The latest and greatest from our team

This report investigates three of the most promising avenues for closing the gap on eight open questions we identified around foundation models beyond text for finance.

This a survery of Foundation Modal landscape as of April 2026 across five interconnected domains: (1) financial language models, (2) financial time-series models, (3) banking and transaction event-sequence models, (4) tabular foundation models, and (5) healthcare and behavioral event-sequence models.

Build a personalized Polymarket prediction markets feed by copy-pasting 5 prompts into an AI coding assistant. Covers search, ML ranking, infinite scroll, and live price charts using the ❜embed Studio SDK.

With 62% weekly churn and just 2-3% D30 retention, prediction markets face a retention crisis. ❜embed's Prediction Market Recommendations API brings personalized discovery to help platforms keep users engaged, bridging the 4x retention gap with traditional fintech.

From 50 algorithms to 2,000+. From 200 developers to 650+. How embed built enterprise-scale personalization infrastructure that's now powering the Base app and ready for 2026.

Linear Boost is a new scoring mode in the embed console that lets you design exactly how your feed ranks content. By choosing the signals you trust like interests, social proof, and content type and assigning simple weights, you get a transparent, customizable 0–1 score for every item. This gives feed builders fine‑grained control over what surfaces first in their user's feeds, and lays the foundation for more advanced ML‑powered ranking tools to come.

This post is our last in our series for now. It explores how Mutual Information–based mechanisms can incentivize truthful reporting in decentralized moderation systems. We walk through the core ideas, assumptions, and failure modes that shape the design of next-generation peer-prediction tools.

❜embed users can now control feed freshness, quality, and fairness by ordering candidate pools at the candidate generation step, shaping which items are pre-sorted before final ranking. This new feature allows feeds to be ordered by criteria such as chronological, popular, trending, or random, giving creators more flexibility to balance discovery, engagement, and content quality.

Discover how “Zora filters” help you surface better coins from the Zora ecosystem—fast. This guide shows which fields to combine (price, volume, market cap, momentum, holders), starter thresholds for liquidity and quality, and guardrails to avoid over‑filtering. You’ll also get success metrics for freshness, diversity, and conversion, plus a live example via Superfeeds.ai’s “Zora Coins on Farcaster” feed.

Measure feed health with DAU, impressions, post engagement %, user engagement %, unique authors, and unique posts. Learn how to interpret these metrics together to diagnose issues and iterate confidently.

This post explores peer prediction — mechanisms that incentivize truth-telling without external verification. It defines key properties for decentralized moderation and reviews leading models, concluding that the Mutual Information Paradigm offers the most promising balance between truthfulness, simplicity, and openness.

Prediction markets are still gaining momentum and onboarding a new cohorts of users. Whether these users are traders or simply consuming predictions as content, prediction market apps face a number of challenges like thin liquidity and poor discovery. These can can be solved with recommendation algorithms. Our new prediction market models help apps show a broader set of more relevant markets to their users.

This post introduces the Bayesian Truth Serum (BTS), a mechanism that rewards honest reporting in surveys by linking payouts to how “surprisingly common” an answer is. We explain how it works, explore its applications across fields like marketing and machine learning, and examine why it struggles with collusion and scalability—key challenges when applying it to decentralized moderation.

In this post we break down how mastering feed mixing gives you full control over the experience that you want to create for your users. By coupling multiple feeds together you can create an exciting, diverse and balanced experience.

Prediction markets are currently onbaording a new wave of users. Platforms are facing discovery as a major bottleneck. In this post we explore how personalized discovery can take prediction markets to the next level.

This post breaks down our approach to fighting spam in open protocols. It exposes the limits of legacy batch analysis, and introduces a new AI-powered approach that scores content and users instantly—making feeds more authentic, flexible, and resilient for both builders and communities.

This third post in the “Quest for Social Sovereignty” series explores prediction markets as mechanisms for truth discovery and decentralized moderation. It examines how markets can aggregate beliefs, extend epistemic power (“omniscience”) to moderators, and inspire novel dispute-resolution schemes like Hanson’s Double-or-Nothing lawsuits and “moderation poker.” The piece concludes by pointing toward oracle-free systems for truth elicitation.

This is the #2 in our series "The quest for social sovereignty". It explores social epistemology—the study of how communities generate and judge knowledge—through the lens of Birdwatch/Community Notes, its strengths, vulnerabilities, and possible improvements for truth-seeking in social media.

Today’s social media platforms strip users of sovereignty over their content, data, and identity. As decentralized social networks emerge as a credible alternative, they offer new freedoms—but also face challenges like moderation and spam. In this blog series, we at ❜embed explore how to make decentralized social usable, transparent, and incentive-aligned, starting with recommender systems and moving toward credibly neutral moderation. Join us as we unpack the future of social algorithms and how we can build them in service of users, not platforms.

