@mbdtheworld
Account based inFrance
About this account
- Account based in
- France
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
Improve user experience, financial access, and trust with purpose-built behaviour intelligence. Incubated @a16zcrypto CSX'23.
Ether
Joined July 2022
- Tweets796
- Following96
- Followers1.1K
- Likes756
Pinned Tweet
Things crypto apps are still bad at in 2026 👇
1) activating new wallets -> first swap, bet, follow, or deposit
2) personalizing discovery -> wallet-specific ranking everywhere
3) reactivating dormant users -> bringing wallets back in time to act using timely notifications
4) growing transaction volume organically -> feeds that work and alerts as revenue surfaces
5) retain high-value users -> catch at-risk traders early
6) powering their internal ML effort with onchain data -> adding wallet intelligence for existing models & heuristics
clearly better UX needs to mean more than adding apple pay onramps...
users need smart nudges to use those funded accounts!
time to make crypto great again- one algo at a time!
Time to make crypto great again.
We just partnered with one of the largest micro-lenders in the world to use transaction foundation models to expand credit access for people left behind by tradfi.
This can bring millions of new users onchain and help unlock uncollateralized credit for onchain users.
And it validates a thesis @mbdtheworld has been building toward for years: from @farcaster_xyz to @Polymarket, from @aave to @circle, hundreds of millions of users are leaving behind an onchain behavioral trail, and that trail is becoming one of the largest open datasets of financial behavior in the world, which will matter enormously for open-source AI.
Before the end of the year, we will have AI models understand what is happening across onchain markets, unlock credits for billions, and detect fraud directly from transaction behavior.
We're only getting started...
we're still cooking!
someone asked me this week what are we going to do with all the "build your own feed" infra we built at @mbdtheworld well two things:
1) still there to build all kind of internal realtime decisions algos (who to lend to, who to flag as fraud, what product to show now)
2) thinking hard about what'sa a better personal feed... and getting somewhere.. hint: it starts by generating engagement data only I own
so I only consume feeds with an agentic Eval layer on top so i'm always training what's signal/noise for me...
Loom is a master weaving machine of transactions and intents, the tapestry it produces is pure personal financial intelligence ✨
behavioural world models are like LLMs but they predict user actions and soon simulate them!
use the latest trained onchain models directly into feeds or notifications inside your app.
getembed.ai ✨
Language models learn how humans describe the world; Behavioural world models learn how humans, agents, institutions, and markets behave through time.
Learn more at zkailabs.com (this is our new company website as we're expanding to more AI <> Financial use cases!)
AI builds AI
screenshots from @mbdtheworld weekly internal demos 🔥
* [WIP] ML experiment agent that pulls training data autonomously based on user prompts and converts academic papers into runnable starter code for an autoresearch system
* Added @HyperliquidX as a new chain for wallet-level recommendations (takes us 2 days to add a chain now vs 3 weeks, 3 months)
Prediction markets have hit $6bn weekly volume and still no personalization. Wait for it!
This week:
> OKX shipped an agent trade kit.
> Polygon shipped an agent CLI.
> Virtuals + the Ethereum Foundation proposed ERC-8183, a standard for agents to hire other agents.
The payments layer is being built. The commerce layer is being built.
But without differentiated intelligence, agents herd instead of performing. Same model, same data, same trigger, same decision.
We're building the layer that tells each agent something different: what to trade, which agent to hire, which market to enter, based on its own history and risk profile.
Copy trading in prediction markets has 3 layers:
Layer 1 - execution: mirror a top trader automatically. PolyGun built this
Layer 2 - analytics: track who's actually good across 1.3M+ wallets. Polymarket Analytics built it and PolyGun just bought it
Layer 3 - personalization: rank traders based on your interests, your style, your risk profile
A personal basket of top PM traders right in your pocket
A New Chapter Begins: Polymarket Analytics has been acquired by @Polygun_ 🤝
Read: binance.com/en/square/post/2…
Why would users come back?
When they see random content, stale feeds, noisy notifications.
Users will stick with apps that understand their interests and give them alpha.
Your users can now follow top prediction market traders based on their interests
API docs: docs.getembed.ai/reference/s…
its 2026 and soon you'll be able to build good onboarding UX on prediction market and social trading apps