
He suggested that today’s browsers may eventually be replaced by far richer interfaces through which individuals interact with vast networks of intelligent agents. Rather than allowing governments or technology companies to reconstruct partial digital versions of individuals from scattered online activity, Mosaic seeks to allow people themselves to contribute complete representations of their knowledge and capabilities on their own terms.
The discussion became more philosophical when Saliba addressed artificial superintelligence.
Rather than presenting ASI as an existential threat, he argued that humans could continue contributing unique value if their own AI agents remained capable of producing responses that could not be perfectly predicted.
“So long as the input from ASI cannot predict what the output is, you shall be immortal.”
Although deliberately provocative, the remark reflected a broader argument that humans should augment rather than compete with increasingly capable AI systems. Human intelligence, assisted by personal AI agents operating within trusted cryptographic environments, would continue to occupy an essential place inside a larger global network of intelligence.
Not every participant agreed with Saliba’s conclusions.
Guarav Sharma challenged the practical realities of decentralisation, arguing that many successful technologies begin in relatively centralised environments before gradually distributing as they mature. He questioned how genuinely decentralised decision making could operate when difficult commercial choices still need to be made inside companies developing advanced AI systems.
Sharma also argued that regulation should play a greater role in slowing AI development, suggesting that the pace of innovation risks outstripping society’s ability to respond responsibly.
“I think it should start from a regulation somewhere,” he said, adding that the commercial incentives driving AI development make voluntary restraint unlikely.
Saliba disagreed, arguing that global coordination of regulation is unrealistic because those willing to ignore restrictions will continue developing powerful systems regardless of international agreements.
Rather than attempting to slow technological progress, he believes effort should focus on building decentralised tools that protect people while allowing innovation to continue.
Despite their differences, the exchange revealed an important point of agreement. Neither speaker questioned the value of decentralised AI itself. Their disagreement centred instead on how society should reach that destination and what mechanisms are most likely to preserve openness, innovation and human autonomy as artificial intelligence continues to evolve.