Our ranking algorithms just got an upgrade and understand image and video content in depth. As a result you'll obtain smarter and better recommendations.

Learn how to quickly build personalized Zora media feeds with ❜embed’s new feed template and SDK. Our developer guide shows you how to leverage powerful tools that put your app at the forefront of onchain content, enabling your users to discover Zora media effortlessly.

We’re partnering with Base to power personalized onchain social feeds directly in The Base App! Together, we’re redefining content discovery with state-of-the-art personalization, media-rich experiences, and innovative onchain monetization. Learn how ❜embed and Base are shaping the future of social algorithms—and how you can build with us.

Kick-start your Farcaster build: explore six miniapp ideas—gaming to dating—plus monetization tips and a step-by-step guide to launch feeds with Embed’s API.

Discover how Ben Baessler leveraged Embed’s Feed Builder to rapidly create Dash, a personalized, video-focused mini app for Farcaster. This guide details Ben’s exact feed configuration—candidate generation, ranking, visibility filtering, and feed construction—showcasing how quickly and effectively developers can launch tailored experiences using Embed.

Learn how to build powerful, personalized feeds using ❜embed’s no-code Feed Builder. This guide breaks down the recommendation system behind the scenes—from candidate generation to ranking, filtering, and feed construction—and shows you how to customize a “For You” feed using Farcaster data. Whether you’re building a client, a mini-app, or a content layer, this article will help you get started quickly with smart defaults and flexible filters.

The Base team just launched the new Coinbase Wallet app in limited beta, introducing a built-in social feed powered by ❜embed.

New look, new website, same mission. We’ve rebranded from ❜MBD to ❜embed: because your Web3 fingerprint deserves a name that sticks.

Zo.me is building a marketplace for ai agents that feel personal, contextual, and action-ready. In this post, we explore how their farcaster discovery agent uses the ’mbd semantic search api to deliver tailored insights from onchain social data—showcasing the future of agentic interfaces powered by personalization and retrieval.

The wallet retention opportunity: While crypto focuses on "onboarding the next billion users," wallets face a critical challenge, 5-10% of crypto buyers actively participate in onchain activities. The solution? Social graph integration. By leveraging decentralized networks like Farcaster and Zora, wallets can offer personalized token swaps, NFT recommendations, and content based on users' social connections. Applications like Interface and Warpcast have already proven this approach drives explosive growth and engagement, positioning wallets as potential leaders in social commerce, a $350 billion market pioneered by Douyin and WeChat.

0xRec: Empowering Web3 startups with AI to challenge Web2 giants. This acceleration program offers selected crypto consumer applications up to $100K in credits to leverage 'mbd's recommendation models—delivering personalization that outperforms traditional approaches by 3X. Participants receive dedicated AI strategy support from experts with backgrounds at AWS, X, and Chainlink, plus enterprise features, community growth incentives, and marketing support. Applications close December 30th for teams ready to bring best-in-class personalized experiences to mainstream audiences.

'mbd secures $3M pre-seed funding to power Web3 social intelligence. The strategic round, co-led by Mask Network and Polymorphic Capital with participation from a16z crypto CSX and others, will accelerate development of their AI recommendation protocol that delivers personalized, spam-free content for decentralized social platforms. With 100k+ monthly API requests and models trained on Farcaster, Lens, and other blockchain data, 'mbd helps developers overcome the content discovery challenge while matching Web2 giants in recommendation quality—boosting relevancy 3x and content coverage 10x compared to popular feeds.

Inside your social feeds: AI algorithms decide what you see by analyzing your behavior and demographic data. While these models excel at predicting engagement (guessing what you'll click with 90% accuracy after just 5-10 interactions), they've evolved beyond simple recommendations to maximize platform profits—often promoting viral, polarizing content over meaningful connections. Web3 social platforms offer a solution through transparency, user control, and bridging algorithms that foster understanding across diverse perspectives.

The 'mbd manifesto: Accelerating AI algorithmic transparency and choice by building personalized recommendations that return value to data owners. Starting with Web3 social feeds, this platform enables developers to implement cutting-edge content personalization in minutes while ensuring users maintain control through natural language customization, clear feedback mechanisms, and fair rewards for data contributions—sowing seeds for a community-owned AI revolution.

Step aboard the Web3 Guardian's Arch, a historic Thames houseboat where AI and Web3 innovators gathered to reshape social media's future through decentralized technologies during ETHGlobal London 2024. Though concluded, the community continues building a more equitable digital world.

Decentralized data's explosive growth: Web3 storage platforms now outpace major social networks, hosting 1.86EB on Filecoin and 155TB on Arweave. This expanding ecosystem—already larger than Wikipedia—creates unique opportunities for AI advancement in content personalization, moderation, and generation while avoiding Web2's algorithm pitfalls.