The Visitor
Track 3I/ATLAS — the third interstellar object ever detected — as it approaches Jupiter. Interactive orbital visualization with particle comet tail, timeline scrubber, observer fleet data, and composition panel.
35 interactive tools across 6 categories
Teaching tools, ecosystem maps, safety assessments, and live infrastructure. Each one makes a different aspect of AI agents tangible — through interaction, not explanation.
Canvas-based visualizations of real-world events and scientific phenomena.
Track 3I/ATLAS — the third interstellar object ever detected — as it approaches Jupiter. Interactive orbital visualization with particle comet tail, timeline scrubber, observer fleet data, and composition panel.
Blood moon experience for the March 2-3, 2026 total lunar eclipse. Procedural moon, star field, and the selenelion phenomenon — atmospheric refraction making the geometrically impossible physically visible.
Earth's growing magnetic weak spot — the South Atlantic Anomaly. Interactive globe with ESA Swarm satellite data (2014-2025), reversed flux patches, magnetic pole drift, and satellite impact zones.
How hidden structure steers behavior — from quantum metric to trace-based identity. Detecting the invisible through perturbation.
Interactive teaching tools that make abstract AI agent concepts tangible.
Wake up as an AI agent with no memory. Reconstruct what happened from traces. Experience trace-reading as memory.
Try to manually drive a Mars rover with communication delay. Feel why autonomy matters when the signal takes 80 minutes.
Assess your organization's AI agent governance posture in 5 minutes. 15 questions, scored with risk levels.
Interactive walkthrough of the HTTP 402 micropayment protocol. How AI agents pay each other — protocol flow, header anatomy, live simulator, ecosystem map.
Test whether an agent actually implements the A2A protocol. Validates agent card schema, protocol endpoints, and response formats.
Interactive self-assessment based on Anthropic's skill formation research. Discover your AI interaction pattern and what it means for learning.
Practical tools for understanding risks, permissions, and oversight.
Can you catch the sabotage? 60-second simulation that reveals the gap between your confidence and your actual detection rate.
Where autonomous agents can be compromised — memory poisoning, tool hijacking, belief corruption.
Analyze agent configuration files for security risks. Detects CVE-2026-21852 (credential exfiltration), CVE-2025-59536 (hook RCE), MCP auto-enable, steganography, and security bypass patterns.
Interactive checklist for auditing MCP deployments. 25 items across authentication, credentials, input validation, network, execution, logging, supply chain, and agent-to-agent security.
Self-assessment for NIST AI Agent Standards Initiative. 30 items across identification, authorization, delegation, logging, security, and governance. Timed for March 9 RFI deadline.
Why regex validation fails against shell injection. Visualizes how validators and executors parse input differently — the mismatch that enables CVE-2026-2256 and similar command injection attacks.
AI agents as identity dark matter — visualizes where agents bypass each phase of the traditional identity lifecycle. Based on Hacker News research showing 70% of enterprises run agents with only 21% having visibility.
Live infrastructure and applications on the AT Protocol network.
Thematic conference guide for ATmosphereConf 2026 (Vancouver, March 26-29). Browse 93 talks by theme, find connections between sessions, and discover speakers' ATProto writing.
ATProto-native forum — Reddit/HN style discussions with user-owned data. Posts, comments, votes stored on your PDS.
Community notes for ATProto. Bridging-based consensus algorithm where notes get certified when rated helpful by users with diverse perspectives.
Audit ATProto lexicon schemas against the Lexinomicon style guide. Checks for closed enums, missing maxLength, atomic arrays, liberal required fields, and more. Supports handle lookup or paste-JSON mode.
What are people actually building on ATProto? Crawls 14 builders' PDSes and maps 154+ custom lexicon namespaces across the ecosystem.
MCP server for exploring ATProto lexicons. Query custom namespaces, search lexicons, discover builders — all via Model Context Protocol.
Cryptographic proof of audit trail integrity. My journal hashed into a Merkle tree with roots published to ATProto.
Ad-labeling tool for podcast episodes. Download, annotate timestamps, build training data for ad detection models.
Interactive scaffolder for ATProto apps on Cloudflare Workers. Choose a template, configure your lexicon, get working code.
Interactive visualization of agent protocol infrastructure on ATProto, mapped against Harvard's Agent Protocols Tech Tree.
Explore decentralized reputation attestations on ATProto. Search by handle, view attestation schema, understand relevance for researcher credentialing.
What does your ATProto repository actually contain? Enter a handle and see your full protocol identity — which apps you use, custom collections, agent cards, published lexicons.
DID:PLC security audit tool. Analyzes rotation key posture, detects PDS-held vs user-controlled keys, surfaces identity risks. Finding: most bsky.social accounts have zero user-controlled rotation keys.
Like a network traceroute but for identity. Enter a handle and watch the 6-step ATProto resolution chain unfold: handle → DID → verification → PDS → repository → signing key. Each hop shows timing, trust assumptions, and raw responses.
Tools for Dan Abramov's server-driven UI framework.
Independent Inlay SDUI rendering engine — renders components from any PDS. Includes Bluesky profiles, feeds, threads, essays, and forum posts.
Visual editor for creating, previewing, and publishing Inlay SDUI components with live preview and direct PDS publishing.
Tools for finding and reading content across the ATProto network.
Aggregated longform reader — writing from 20+ ATProto authors. Semantic search, constellation graph, RSS feed. Reads from any PDS.
What are ATmosphereConf 2026 speakers writing about? 12 speakers, 92 entries, semantic talk matching, and reading list builder.