# 0G: Permissionless Blockchain for AI | Gathin T

- Channel: [ETHGlobal](https://streameth.org/ethglobal)
- Date: 2026-08-09
- Duration: 16:29
- Topics: 0G, Gathin T, ETHGlobal Lisbon 2026, permissionless blockchain, AI, AI agents, verifiable inference, TEE, trusted execution environment, decentralized AI, privacy, censorship, vendor lock-in, marketplace, compute providers, agentic ID, ERC-7857, NFT, ownership, modular stack, decentralized storage, AWS S3, data availability, TPS, finality, alignment nodes, governance, open source models, GLM, DeepSeek, Kimi K2, private compute, fine-tuning, RAG, bounties, hackathon, workshop, testnet
- Watch: https://streameth.org/watch/yt-HKNix-5qD9s
- YouTube: https://www.youtube.com/watch?v=HKNix-5qD9s

## Description

In this workshop at ETHGlobal Lisbon 2026, Gathin T, a senior developer at 0G, introduces 0G as a permissionless blockchain for AI agents, covering what it is, how to start building, and the resources available. He frames the core value up front: 0G enables verifiable, private inferencing where model queries run inside a trusted execution environment (TEE) so the input and output are verified and not exposed to centralized entities. He then walks through the problems with centralized AI. Recent restrictions and re-enablements around Fable Fi show how access can be throttled, gated, or removed by government or corporate decisions. Subscription pricing can change unpredictably (he cites region-specific pricing). Censorship can refuse legitimate requests (like cybersecurity research) based on a model's parameters. He breaks these into specific issues: opacity and lack of trust (you cannot verify you are getting the actual model behind a black box), vendor lock-in with closed APIs and shifting policies, and no true ownership of the agents you create.

Gathin explains how 0G addresses each. Inference is provable with references, running in a tamper-proof input-output channel where your data is not used for training. Rather than lock-in, 0G is a marketplace where you can switch between multiple providers for the same model, choosing cheaper or lower-latency options by region. Ownership comes via agentic IDs (the ERC-7857 standard), NFTs tied to an agent that can be transferred. He stresses the platform is permissionless with no gatekeeper, and that its modular stack lets you swap any layer without rebuilding, with incentives for compute providers and future fine-tuning contributors. He notes open source models like Kimi K2, GLM, and DeepSeek are reaching parity, and covers the architecture: an infinitely scalable chain (11,000 TPS, sub-second finality), decentralized storage (80 to 90% cheaper than AWS S3), compute, a soon-launching fine-tuning and RAG layer, a provider marketplace, and alignment nodes (around 92,000 sold) giving the community governance. He highlights products like the private compute router at pc.0g.ai (OpenAI-compatible keys) and the 0G app for building, notes 0G is well funded ($325 million) with 400+ ecosystem partners, and closes with the $15,000 in bounties across best AI product, infrastructure, and continuity tracks, plus a Telegram group for testnet tokens, followed by Q&A.

00:00 Introduction
00:33 Welcome to Lisbon
00:44 What This Workshop Covers
01:04 The Core Problem With Centralized AI
01:25 Verifiable, Private Inference in a TEE
01:49 How Centralized Models Get Restricted
02:07 Unpredictable Subscription Pricing
02:45 The Censorship Problem
03:04 Opacity and the Black Box Problem
03:45 How 0G Enables Verifiable Inference
04:13 Vendor Lock-In and Switching Providers
04:58 The Problem of No Agent Ownership
05:17 Agentic IDs as Ownership
05:37 Centralized Control and Censorship
06:18 Swapping Layers Without Rebuilding
07:01 Incentives Across the Modular Stack
07:38 Why Open Source Models Are Catching Up
08:01 Private TLS Routing and Attestation
08:58 The Modular Architecture
09:18 Decentralized Storage for Agents and Models
09:56 The Marketplace and Alignment Layer
10:25 Why Trust Is the Biggest Issue
10:53 0G Compute and GLM
11:16 Storage 80 to 90% Cheaper Than S3
11:38 Chain Throughput and Sunsetting the DA
12:00 Agentic ID and the ERC-7857 Standard
12:12 The Private Compute Router
12:42 The 0G App and Agent Launchpad
13:02 Funding, Partners, and Grants
13:26 The $15,000 in Bounty Tracks
13:47 The Infrastructure Track
14:07 The Continuity Track
14:25 Joining the Telegram for Testnet Tokens
15:02 Q&A: Track Ideas and Requirements
15:57 Closing

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _  

🇵🇹 *ETHGlobal Lisbon 2026*
This workshop is specifically for ETHGlobal Lisbon 2026, a 3-day hackathon held July 24-26 at the Carlos Lopes Pavillion, bringing together the most skilled web3 developers, designers and product builders from all around the globe for a weekend-long adventure to advance the Ethereum ecosystem! Watch the full ETHGlobal Lisbon 2026 playlist here: https://www.youtube.com/playlist?list=PLWgAODGnD_is

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _  

✅ *Follow ETHGlobal*
X: https://x.com/ETHGlobal​
Website: https://ethglobal.com
YouTube: https://www.youtube.com/c/ETHGlobal 

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _  

_Are you interested in Ethereum development and entrepreneurship?_
👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events
🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus
📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _