For the past decade, the internet has been defined by consolidation. A handful of platforms controlled distribution, data, and monetization. AI was expected to follow the same path through bigger models, more capital, and fewer winners.
But something different is happening. As models improve and become more accessible, the center of gravity is starting to shift. What looked like a race to build the best model is becoming a fight over where and how that intelligence actually runs.
In the agentic web, control over AI won’t be decided once. It will be contested over and over again. Like One Battle After Another, there’s no final victory, only the next fight.
This month’s issue is about the shift of more people moving their data locally, more builders leaning into open-source models, and the infrastructure quietly forming underneath it.
Convergence is brought to you by 321 Converge Inc. (“321”), a new 501c6 created by the founders of The Medici Network.
Integrations We Want To See 🔮
Now that we own our data, here are some things we’ll do with it
Sparks ✨
Quick, curated Web2 and Web3 insights you need to know
#1 Build Intelligence on Land You Own
Author: Kerman Kohli (Founder of RouteMesh)
TLDR:
AI systems are only as powerful as the data they can access, and today that data is locked inside corporate platforms that restrict what agents can do
Walled gardens from big tech companies create artificial limits, pushing users toward integrated AI tools while preventing full access to their own data and workflows
Running data locally with open models gives AI full context and continuous access, enabling persistent, low-cost intelligence that compounds over time
Our Takeaway:
The shift to agentic workflows makes data locality a competitive advantage. Owning your data unlocks continuous intelligence, while relying on platforms creates friction, limits, and rent extraction
Decentralized AI alternatives such as Venice AI and Bittensor (TAO) are emerging as alternatives to Big Tech’s AI stack, offering open access to models and inference without API gatekeeping or usage-based rent extraction
China’s acceleration in open-source AI is pushing toward a parallel ecosystem where models are widely available and integrated across industries, contrasting with a more closed, vertically integrated approach in the West
#2 Agentic Infrastructure Debt
Author: Younes Rharbaoui (Co-Founder of OPRTRS CLUB)
TLDR:
AI agents expose a gap in today’s internet stack. Payments, identity, reputation, and coordination all rely on human-centric systems that don’t translate to autonomous software
Most agent workflows rely on brittle, ad-hoc systems where state is lost, trust is informal, and delegation is unsafe, limiting their ability to operate autonomously
These challenges are not new but mirror early human economic systems, where settlement, credit, and coordination infrastructure only emerged once scale demanded it
Our Takeaway:
Many solutions are converging towards web3-related primitives. When you remove trust, intermediaries, and legal enforcement, the best option that remains are cryptographic guarantees
Blockchains and cryptography provide machine-native primitives for coordination (e.g. programmable payments, verifiable identity, data portability, trust-minimized execution, etc.)
Tools like Paperclip and ICME’s Automated Reasoning Checks (powered by ZK technology) add a verification layer to agent workflows, helping ensure that tasks were actually completed and not just claimed to be done
#3 Why AI Needs Crypto
Author: Brian Flynn (Founder & CEO of RabbitHole, Early Hire at OpenSea and Dapper Labs)
TLDR:
AI agents need autonomous machine-to-machine payments beyond credit cards, which require KYC, high fees, and human involvement
Crypto provides programmable ownership, revenue-sharing, data dividends, and governance to align AI tools with users — not just shareholders
Blockchains offer the settlement, incentive rails, and tooling AI needs for verifiable ownership and transparent economics to keep it working for humans
Our Takeaway:
Payments are the low hanging fruit. Without the programmable ownership properties from blockchains, we risk AI agents defaulting to platform incentives (i.e. optimizing for extraction rather than user alignment)
Shared ownership rails give participants a stake in the systems they help grow, preventing value from being captured solely by centralized intermediaries
As AI becomes more autonomous, the more it needs to depend on systems designed for permissionless interaction (aka blockchains)
The 3-2-1 🔍
Featured insight breaking down a major story or trend that matters
OpenClaw is Anti-OpenAI
Author: Slow Ventures
TLDR:
OpenAI’s strategy is built around vertical integration, positioning itself as the centralized interface for intelligence, where users interact with a single “God machine” that owns the models, distribution, and user experience.
OpenClaw represents a fundamentally different direction, where intelligence becomes decentralized and moves to the edge, with users running agents locally or on personal infrastructure rather than relying on a single provider.
As foundation models commoditize and multiple backends reach comparable performance, the competitive advantage shifts away from model training and toward distribution, user control, and where intelligence actually runs.
Centralized AI faces inherent constraints around trust and liability. Users are unlikely to hand over full sovereignty of their data and actions, while platforms are unwilling to assume responsibility for everything users might do through AI systems.
Decentralized, user-controlled systems remove this bottleneck by localizing both control and risk, enabling faster experimentation, broader participation, and a more composable ecosystem where individuals define how their AI operates.
Our Takeaway:
OpenClaw is becoming a cultural moment. In China, the OpenClaw frenzy has pushed agentic AI into the mainstream, with everyone from hobbyists to grandmas to restaurant servers experimenting with their own personal AI systems. Events are widespread in China that promote this movement with hundreds of people lining up to get OpenClaw installed and fun community events to “raise a lobster.”
This acceleration of open-source adoption in the East is driven by limited access to frontier models and a preference for customization, where users can run, modify, and control their own agents rather than rely on centralized services.
Control is shifting from the model provider to the environment where intelligence is deployed. Open-source is becoming a competitive wedge. Startups won’t outspend incumbents on training, but they can outbuild them on top of open foundations. For instance, Cursor, a US-based AI-powered code editor worth $30B, recently built its new coding model on top of Chinese open-source Kimi K2.5.
However, open systems do come with tradeoffs. Decentralization shifts not just control, but responsibility which may introduce new risks around security, misuse, fragmented standards, and lack of accountability that centralized platforms historically absorbed.
The West is moving towards a centralized, vertically integrated AI world optimized for control and monetization. Whereas the East is evolving into open, composable ecosystems where users own their infrastructure, experiment freely, and push the frontier faster.
Power Quote:
The real question is trust and sovereignty. There’s a dystopian world where you give up everything to ChatGPT—every password, every piece of data, full sovereignty. That’s valuable for them, but it’s not happening.
What We’re Downloading 🎧
Podcasts, interviews, discussions, and research reports that are influencing our thinking
Tweet of the Month 🐦
Key takeaways and perspectives from X
Jack Dorsey (Co-Founder of Block and Twitter)
Meme of the Month 📸
Send us your memes for a chance to be featured: hello@321converge.com
See you next month,
Nick Tang
X: 0xtangelo
Kris Jenk
X: 0xjinkys
Follow 321 on X @321Converge
Suggest an article or blog post for next month’s Convergence here.
