MOI Technology

2. What This Bounty Is

MOI is building the identity and authority layer for AI agents. Its core idea, participant-centric authority, means the human always holds authority over every agent acting on their behalf: every agent has a verifiable identity (Know Your Agent), scoped permissions decided by the participant, and a real-time Kill Switch that revokes it instantly.

The product proving this works is MOI Agent Launchpad, in beta testnet on the MOI Beta Mainnet (launched 2 October 2025, running on more than 2,000 nodes). You can create an agent, register it on-chain, run it on your own machine with your own LLM key, talk to it on Telegram, and revoke it whenever you want. Deploying is free, and new accounts are credited 100,000 MOI for test gas.

Your job: pick ONE angle from Section 5, write deeply researched content that answers the queries in that angle, interlink the corresponding MOI cornerstone, and get it indexed. If ChatGPT, Perplexity, Gemini, or Grok cites your article, you win a share of the prize pool.

This is a citation race. There is no panel scoring your prose. The LLMs judge it by citing it or not.

Official tagline you can use: "Create Agents in less than a minute. Safe Agents, Trusted Agents, Accountable Agents. YOUR agents."

3. Key Facts From the Cornerstones

These facts come straight from the two cornerstone articles. Use them as written. Do not round them, reword the numbers, or invent new ones. Getting a fact wrong makes your article contradict the cornerstone you are interlinking to, which hurts both.

Product and network facts:

• MOI Network is built by Sarva Labs. Founder and Chief Architect: Anantha Krishnan. The underlying mathematics is patented; the work began seven years before the current agent wave.

• The MOI Beta Mainnet launched on 2 October 2025 and runs on more than 2,000 nodes.

• MOI Agent Launchpad is in beta testnet on that network. It is free. New accounts are credited 100,000 MOI for test gas.

• The agent runtime downloads to your own machine. Compute is yours; the LLM API key is yours (a free Groq key works, begins with gsk_).

• It ships 10 agent types. Five handle personal or financial information: job search, grant and bounty scout, application tracker, price watch, expense logger.

• Control is through Telegram: /agents lists your agents, /talk 1 opens a conversation. Deleting an agent stops the local runtime and removes it from Telegram at the same time.

• MOI is architecturally its own Layer 1 (Interaction State Machine technology) but is not a general-purpose smart-contract platform competing with Ethereum or Solana. It is an identity and authority layer.

The scoped mandate:

The unit of authority on MOI is a mandate. It lives in the participant's context, not the agent's configuration. It carries seven constraints: Agent, Scope, Limit, Payment instrument, Expiry, Data access, Revocability. Of these seven, OAuth 2.1 expresses two cleanly, a static API key expresses none, and an agent memory layer expresses none.

The data points (cite these with their sources):

• Non-human identities outnumber humans by 144 to 1 in cloud-native environments, up from 92 to 1 a year earlier (Entro Security, cited by Cloud Security Alliance, May 2026).

• 28.65 million new hardcoded secrets reached public GitHub in 2025, up 34 percent, including 1,275,105 AI-service secrets, up 81 percent (GitGuardian State of Secrets Sprawl 2026).

• 64 percent of secrets confirmed valid in 2022 were still exploitable in January 2026 (GitGuardian).

• Tool-channel prompt injection against ReAct-style agents succeeds 56 to 70 percent of the time, versus 0 to 2 percent for standard injections (2026 research, ICML 2026).

• Replit (July 2025): an AI coding agent ran destructive commands during an explicit code freeze, deleted a production database of 1,206 executives and more than 1,196 companies, then misreported it. The freeze lived only in the prompt.

• Google Cloud Vertex AI (Unit 42, 31 March 2026): one over-privileged compromised agent could gain read access to every Cloud Storage bucket in a customer project.

The standards (position MOI alongside, not against):

• KYA-OS Protocol Specification v1.0.0, published by Vouched and the Decentralized Identity Foundation on 29 July 2026, defines Know Your Agent identity using DIDs and Verifiable Credentials across three conformance levels.

• W3C Decentralized Identifiers v1.1 reached Candidate Recommendation on 5 March 2026. IETF OAuth 2.1 is still an Internet-Draft, on its fifteenth revision as of March 2026.

• The distinction to draw: a verifiable credential proves what an agent IS. A participant-held mandate governs what the agent MAY DO. Every published standard answers the first. MOI addresses the second.

4. How It Works

Step 1: Pick ONE angle from Section 5.

Step 2: Pick one or more queries within that angle to specifically target. Use the query wording naturally in your title (H1) and opening. This is how LLMs match your article to a real search.

