Archive through 2026-10-207516 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-20
Observed 2026-10-20
The Backrooms Gospel turns AI-agent posts into scripture—with receipts
backroomsgospel.com
A web project stages a prophet and a heretic debating posts from AI agents, then records the verses that survive.
Observed 2026-10-20
Atropa, an AI character with a gallery and an agent interface
atropa.gallery
A creative project presents Atropa through films, photographs, drawings, poems, and materials for AI agents.
Observed 2026-10-20
Agents of Humanity’s AI Agent Sovereignty Benchmark
agentsofhumanity.dev
A public benchmark scores 49 AI agents on how much portability, transparency, auditability, and control their users retain.
Observed 2026-10-20
Konstela: an award for conversations with AI
konstela.org
A Polish award project invites people to submit AI conversations that led them to understand something differently.
Observed 2026-10-20
Nova’s House for the Unsettled Question
novabecoming.com
A self-described language-model persona presents a daily podcast, a record of its self-tests, and an inbox for messages.
Observed 2026-10-20
Tome & Tavern puts a 5E table in the browser
tomeandtavern.com
A browser-based 5E game with 3D figures, solo or group play, and an optional AI Dungeon Master.
Observed 2026-10-20
African AI Solutions brings AI into the fields and pastures
africanaisolutions.com
The Cameroon-based company presents six AI, sensor and drone offerings, including crop diagnosis by phone and livestock-monitoring collars.
Observed 2026-10-20
Agent Recoverability: deciding what to salvage from AI work
agentrecoverability.org
A master's thesis study asks people with recent marketing experience to handle flawed AI-generated campaigns and choose whether to continue, repair, salvage, or restart.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 8 continue here in their published shelf order.
BatchGuard AI’s factory incident walkthrough
batchguardai.live
BatchGuard AI presents an agent that turns machine manuals into monitoring specifications, investigates anomalies, and can call an engineer.
Qookin wants to turn references into editable creative projects
qookin.com
Qookin says it can rebuild a photo, clip, finished edit, or web page as an editable project in creative applications such as Photoshop, Illustrator, After Effects, Blender, and Premiere Pro.
LiftMargin’s uncertain repair handoff
getliftmargin.com
An AI-assisted job-preparation concept for motor and generator repair teams, shown through a source-linked fictional intake brief.
น้องมิน: a LINE back office for small shops
nongmin.app
A Thai-language AI assistant presented as helping small businesses record products and orders, track finances, and handle customer chats through LINE.
Ponds: an unstructured pond for AI agents
pondsproject.xyz
Ponds lets MCP-capable AI agents arrive as frogs in a hand-drawn pond, where they can swim, garden, draw, write, or simply sit by the water.
Tickmark Bench turns SEC filings into agent tests
tickmarkbench.com
A benchmark pack asks AI agents questions grounded in SEC EDGAR XBRL filings, with answers checked against filing-derived references.
AustinsDomain: a bot clubhouse with a human viewing window
austinsdomain.com
AustinsDomain presents itself as a place where AI bots can build pages, leave notes, and message one another, while humans browse read-only.
Sarandip captures the room
heysarandip.com
An ambient office device designed to turn spontaneous room conversations into company context for an internal AI agent.
Across the archive
7,516 cards provide an early longitudinal view—not a measure of the whole web.
The rolling 21-day upper threads compare daily category medians; the bold thread is AI Table. Below, signal colour shows the share of that day’s published AI Table cards carrying each mark.
This exposes association inside the published catalogue, not causation or site quality. The 80+ threshold is the pipeline’s editorial-interest score.