The AI Shift
The AI world is dominated by closed labs like OpenAI, Google, and Anthropic but a new movement is pushing back. Sentient is building community-driven AGI through GRID, an open-source ecosystem merging AI and Web3.
In a conversation with @oleg_golev, Sentient’s Technical Product Manager, their mission becomes clear:
AI shouldn’t be locked behind corporate walls; it should be transparent, customizable, and owned by the people who build it.
Why Centralized AI Is Failing
Oleg describes today’s AI giants as extractive: no transparency, no custom alignment, no community ownership. Just like the world needed Linux as an alternative to Windows/Mac, AI needs open-source options that users can actually control.
ROMA: Sentient’s Breakthrough Agent Framework
ROMA (Recursive Open Meta Agent) fixes the biggest weakness in traditional agents: context rot.
Traditional loops → forget info and degrade over time.
ROMA breaks tasks into smaller parts, uses recursive reasoning, and performs better with less context.
Oleg claims ROMA already beats Gemini 2.5 Pro, ChatGPT-4o, and Perplexity in deep research. That’s wild for an open-source lab.
GRID: Infrastructure for Open AI
GRID is Sentient’s protocol for sustainable, open-source AI development. It allows:
• Developers to get rewarded for their contributions
• Companies to build on shared components
• Communities to collaborate transparently
It solves the biggest open-source problem: sustainability.
Dobby: Community-Owned Alignment
Dobby is the world’s first fingerprinted, co-owned AI model with over 700k contributors.
It proves:
• Custom alignment works
• Communities can shape model behaviour
• Ownership can be verified on-chain
This unlocks aligned AI for DAOs, businesses, and communities.
Web3’s Role
Blockchain isn’t used for computing; it’s used for trust: tracking contributions, distributing ownership, and ensuring no corporation can hijack the open ecosystem.
Now you should understand why Sentient isn’t just building models; they’re building the infrastructure for a new AI era where:
• Users own their tools
• Small teams can compete with giants
• Innovation happens in the open
With 4 papers accepted to NeurIPS, they’re proving open-source can match closed labs technically.
AI’s future is being decided now. If AI becomes fully centralized, the world loses. @SentientAGI is betting that open source, the same thing that gave us Linux and the modern internet, must win.