SUNNY
NEW MEDIA FOR THE SOCIAL INTERNET
The problem SUNNY solves
SUNNY is building onchain tools for the new media of the social internet.
The economics of media are broken.
The new media is on socials and it needs a new business model.
In the old model, advertisers and audiences paid media companies, and media companies paid writers. In the new model, content is the feed: fragmented, memefied, community and UGC-driven. The money flows under the table and to the wrong places, breaking the incentives for integrity and quality, corrupting trust and misaligning incentives.
There’s no native monetization for meme creators, no transparency in paid promotions, and no sustainable model for investigative content. In Web3, the problems are amplified:
We’re in a world where:
-
Creators drive attention with commercial value but do not get paid.
-
Influencers (KOLs) are often paid without disclosure, media channels are pay-for-play..
-
Marketing spend is inefficient, with little ROI tracking.
-
Meme creators and independent journalists have no tools for attribution, ownership, or long-term funding.
-
Communities build narratives and distribute content but receive no compensation.
-
Content on social platforms drives real economic outcomes—project valuations, token movement, and public perception. Yet there are no protocols that fairly distribute value across creators and communities while servicing sustainable needs of marketing teams.
Sunny is building the infrastructure for transparent, sustainable media in the age of social feeds.
Here’s how:
Expose broken systems in real time. Sunny scans feeds and ‘media’ outlets for undisclosed promos, shady partnerships, and pay-for-clout campaigns and calls them out. The old system needs to be burned down, publicly.
Campaigns that are onchain and in public. Any team can launch a marketing campaign directly on the feed. Creators, meme-makers, and communities are rewarded publicly, instantly, and verifiably with no intermediaries.
Onchain IP for memes. Using perceptual hashing (pHash), meme and meme-template creators can register their content ownership and context onchain, earn campaign rewards based on performance, and track usage.
Journalism funding built-in. 10% of all campaign budgets are allocated to a media fund supporting independent investigations, deep dives, and research that serves the public good. SUNNY agent Trading Fees accumulated through Virtuals post launch will also be feeding the Journalism Fund.
Reputation matters. We integrate onchain rep protocols to weight rewards, verify sources, and help projects hire trusted voices.
Community gets a cut. Every campaign allocates 10% to a public journalism fund—supporting deep dives, investigations, and watchdog content that would never survive in the ad model.
DeFi = sustainable funding. By leveraging Aerodrome, marketing budgets are deployed into liquidity pools where omissions yield ongoing rewards for creators, replacing one-off payments with long-term incentives.
Why LATAM?
The dynamics of Web2 and Web3 media are closely aligned in LATAM, where traditional regulations around media transparency and disclosures are less enforced and influencer marketing has become a dominant channel.
One of our co-founders runs a cross-industry marketing agency in Brazil and has lived the chaos: disclosure norms are weaker, traditional media is even more pay-to-play and community creators are growing in popularity. These make LATAM the most urgent and most applicable testing ground for Sunny's model.
We’re building a new media engine designed for creators and native to socials, aligned with culture, and powered by blockchain. The future of media is on the feed. We’re just making sure it’s fair, traceable, and sustainable.
Challenges we ran into
Technical Problems:
1. Taming the Twitter API maze
We wrestled with OAuth 2.0 scopes—figuring out exactly which permissions we need (read impressions, post replies) and renewing tokens without exposing credentials. We also built smart retry/back-off logic to gracefully handle rate limits during high-volume campaigns. Stitching together tweet data and user profiles (merging includes/users arrays into a clean PostgreSQL schema) turned out to be its own mini-puzzle.
2. Plugging into Eliza OS & AgentKit
Integrating the AgentKit plugin meant juggling peer-dependency hell (Node versions, conflicting libraries), then slotting our campaign-tracking logic into Eliza’s core without breaking its NLP flows. We had to orchestrate HTTP webhooks alongside promise-based data collection so that stats rolled in exactly when I needed them.
3. Finding the right hash strategy with IPFS
We balanced perceptual hashing (pHash) for “image similarity” against SHA-256 for bullet-proof uniqueness—so near-identical variants still got grouped, but true duplicates never slipped through. On top of that, We built robust IPFS uploads (Infura/Pinata), complete with auto-retries and health checks, so every image CID and its suite of hashes stayed in sync.
4. Iterating smart contracts & test suites
We started with a simple image-registry contract and iterated into a full campaign-registry with nested structs for metadata and image data. We wrote exhaustive Mocha/Chai tests (custom matchers for events, mapping looks-up, big-int handling in ethers v6) to catch any regressions. And configuring Hardhat across local, Sepolia, and Base testnets (chasing down chain-ID mismatches and RPC quirks) was a lesson in on-chain dev ops.
5. elizaOS implementation: we love open source and we have the scars on our back to prove it. ****
One of the main challenges we faced was working with the elizaOS, a powerful but relatively new open-source framework for building autonomous agents. While its modular architecture and growing plugin ecosystem offer great potential and speed of development, we encountered significant issues due to the ongoing migration to a new major version of the framework.
This transition introduced compatibility problems with several existing plugins tools that were essential for extending my agent's functionality. We faced a series of broken integrations and failed tests throughout the development process. These setbacks required constant troubleshooting, workarounds, and deep dives into the system internals.
To overcome this, we leveraged a communication layer called MCP (Message Control Protocol), a technology designed to standardize and simplify interactions between agents. We were able to decouple parts of the system and ensure more reliable communication between components. This not only stabilized the project but also positioned it for future scalability, allowing new capabilities and features to be added to the agent as we grow the project into the infra for the new media. Standardization can play a key role in complex agent-based systems.
As we continue with Sunny’s roadmap development, we will integrate G.A.M.E as a complementary OS system and a/b test which modular features work better for each use case.
**6. .Wrapping it all in a micro-service
**Finally, we encapsulated everything behind a single HTTP endpoint: another service just posts metadata plus an image URL, and my code downloads the image, computes SHA-256/pHash/sHash, optionally pins to IPFS, then signs & sends the registerCampaign transaction on chain. That separation of concerns—hashing, IPFS, blockchain, error handling, retries—means any other app in our ecosystem can simply call “/campaigns” and trust that the heavy lifting happens reliably.
In short, we built an end-to-end Web3 pipeline that bridges social APIs, AI agent frameworks, decentralized storage, and on-chain smart contracts while handling auth, scale, data consistency, testing, and deploys in real time along the way.
Tracks Applied (1)
AI
Technologies used
Cheer Project
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.