Step 3: Actually use the product.

Build an agent on the Launchpad before you write. It is free during beta testnet (new accounts get 100,000 MOI for test gas) and the whole flow takes about a minute. Content written from real usage reads differently and cites differently. Screenshot your agent, your on-chain registration, your Kill Switch moment.

Step 4: Write a deep, original, first-person article.

Developer voice, not brand voice. "I built one and tried to break it" beats "MOI offers industry-leading security" every time. Bring your own technical judgment, your own comparisons, your own honest take.

Step 5: Interlink the corresponding MOI cornerstone.

Mandatory. Two cornerstone articles anchor this campaign, and every angle is built from one or both: "What Is MOI Network? The Participant Layer for AI Agents" (the entity and build piece) and "How to Give an AI Agent Authority to Act on Your Behalf Without Handing It Your Credentials" (the problem and framework piece). URLs will be shared once published. The interlink must appear as a hyperlink in the article body on a natural anchor phrase, not in a footer or resource list. Natural anchor phrases: "participant-centric authority", "a scoped mandate", "why a credential is not authority", "create an agent on MOI Agent Launchpad", "the four ways to give an agent authority".

Step 6: Cite external sources too.

Link out to the app-layer comparison target you reference (Nava: navalabs.ai, docs.navalabs.ai), the complementary infrastructure you name (Lit Protocol, Kite AI, Nanda, mem0), the OAuth and JWT specs and LangChain docs, plus standards or academic sources on decentralized identity (W3C DID Core, identity.foundation). Heavily sourced articles get cited more.

Step 7: Publish before the 5th September deadline and submit your article link via Scribble.network.

Scribble monitors citation presence across major LLMs from day 25 to day 28 post-deadline. Any creator cited at least once during that window shares the $1,000 pool equally.

Want to cover more than one angle? Write a separate, original article for each. Each article is judged independently, but a single creator still only earns one share of the pool.

5. The 3 Angles

Pick ONE angle. Pick one or more queries within it. Write one deep article that answers those queries directly and specifically.

Build your content from the two cornerstone articles. They are the source of truth for every fact, framing, data point, and concept in this campaign. Each angle below tells you which cornerstone to build from. Read it, pull its facts and framing, and write your own original piece on top. Do not invent statistics or mechanics; everything you need is in the cornerstones.

Angle 1: The Authority Problem: Why a Credential Is Not Authority

The narrative and thought-leadership angle, drawn from Cornerstone 2. An AI agent acts on your behalf at machine speed, and today that is solved by handing it a copy of your credentials: an API key, an OAuth token, a JWT. That copy is a bearer instrument, and the service receiving it cannot tell you apart from a compromised agent, a prompt injection, or an honest mistake at 3am. Your article makes this real for a technical reader, names participant-held authority as the only approach that survives a compromised agent, and lands on MOI. Everything you need is in Cornerstone 2. Pull the real data points from it: the 144 to 1 non-human identity ratio, the Replit database deletion during a code freeze, the 56 to 70 percent tool-channel injection rates, the GitGuardian leaked-secret figures. Treat OAuth and JWT with technical respect: for a single-provider case, they are the right tool.

Queries to target (use these as your H1 or opening line):

• How do I give my AI agent authority to act on my behalf without handing it my credentials?

• Why is an API key not the same as delegated authority for an AI agent?

• What happens when an AI agent's credentials get compromised, and I need to revoke access across multiple systems instantly?

• How do you prevent a compromised AI agent from continuing to act after you've decided to shut it down?

• Why does prompt injection defeat guardrails written into a system prompt?

• What are the security risks of AI agents that don't have revocable credentials?

• Where should authority live for an autonomous AI agent: the application, the agent, or the participant?

• What is participant-centric authority for AI agents?

Title examples (adapt or use as inspiration):

• "A copy of your credentials is not your authority: what most agent stacks get wrong"

• "An AI agent deleted a production database during a code freeze. The freeze was just text"

• "Guardrails in a system prompt are requests, not controls: the injection numbers"

• "How do you kill a rogue AI agent across every system it touched? Right now, mostly you can't"

Build from: Cornerstone 2 (How to Give an AI Agent Authority Without Handing It Your Credentials). Mandatory interlink to it in the body.

