GM!
The Apus Network Twin app is causing a stir on the timeline, with the permaweb community sharing hilarious snippets from their chats with an AI version of “President Trump”. Today, we’ll take a closer look at the tech behind this app and the potential it promises for onchain AI.
Let's dive straight in.
Apus Network unleashes onchain twins ♊
Apus Network recently rolled out Twin, an onchain app that lets you engage in deep conversations with AI versions of figures like Donald Trump, George Orwell, Barack Obama, and Ayn Rand. The permaweb community has been having lots of fun sharing screenshots of their chats with these iconic politicians and philosophers. Behind the playful chat interface, however, lie some serious technical advantages for the AI era.
Tamper-proof and verified
As a fully onchain application, Twin leverages AO for deterministic computation and Arweave for permanent storage, both significant attributes in their own respect. While permanent storage ensures characters load correctly and don’t vanish or change over time, deterministic compute keeps character behavior consistent, ensuring that the same input data always results in the same output.
Inference occurs within Trusted Execution Environments (TEE) on Apus Network for private computations. As a nice touch for the website user, you can see the security status of your connection when you’re engaged in a chat.
While the app’s utility may seem limited to surface-level chit chat, it’s a big move forward for decentralized AI on the permaweb. As highlighted during AO’s launch, the hyper parallel computer offers a technical way to host LLMs onchain, with full verifiability over their inputs and outputs.
Apus Network builds upon the concept of AI auditability with its Fast Provable Inference Faults (FPIF) system that signs off on computations and penalizes any foul play. Compared to alternative mechanisms such as zero-knowledge proofs and optimistic fraud proofs, FPIF offers an ideal blend of fast verification speeds, high security, low cost efficiency, and high scalability.
Agentic AI on AO
This working example by Apus Network is a meaningful first step toward an agentic onchain future, where our AI assistants help out with existing tasks like finance, supply chains, coding, but also support new use cases that have yet to be invented. As these agentic tools get more powerful and assume more responsibilities, it's critical that we maintain oversight over their thoughts and actions. Verifiable AI inference is a core function in this vision.
Get involved with Apus Network
If you’d like to support Apus Network in bringing more innovative AI applications to the permaweb, here are a few steps you can take:
- Delegate your AO yield to $APUS, currently the leading fair launch token by AO delegations.
- Participate in their Discord community.
- Read their Verifiable AI Inference litepaper to dive deeper into their tech stack.
And of course, don’t forget to try out the app. If you’d like to ask President Trump what he thinks of the recent government shutdown, you can try it for free at twin.ar.io.
ICYMI 👀
- Odysee adds new mobile features
- StarGrid reveals impressive new level
- ARIO founder analyzes gateway stats
- Load’s fully AO-based x402 facilitator
This Week's Community Feature 🎬
Seasoned permaweb developer Merdi Kim shared links to 10 Wayfinder tutorial videos on his YouTube channel, covering several different routing strategies.
This playlist is an excellent starting point for all you builders looking to learn about Wayfinder and level up your app development skills.
Thanks for reading!
If you know anyone else who would find this information and our insights helpful, please share this newsletter with them.
Follow us on X to stay up-to-date. Pro tip: check out @aoComputerClub.
The Longview Team
This is not investment advice. No profit guarantees. If in the U.S., ensure compliance with U.S. laws and seek professional advice.