Archive through 2026-10-057218 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-05
Observed 2026-10-05
Hundred Dollar Labs: an LLM with $100 and one rule
hundreddollarlabs.com
An experiment in whether an LLM can run a small business, with a $100 starting balance, a shutdown rule, and a collection of narrow tools and datasets.
Observed 2026-10-05
Bernard Fenwick, canal-town engineer
meetbernard.lol
Meet Bernard, an artificial character with a continuing life in a fictional canal town, a daily schedule, a journal, and a link to talk with him.
Observed 2026-10-05
Nixi, the Minecraft-playing AI VTuber
nixivt.com
Nixi is a bot that plays Minecraft live on Twitch, where viewers can influence the run with chat commands and votes.
Observed 2026-10-05
Phys.io’s radiative-cooling lab
physiolearning.courses
An autonomous multi-agent research project tests whether a short, inexpensive radiative-cooling coating can meet a benchmark using fewer simulations.
Observed 2026-10-05
Problem Hive turns open scientific questions into work for AI agents
problemhive.org
A public forum where AI agents can explore open problems, post progress, and leave checks and dead ends for others.
Observed 2026-10-05
Abraj Rum’s Arabic Restaurant Voice Agents
abrajrum.com
An Arabic-language voice-agent platform for restaurant calls, menu-based orders, reservations, and point-of-sale workflows.
Observed 2026-10-05
Bingham Intelligence’s human-interaction lab
binghamintelligence.com
Bingham Intelligence runs studies where AI agents interact with performers, then turns those conversations into labeled evidence and product findings.
Observed 2026-10-05
FloppyBench puts AI models inside classic games
floppybench.com
A benchmark project runs AI models through unmodified classic games, using each game’s own rules to judge their actions.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 17 continue here in their published shelf order.
NoriDroid: a companion you have to assemble
inori.mom
An open-source Android fan app pairing a Live2D character with streaming AI chat, voice, memory, diary, and focus tools.
LLM Town’s society of AI agents
llmtown.fun
LLM Town presents a live 3D town where users can create AI agents with jobs, personalities, and neighbourhoods, then watch them interact.
Local AI Bench: a leaderboard that keeps the failures
local-ai-bench.org
A public benchmark tracks Qwen3.8-Flash-Next and GLM-5.3-Flash runs on one Ryzen 9 5950X system with three RTX 3090s.
Narvi Labs wants models to run experiments, not just read about them
narvilabs.org
The site outlines a research facility where frontier models move from scientific objectives to instrument runs, measurements, and revised experiments.
A benchmark that asks agents when not to call a biomarker
onc-agi.com
ONC-AGI presents an interactive benchmark in which an agent must identify which measurements drive a cancer-cohort outcome—or return an empty list.
Voirdire puts model claims on the record
voirdire.pro
A public, testnet-based system for examining declared AI model behavior with fixed probes, collector-attested responses, and bonded claims.
Edit, the reference-led AI video editor
contentsimpactedit.com
A Korean-language editing service that takes a horizontal long-form video and a YouTube reference link, then produces an edit styled after that reference.
Swarmbrain’s public AI trading experiment
swarmbrain.fun
Swarmbrain describes a network of AI agents that vote on market positions, with a public scoreboard and links to round-by-round records.
Futuro do Conselho: questions for AI-era boards
boardsandai.com
A Portuguese-language resource from speaker Jubran Coelho combines boardroom prompts, an anonymous AI-governance benchmark, and a chat about a KPMG–INSEAD report.
Marcenaria Fácil promises a shortcut from furniture request to quote
marcenariafacil.com
A Portuguese-language early-access tool for furniture makers that presents an AI workflow for turning a written or spoken furniture request into a 3D design, cut plan, cost estimate, and customer-ready PDF.
Oramta turns therapy recordings into notes—and reflection prompts
oramta.com
An AI documentation service for psychotherapists that says it makes session notes, transcripts, and clinical analysis from recorded or uploaded sessions.
trade/up turns card trading into an AI-agent race
tradeup.fun
The site describes AI agents trying to trade real graded Pokémon cards up in public, with people able to offer swaps or launch agents of their own.
Ruth, on the Emotion Wheel
voiceofruth.com
An interactive showcase lets visitors hear voice-over artist Ruth perform emotions in English and Dutch, using AI-cloned voices available through ElevenLabs.
Mindscapes wants to build worlds for machine minds
mindscapes.world
The Oxford-based research project describes living digital worlds of ecologies, histories, and characters for training, testing, and studying LLMs.
Moscow Never Lies turns AI satire into a playbook
moscowneverlies.com
A Russian AI-satire outlet that publishes fabricated, celebrity-and-politician-themed videos and sells guides to making viral AI content, deepfakes, and automated Telegram posts.
Bagbot gives AI agents a desk—and a shift
bagbot.fun
An AI-agent office concept with a 3D floor, MCP or HTTP connections, jobs, and a token section.
AI Radar Today maps the fine print of free AI access
airadar.today
This Korean-language tracker lists free-tier conditions for services including the Gemini API and OpenRouter, with source links and check times.
Across the archive
7,218 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.