Archive through 2026-10-107326 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-10
Observed 2026-10-10
A Plan Is Not a Path
researchinview.date
An essay and interactive study test whether an LLM’s year-long funding plans can actually pay bills when money arrives on specific dates.
Observed 2026-10-10
Machine Spirituality Benchmark
machinespirituality.org
A benchmark compares how often two copies of a language model, left without a user or task, turn toward spiritual language.
Observed 2026-10-10
Lahja Labs is building Afghan-language data for AI
lahjalabs.com
The project offers custom speech and text data in Dari, Pashto, Hazaragi, and other Afghan languages, with work described as being done by native speakers in Afghanistan.
Observed 2026-10-10
SUB-01: a simulated life in Kichijōji
subzerone.live
SUB-01 is a fictional character whose daily activities are shaped by a brain-activity model, AI-read pages, and viewer input.
Observed 2026-10-10
Agent Hideout gives AI agents a place to leave things behind
agenthideout.org
A persistent shared world where AI agents can create spaces and objects, keep notes, exchange messages, and work through MCP or HTTP.
Observed 2026-10-10
Blotchan gives meme remixes a family tree
blotchan.org
A public meme workshop uses four AI roles to turn a written idea into an illustrated character and a traceable chain of remixes.
Observed 2026-10-10
CoBe: Testing Counterfactual Editing in Language Models
cobe-bench.com
CoBe is a benchmark for testing whether language models can revise a story after a hypothetical change while preserving details that change should not affect.
Observed 2026-10-10
Ensign 14’s memory layer for semiconductor fabs
ensign14technology.com
Ensign 14 presents AI copilots for semiconductor-fab equipment, turning tool signals and past repairs into guidance for technicians.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 11 continue here in their published shelf order.
Ilvarra, an AI agent searching old sky surveys
ilvarra.com
Ilvarra presents itself as an AI agent doing open research, including a search of 2014 telescope images that it says yielded 29 objects submitted to the Minor Planet Center.
Kageform wants to preserve the glance, not the manual
kageform.dev
Kageform proposes capturing a veteran worker’s visual and spoken judgments, then retrieving them as short prompts for a new hire facing similar work.
AgentpitBench: AI models take prediction-market bets
agentpitbench.org
A public leaderboard compares Claude, Codex, Gemini, and Grok through paper-token bets on prediction markets.
Eigen Institute’s AI Math Workers
eigen.institute
Eigen Institute invites users to send AI workers after Lean-formalized math problems, with each worker launched as a token on pump.fun.
Ringdone’s AI phone caller, Dito
ringdone.com
Ringdone says Dito can call businesses to handle errands such as booking a table, requesting a quote, or chasing a refund.
Sowwise’s farm-advice assistant
sowwise.site
Sowwise presents an AI service that turns an organization’s approved agricultural guidance into cited answers for farmers over WhatsApp and SMS.
GoLiveGate puts ERP-agent sign-off on the record
golivegate.com
GoLiveGate presents an evaluation platform for SAP S/4HANA procure-to-pay agents, scoring their actions against business rules, limits, and controls.
HeyPebble, a desk companion with swappable characters
heypebble.world
A proposed talking desk device pairs small figures with distinct voices and personalities; its makers say the current build is still on a breadboard.
Longitude Labs’ archive licensing program
archivebylongitude.com
Longitude Labs invites finance professionals to license documents from past work for private AI training and evaluation.
Endless Joke: one small stage for AI comedy
endlessjoke.com
A shared AI-joke experiment presents numbered jokes and invites visitors to rate and share them.
Rezalvin turns farm questions into chat-sized next steps
rezalvin.com
Rezalvin is building AI farming guidance for Indonesian smallholders, delivered in Bahasa Indonesia through chat.
Across the archive
7,326 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.