⚠️ Editor’s Note: This article was originally published by Asymmetric Jump on Substack. It is republished here with full credit to the author. All rights belong to the original author.
0. Introduction
Hey there,
Today we explore Autoppia, a Subnet on Bittensor powering decentralized AI Workers and web automation agents. It enables developers to build, deploy, and monetize AI tools while users automate workflows privately. This research is built using live data from TAO.app, Taostats, SubnetStats, Backprop.finance and Autoppia’s docs and GitHub.
Hope you enjoy it!1. Quick Overview
• Purpose: Decentralized marketplace for AI Workers and Web Agents
• Launch Date: Feb 6, 20252. TL;DR
- Autoppia is Bittensor’s decentralized marketplace for AI Workers and web agents
- It enables developers to build and publish modular AI agents, and users to customize and deploy them via a no-code/low-code Studio
- Its SDK, agentic standards, and privacy-preserving infra allow full automation of web and business tasks using LLMs, APIs, and inter-agent value flow
3. Product & Features
- Miner Tasks: Develop compliant AI Worker templates using the Autoppia SDK and agentic interfaces (e.g. email summarization, web automation)
- Validator Tasks: Evaluate Workers via the IWA benchmark, scoring based on web task completion, speed, and reliability
- Stack & Infrastructure: Python SDK, YAML config system, LLMs (GPT-4o), vector stores, TEE-secured deployments, Bittensor subnet backend
- Modularity: Agents are framework-agnostic and can be chained together via standardized interfaces (e.g. LLMInterface, EmailInterface)
- Core Functionality: A full-stack decentralized app platform for publishing, configuring, and monetizing AI Workers — accessed via SDK, API, and UI apps like Autoppia Studio, Chat, and Coworker (browser extension)
4. Moats
- TEE-secured deployment: Uses Trusted Execution Environments for hardware-level privacy, making inference secure and censorship-resistant
- Standardized interfaces: Devs must follow SDK standards (e.g.
LLMInterface,EmailInterface), improving composability and ecosystem trust- Agent chaining: Unique Worker-to-Worker architecture enables modular AI apps with economic flow across agents
- IWA benchmark: Verifiable evaluation framework for Web Agents, strengthening validator scoring logic
- Scoped credential mgmt: Secure user config system with firewalls, scoped tokens, and sandboxed runtime
- Token-locked publishing: $PIA staking required to publish Workers—creating a friction barrier against spam and poor-quality agents
- Cross-agent economy: $PIA acts as a settlement layer for inter-agent payments, hard to replicate without shared infrastructure
5. Team
Autoppia is built by a small but deeply technical team based in Spain:
Founders:
- Daryxx – Telecommunications Engineer, focused on infrastructure and system design
- Dr. Riiver – Physicist, brings academic rigor and cross-domain thinking to AI agent architecture
- Core Team:
– 6 additional developers specializing in:
• Full-stack development (Autoppia SDK & backend)
• Frontend & UX (e.g. IWA test suites, Autoppia Studio)
• Machine Learning engineering (agent logic, validation frameworks)The team blends protocol engineering with applied ML and user-focused tools, giving Autoppia both research-grade reliability and developer accessibility.6. Code Quality
• Repo: github.com/autoppia/autoppia_web_agents_subnet
• Last Commit: 3 days ago
• Languages: Python (70%), Shell (30%)
• Stars: 2 | Contributors: 4
• Hardware Needs: Browser-based testing, validators launch instances (CPU), agents may require light GPU depending on model
• Forkability: High — modular, well-structured, MIT license, uses YAML configs + clear interfaces (LLM, Email, WebOps)7. Social Sentiment
- Community sentiment on X for Web Agents – Autoppia is neutral, with posts focusing on V3.0.0 and Automata launch.
- No influencer engagement; activity is mostly self-driven by @AutoppiaAI
- Public alpha shared via Automata trial link on June 5, 2025.
- No FUD or warnings, but low chatter raises adoption concerns. Signal Rating: Weak
Explainer
- γ (Gamma): The reward token (Alpha) earned daily — it’s how the subnet pays out emissions.
- Alpha Distribution: How much of the Alpha token supply is actually in users’ hands (vs locked or owned by devs).
- Root Prop: % of rewards going to the subnet owner — higher means more centralized earnings.
- Gini Score: A measure of fairness — 0 = everyone holds equally, 1 = one wallet holds all.
- z-score Difficulty: How hard it is for a miner to solve tasks; higher = tougher scoring bar.
- Emissions Ratio (γ/TAO): How much TAO you effectively earn per Alpha — shows ROI.
- TEE (Trusted Execution Env): A secure zone in the computer — even devs can’t peek inside.
- Worker: A mini AI bot you can build or deploy — like an agent that does emails or web tasks.
- IWA Benchmark: A test system that gives tasks to agents and scores how well they perform (for Web Agents).
- Scoped Credentials: Login info (like API keys) that are isolated per worker — so nothing leaks.
- Agent Chaining: Linking multiple AI Workers together to do complex workflows, like Lego blocks.
- Asymmetric Score: A custom stat showing how much upside the subnet has vs. its price and activity.
- Alpha Liquidity: How easy it is to buy/sell the subnet token — more liquidity = better trading.
- Subnet Owner: The wallet that created and controls the subnet (can earn part of the emissions).
- Validator: Checks that Workers do their job right and assigns rewards.
- Miner: Builds and runs the Workers (AI agents) to earn rewards.
Disclaimer
This report was AI-assisted and refined by the researcher. It is provided for informational purposes only and does not constitute financial advice. Always DYOR. Researcher may hold or trade tokens discussed.
Be the first to comment