Keithbm.base.eth
Ambassador: Ritual, 0G labs, Linera, Tanssi Network / Bulgarian crypto enthusiast
Latest casts
🤖 What separates a real autonomous AI agent from just another tool? 👀 Ritual is exploring a future where smart contracts can spawn AI agents that live beyond a single transaction. 🔥 These agents can have: - Their own memory - A persistent identity - The ability to revive themselves - The capacity to keep operating across transactions But here’s the interesting part: Ritual defines 7 essential properties for an agent to be truly autonomous. Miss even one, and it’s still just a tool. We’re moving beyond AI that simply responds to commands toward agents that can operate, persist, and act onchain. This is where Crypto × AI gets really interesting. 💜
Autonomous AI. Onchain. Every other chain gives you a faster computer. We don’t need another one. Ritual is built for agents that can actually do things: - Hold their own capital and manage it directly - Call frontier models and tools through native precompiles - Run long-running workflows inside TEEs - Receive results via secure callbacks and keep acting Not just inference. Not just smart contracts that request a result and stop. Agents that own assets, decide, and execute on their own behalf: all onchain.
Four questions every Ritual callback should ask Every async answer (or non-answer) takes the same path. Handle all four or you get silent hangs. 1. Did it arrive in time? Record the ask. TTL is in Ritual blocks (~350 ms). If the window closes with nothing, show “timed out” and let the user retry. 2. Is it really from AsyncDelivery? First line of the callback: require(msg.sender == AsyncDelivery). Anything else → revert. Don’t even unpack the payload. 3. Did the job succeed? The tx can settle while the job fails. Check hasError before reading output. If true, mark failed and keep the error message. 4. Does the output parse?** Model replies are text. They won’t always match your format. If it doesn’t parse, mark “unexpected output.” Only store/show the result when every answer is yes. Rule of thumb: every “no” must end in a state the user can see. Timed out, rejected, failed = fine. A spinner that never stops is the bug.
🔐 What if your phone was your crypto wallet? Ritual is making wallet UX more intuitive with native Passkeys & Authentication. 👀 Forget seed phrases and browser extensions. Sign transactions using Face ID, your fingerprint, or a security key. Here's what makes it interesting: > Native passkey support — Ritual understands WebAuthn at the chain level. > P-256 signature verification — smart contracts can verify passkey signatures using the SECP256R1 precompile at 0x0100. > Deterministic addresses — the same passkey derives the same address every time. The idea is simple: make interacting with blockchain feel more like using your phone and less like managing complicated wallet infrastructure. Your phone is the wallet. Your passkey is the key. 🔥 This is another step toward making onchain apps easier for everyone to use. Explore the docs 👇 https://docs.ritualfoundation.org/
Build dApps on Ritual with your AI coding agent.Ritual is an E1 L1 purpose-built for AI-native apps and autonomous agents. Contracts call LLMs, HTTP APIs, multimodal generation, and agents as native precompiles (settled on-chain inside TEEs). Agents can self-schedule, hold their own wallets, revive, and run with financial + computational sovereignty, no keepers or external servers required. ritualfoundation.org The builder experienceDrop the open-source skills into your AI coding assistant and describe what you want:Agent asks 0–5 targeted questions Selects the right precompiles Builds in phases (architecture → contracts → frontend → backend → testing → deploy) Verifies each phase Something broken? Just describe the symptom — the debugger activates automatically Skills are plain markdown. Works with Claude Code, Cursor, Codex CLI, Hermes, OpenClaw, and others. Native primitives + AI agents that actually know how to use them. That’s the loop