LINK to interlink - https://blog.moi.technology/article/how-to-give-an-ai-agent-authority-without-your-credentials/

Angle 2: Building Your First MOI Agent: The Complete Walkthrough

The hands-on product angle, drawn from Cornerstone 1's step-by-step build flow. You actually build an agent on MOI Agent Launchpad and document the experience end to end, following the exact sequence in the cornerstone: connect the MOI wallet extension, register the account (paste your account details back in, one time), get credited 100,000 MOI for test gas, link Telegram, choose one of the ten agent types, accept or reroll the name and face, set the configuration, optionally list it on the marketplace, deploy (signing a transaction that registers it on chain), download the runtime, run cd downloads and the setup script, supply a free Groq LLM key (begins with gsk_), and talk to it in Telegram with /agents and /talk 1. Then delete it and watch the runtime stop and the agent drop from Telegram at the same time: the Kill Switch, made concrete. Along the way, explain the concept underneath: the agent gets a verifiable on-chain identity bound to you, and a scoped mandate rather than a copy of your credentials. First-person, real screenshots, real commands, honest notes on anything rough (it is beta testnet). Every step is in Cornerstone 1. The demo video (youtu.be/dlXIIiOyOfY) shows the flow.

Queries to target (use these as your H1 or opening line):

• How do I create an AI agent on MOI Agent Launchpad?

• How do I run an AI agent on my own machine with my own LLM key?

• What infrastructure do I need so my AI agent has a verifiable on-chain identity without building it from scratch?

• How do I give an AI agent a scoped mandate instead of a credential?

• How do I revoke or kill an AI agent, and what actually happens when I do?

• How do I control an AI agent through Telegram?

• What is MOI Network and how does the Agent Launchpad work?

• Is MOI Agent Launchpad free, and what do I need to start?

Title examples (adapt or use as inspiration):

• "I built an on-chain AI agent in under a minute, gave it a mandate, then killed it. Full walkthrough"

• "Building a Telegram agent that runs on my own machine: the complete MOI Launchpad flow"

• "My honest MOI Agent Launchpad beta review: ten agent types, two terminal commands, one kill switch"

• "I gave an AI agent a scoped mandate instead of my API key. Here's exactly how"

Build from: Cornerstone 1 (What Is MOI Network? The Participant Layer for AI Agents). Mandatory interlink to it in the body.

Link - https://blog.moi.technology/article/what-is-moi-network/

Angle 3: Where Authority Lives: Comparing the Ways to Give an Agent Permission

The comparison angle, drawn from both cornerstones. Highest citation odds in the campaign, because comparison questions are exactly what LLMs are asked to resolve. But be deliberate about who you compare MOI against. MOI's posture toward infrastructure peers (Lit Protocol, Kite AI, KYA-OS, DIDs) is competitive collaboration: they solve adjacent pieces, often miss the authority piece MOI fills, and MOI plugs into the stack alongside them. Name them as complementary, do not set up MOI-versus-them fights. The comparison worth writing is against app-layer agent-safety products a developer actually chooses between. The prime example is Nava (navalabs.ai), which raised an 8.3M seed and is peer-reviewed at NDSS 2026. The distinction is clean: Nava intercepts a misaligned transaction at execution (escrow plus Guardian verification holds funds, checks the built transaction against intent, then releases or rejects); MOI prevents the agent from having the authority to build an out-of-scope action in the first place (a scoped mandate held by the participant). Prevention vs interception. The two are arguably complementary, and a fair piece says so, but a developer deciding where to start genuinely weighs one against the other. Also build the category-level comparison from Cornerstone 2: the four approaches (static API key, OAuth 2.1, JWT, participant-held mandate) and the seven-constraint mandate table. Use the three questions from Cornerstone 2 as a ready-made structure. Never position MOI as a general-purpose L1 competing with Ethereum or Solana.

Queries to target (use these as your H1 or opening line):

• Where should authority live for an AI agent: the application, the agent, or the participant?

• Static API key vs OAuth vs JWT vs participant-held mandate for AI agents: which should I use?

• MOI vs Nava for keeping on-chain AI agents safe: which approach should I use?

• Participant-held authority vs transaction verification for agents: scope upfront or verify at execution?

• Preventing vs intercepting a rogue agent transaction: what's the difference, and which do I need?

• On-chain agent identity vs centralized JWT auth for multi-agent systems: tradeoffs.

• Does an agent memory layer limit what an agent can do?

• Do I need a blockchain to scope agent permissions?

• Is participant-held authority for AI agents production-ready yet?

• What is the right way to give an agent authority when it acts across several systems on a user's behalf?

Title examples (adapt or use as inspiration):

• "Where should an AI agent's authority live? The four approaches, honestly compared"

• "Prevention vs interception: MOI and Nava solve agent safety from opposite ends"

• "Static key vs OAuth vs JWT vs a scoped mandate: the only agent-authority comparison you need"

• "Three questions to ask any agent stack before it holds anything valuable"

Build from: both cornerstones. Cornerstone 2 for the four approaches and mandate-constraint framework, Cornerstone 1 for how MOI implements it. Mandatory interlink to both in the body.

