Swept Interviews: Jack Vinijtrongjit from Saakuru Labs
It got off to a bumpy start, our call.
Try as we might, we couldn’t force Google Meet to record our call without upgrading the account again. This was a feature which we each erroneously thought was offered on any paid tier.
After other attempts to record what Saakuru was up to in the AI space with CEO Jack Vinijtrongjit, we became worried that it just wouldn’t be possible. Where there’s a will, there’s a way.
The Saakuru Agentic Framework
Saakuru has been building an agentic framework, known as the Saakuru Agentic Framework. This framework aims to make AI agents applicable to games on Saakuru and offer integrations for other consumer-centric applications across the protocol.
Jack also expects it to significantly streamline agent deployment on Web3 games to volumes topping thousands of agents per game deployed in a matter of minutes.
Jack affirmed that Saakuru isn’t pivoting. Realistically, if AI product offerings indicated a pivot at tech and financial firms, the market would be reeling at all the jarring announcements of the like. As with many others across Web3 and beyond, AI is being used to enhance existing technologies and future-proof them. Within that context, Jack said, “It doesn’t change our target userbase.”
“It doesn’t create a new problem, it creates a new solution for the same problem, including accessing the full functionality of a system. As it relates to gaming, I want to do something that I was told was impossible.”
What may set Saakuru’s agentic framework apart from many others is that, by design, anyone should be able to use it, including businesspeople who notoriously lack coding proficiency. This, Jack said, should increase accessibility, as anyone can describe to the framework what they want their agent to do in plain language, and it will be built.
Access and the AI Swarm
Currently, deploying an AI agent requires a developer to exert some, albeit a minimal amount of, manual labor on the part of a developer. For example, ElizaOS on the Web3 side and N8N make deployments easier by having set up much of the infrastructure developers need to get their agents live, but Jack aims to replace them because they are “quite limited.”
He also expects his framework to solve accessibility by reducing the required manual labor of building an agent down to as close to zero as possible. “First, we wanted to solve one problem,” he said, “AI swarm.”
“What if you have an RPG game with 10K agents that all have interactions with each other? Our framework is trying to solve efficient deployment of thousands of agents at a time for games inclusive of multiple languages.”
AI swarm refers to the exponential growth of AI agents across a network that interact with each other to exponentially increase the amount of data scraped and structured into a database, essentially improving interactions with each other and their environment. Think about it like the growth of neurons through the brain when a child acquires a new skill.
However, what use does such an agentic framework have if it isn’t powerful enough to accurately interpret the relative scribbles those uninitiated in prompt engineering tend to throw at chatbots? After all, it took 15 minutes for we two technically inclined individuals to figure out how to record a video call – how long might it take a middle manager at a corporation who struggles to order food via app to adequately prompt a chatbot?
Jack’s slated solution to the classic luddite bottleneck is to add humanity to the machine in a manner of speaking.
Paging Dr. Cope
A feature of the Saakuru Agentic Framework under development is Dr. Cope, an agent trained in human psychology that will inform other agents deployed as NPCs in games on how to behave and remember what human players have done in the game to and around them.
Games like Red Dead Redemption 2, as entertaining as they are, remind players just how far from the uncanny valley games with an expansive world still are. If you can rob a shopkeeper at gunpoint, lose a marginal amount of reputation, then return a few days later to repeat the deed as if nothing happened, then you’ve gotten off too easy – even for an outlaw.
Dr. Cope aims to improve upon that trope. “Within that context,” added Jack, “Compare Dr. Cope to Wendy Rhoades from Billions.”
Furthermore, Jack said “Dr. Cope is designed to store real memories” of its interactions with other humans and agents.
“Dr. Cope will allow agents to learn from each other, including recognizing emotional responses by people.”
Not only would such a system need a spectacular amount of storage space to keep its interactions within a game cached and stable, if they are on an individual basis for each user’s instance of the game, that could mean that each user would need massive hard drives to properly play games with Dr. Cope.
On the other hand, Saakuru could potentially host storage itself either in on-premises servers or in the cloud – most likely in the cloud. Considering the human brain roughly has 2.5 petabytes of storage space to house memories and the emotional responses associated with them, I expect Dr. Cope to require a mind-boggling amount of storage, but Jack did not share details on this aspect of its design.
How to Compete?
Currently AI agentic frameworks from ElizaOS and Virtuals have captured the gross mindshare. For Saakuru to compete, Jack and his team face an uphill climb to relevance in a space that has its darlings.
In addressing this, he appears to believe that focusing on building the best product that manages to solve a problem that real people are facing will be the path to success. And if Jack’s sentiments towards the formerly mentioned frameworks are any tell, he believes there is plenty of opportunity for Saakuru to rise.
In a highly competitive Web3 landscape where firms are rapidly finding new ways to integrate AI into their operations, fortune may favor the bold. Base network has benefitted from the $700M+ in total value locked (TVL) on-chain from coins launched by Virtuals, an AI agentic framework and launcher. ElizaOS $ai16z added $2.5B in TVL at its peak to Solana itself.
Despite their early successes, Jack believes he and his team can do better. “Virtuals sucks,” he said, “and Eliza is quite limited.” Usually the best option [when using either of them] is to code it yourself” which many prospective AI builders still cannot do.
If truly no-code options are the key to long-term success, Jack says he is well on that path as his competitors lack such ease-of-use. His goal affirms his impetus for creating the Saakuru Agentic Framework: access.
“The more access given to everyone, including kids, the more innovation we’ll see from every sector. What we see with the examples of device familiarity through Apple is that kids can go and do tremendous things from their kitchen table with tablets in their hands while Mom and Dad are busy.”
By improving access by addressing the AI swarm issue, Jack expects Saakuru will rise to the top of the mindshare at the convergence of AI and Web3. “Put that in the hands of anyone and you’ll see millions of apps popping up every day.”
3 Comments