How Peter Thiel-backed unicorn Sentient AI plans to compete ChatGPT, Perplexity to build open source AI

Himanshu Tyagi, Co-founder, Sentient AI

Just when it seemed like the AI race had already been claimed by the usual giants — OpenAI's ChatGPT, Google’s Gemini, Microsoft’s Copilot and Perplexity, which among others counts Jeff Bezos of Amazon and tech company NVIDIA as investors — a new challenger is stepping onto the field with a very different playbook. Sentient AI, the Peter Thiel–backed upstart, is not here to play nice. Its mission? To break AI out of the hands of a few and bring decentralisation to the forefront.

Started in 2024, Sentient Foundation has been working on several areas towards its goal of preventing AI being controlled by a handful of private entities, including research, building a blockchain protocol, decentralised artificial general intelligence (AGI) and AI chatbots.

The San Francisco-based startup is now gearing up to launch its initial offerings, one of which is an open-source AI platform—Global Research and Intelligence Directory, which abbreviates to the catchy Grid—that integrates multiple AI models from other developers.

Launched earlier this month, Grid is an open-source AGI platform allowing any developer to contribute agents, data or models. It also lets composable workflows by combining multiple agents, models and tools into functional applications. The developers can earn token-based rewards, with usage fees and subscriptions as optional monetisation routes.

As of now, Grid has over 40 specialised agents, 50 data sources and 10-plus models, spanning Web2 and Web3 applications, the startup shared. It’s like Amazon, but for intelligence. “Anyone building a smart agent, whether it writes code, plans your day or teaches a kid, should be able to offer it up, and get paid when it’s used,” the startup explained.

Speaking to Moneycontrol, Himanshu Tyagi, co-founder, Sentient, said, “We want to make sure that people who build in open source can monetise what they have built by finding distribution and then getting back rewards when the things get used. We also ensure that the things in open source remain safe. (This entails) open auditing all kinds of alignment, ensuring the models/ tools on the platform have proper governance that is not controlled by a single player and they remain decentralised in their governance.”

“So toward these three goals, we are building a blockchain protocol and we are doing research to fill the gaps wherever they are, especially in safety and alignment,” he added.

Another innovation the startup is currently building in private beta is Sentient Chat, which is expected to be live in another two weeks and has around 15,000 daily active users in beta phase.

In an earlier interview in June, Sandeep Nailwal, co-founder of Sentient and founder of blockchain tech firm Polygon Labs, explained that the startup was building a planner and coordinator model (Sentient Chat), which sees addresses user queries on the platform by suggesting the right model to answer that query.

“Somebody could have a finance expert AI model, or a health expert AI, or a writing expert and all that. Sentient’s planner and coordinator model will break that query into different chunks and then give you one combined answer after getting responses from various models,” he said during that interaction.

The startup wants to differentiate itself from Perplexity and ChatGPT also through adding many more agents and data sources to coordinate on its platform. One such use case Sentient wants to focus on is crypto data indexing.

“Crypto is one such place where people are crazy about indexing any new kind of data. Now, not all such data sources can be integrated within Perplexity or ChatGPT, which are closed source. But here there's an open protocol where anybody can come and integrate this. This is the differentiator we bring in through deep research and journeys which involve many agents and many data sources,” Tyagi said.

For instance, he said, both Perplexity and ChatGPT’s crypto data is coming from digital assets aggregator CoinGecko, but Sentient will be able to integrate data from hundreds of other application programming interfaces or APIs and token terminals such as crypto exchange Arkham Exchange.

Sentient was co-founded by IISc professor Tyagi, Nailwal and Princeton professor Pramod Viswanath, along with Sensys (an open source AI venture development company).

Tyagi explained that he looks after the product and technology, Nailwal is responsible for the business and strategy side, while Vishwanath, who had also been his postdoctoral mentor, will be spearheading research.

Currently valued at $1.2 billion, Sentient had raised $85 million in a seed round last year from Thiel’s Founders Fund, Pantera Capital and Framework Ventures.

Thiel, an early investor in Facebook, is also the co-founder of PayPal and data analytics firm Palantir Technologies.

Sentient at present has a global team of 30, predominantly in the US, across AI research, product and marketing.

AGI is still seen as a hypothetical form of AI that has human-like cognitive abilities, such as the capacity to understand, learn and perform any intellectual task that a human being can do. But that is still years away.

Tyagi defined Sentient’s AGI as one for the internet or AGI for digital spaces.

“I can very precisely define AGI of the digital space. Think of all the digital experiences that you have in a day across your phone, laptop, home devices. Roughly, you either use your hands, voice, camera gestures to command it to do something. Today this journey is very fragmented…,” he said.

Tyagi explained that in AGI of the internet, this entire digital experience becomes a single interface. It can perform tasks across devices and types of commands, whether voice or text, as they will all get translated to the same embeddings.

“This AGI is now very tangible. It's a lot of hard work. But I think this will be on the horizon over the next two to three years,” Tyagi said.