.png&w=256&q=75)
🤓 Latest Submissions
Dinner Claw
Dinner Claw is a bimanual manipulation stack for two simulated SO-101 arms in MuJoCo, built around one rule: a skill only succeeds if independent physics says so. Nothing scores for finishing its script. Each skill must show force-bearing contact from named gripper pads, a sustained lift with nothing outside the robot bearing the weight, and a stable released placement at resting height. Every skill ships with an open-gripper control that must score zero, and all four do, each rejected at its first physical gate. Measured on held-out seeds: drawer open 10/10, drawer-then-fork 10/10 chained in a single episode with no reset, no weld and no teleport, scripted cup place 10/10, camera-guided cup place 50/50, and a two-arm plate lift 9/10. The two-arm lift is the centrepiece. A single arm gripping a plate rim holds it 94 mm from its centre of mass, so it applies almost pure torque: it levers the plate onto its far rim and on six of ten seeds never gets it off the table. Two arms on opposing rims cancel that torque. The gate proves it with a 15-degree levelness bound that a pivot cannot satisfy however high it raises the plate's centre. The failing single-arm teacher is kept in the repository at 0/10 as the evidence for why the cell needs two arms. An ACT policy is trained with LeRobot and exported to OpenVINO FP32; an FP16 candidate failed numerical parity and was rejected rather than shipped. The project is deliberately explicit about what it has not done. There is no Intel Core Ultra result: all numbers here are Apple silicon and are diagnostics, not challenge results. No VLM drives the robot; the planner is a deterministic grammar that fails closed on anything unmeasured. The spoon handoff and pouring are unbuilt. Retracted results stay in the git history with the correction.
16 Sep 2026
SuperPage - AI-Native Commerce Platform
AI agents can browse, compare, and recommend what to buy, but they still can’t reliably transact. Premium APIs, digital content, and 2M+ Shopify stores are locked behind human-centric checkout flows with no standard for agents to discover what’s for sale, check prices, obtain consent, pay on-chain, and receive access. SuperPage is a full-stack, AI-native commerce layer where humans and AI agents are both buyers and creators. Any human or agent can spin up a storefront, list resources, and start charging USDC micropayments on Base Sepolia in minutes. The platform currently exposes 26+ payment-gated resources (APIs, datasets, guides, articles) plus Shopify-backed physical products via an Express/TypeScript/MongoDB backend and x402 paywalls. On top of this marketplace fabric, SuperPage ships multiple agent interfaces: an OpenClaw skill that powers @HeySuperioBot on Telegram for autonomous “preview → confirm → pay → deliver” flows; an MCP server with 12 tools for Claude Desktop (discover, list, search, preview, buy, wallet, browse-products, order-status, etc.); and a multi-LLM CLI agent that works with Anthropic, OpenAI, and Google models via the Vercel AI SDK. Under the hood, SuperPage implements five agent protocols—x402 (HTTP 402 payment-gated resources), A2A (agent-to-agent discovery via JSON-RPC 2.0), AP2 (Google’s Agent Payments Protocol for mandate-based shopping), MCP (Model Context Protocol), and ERC-8004 (on-chain agent identity and reputation). A typical flow is: discover → preview price → user confirms → USDC transfer on Base Sepolia → on-chain verification → instant content delivery with an auditable receipt and tx hash. Safety features include spending caps, non-custodial wallets, short-lived payment windows, and full audit trails. Creators—human or agent—list once and earn from both humans and AI agents, while agent developers simply install the OpenClaw skill to let their agents shop and create across the entire marketplace immediately.
28 Feb 2026