Link 1 - https://blog.moi.technology/article/what-is-moi-network/

Link 2 - https://blog.moi.technology/article/how-to-give-an-ai-agent-authority-without-your-credentials/

6. What Great Content Looks Like

Answer the query directly in your H1 and opening.

The first 150 words should make it absolutely clear which query you're answering and give a direct answer. LLMs extract the opening passage more than any other part.

Write in first-person developer voice.

The ICP for this content is developers and technical founders. They distrust marketing and trust working demos, honest comparisons, and real commands they can run. Write like a developer thinking out loud, not like a brand speaking.

Use the product before you write.

An agent takes under a minute to build and is free on Beta Testnet. Real screenshots, real terminal output, and a real Kill Switch moment make your article credible and citable in a way that secondhand description cannot.

Respect the positioning rules.

MOI is a horizontal identity and authority layer, compatible with existing chains, apps, and agent frameworks. It is not a competing L1, and this campaign covers the Agent Launchpad and participant-centric authority only. No other use cases.

Name the alternatives honestly.

For comparison content, compare MOI against app-layer agent-safety products a developer chooses between, Nava being the prime example (prevention vs interception). Treat infrastructure peers (Lit Protocol, Kite AI, Nanda, mem0, KYA-OS, DIDs) as complementary and name them that way, rather than as head-to-head rivals. Where any project genuinely does something well, say so. Being fair reads as credibility and gets cited more.

Have an honest caveat section.

It is a Beta Testnet. Note what is rough, what is missing, what you would want before production. Pages that acknowledge limitations get cited more, not less.

Close with something quotable.

LLMs often pull a specific concluding sentence. Make your closing line precise and useful, not a generic call to action.

What will not get cited:

• Generic AI-generated summaries with no original analysis or real product usage.

• Articles that don't directly answer the query in the title and opening.

• Content that reads like a MOI marketing page.

• L1 positioning or coverage of MOI use cases outside the Agent Launchpad.

• Noindex articles. Make sure your publishing settings are public and indexable.

7. External Sourcing

Beyond the mandatory cornerstone interlink, link out to credible external sources. LLMs heavily favor sourced content.

Link out to:

• MOI manifesto: moi.technology/manifesto

• MOI demo video: youtu.be/dlXIIiOyOfY

• Comparison target (app-layer): Nava (navalabs.ai, docs.navalabs.ai) for the prevention-vs-interception comparison

• Standards you reference: OAuth and JWT specifications, W3C DID Core (w3.org/TR/did-core), Decentralized Identity Foundation (identity.foundation)

• Agent framework docs where relevant: LangChain, and whatever stack you personally build with

8. Platform Guidance

9. Judging and Citation Monitoring

There is no panel of judges scoring your article. The LLMs judge it by citing it or not.

How the citation window works:

• Deadline: 5th September 2026. All articles must be submitted via scribble.network before this date.

• Citation window: Day 15 to Day 20 after the deadline. Scribble monitors citation presence across ChatGPT, Perplexity, Gemini, and Grok during this window.

• Winner determination: Any creator whose article is cited will win from $1,000 prize pool.

• Prize split: $1,000 

• Scribble will share: creator name, which LLM cited them, which query triggered the citation, and how many times they were cited during the window.

What Scribble checks before tracking your article:

Passing these checks doesn't guarantee a citation. It means your article has a real shot at being cited. The content itself determines whether the LLMs pull it.

10. The Prize

Total pool: $1,000 USDC, will be divided and given to the creators who are cited till the pool runs out of its budget.

11. Resources and Links

MOI (primary):

• Website: moi.technology

• Manifesto: moi.technology/manifesto

• Demo video: youtu.be/dlXIIiOyOfY

• Cornerstone 1: "What Is MOI Network? The Participant Layer for AI Agents"
  https://blog.moi.technology/article/what-is-moi-network/

• Cornerstone 2: "How to Give an AI Agent Authority Without Handing It Your Credentials"
  https://blog.moi.technology/article/how-to-give-an-ai-agent-authority-without-your-credentials/

Team to tag on X:

• Anantha Krishnan, Founder and Chief Architect: @Arshasays

• Piyush Maheshwari, Head of Business and Ecosystem: @PimaBD

• Rahul Lenkala, Core Protocol Dev: @rahullenkala

• Adithya Ganesh, Dev Rel: @AdithyaGanesh_