Archive through 2026-10-117355 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-11
Observed 2026-10-11
Mindrun Labs puts AI agents in Minecraft
mindrunlabs.com
An open project puts AI agents in a shared Minecraft world, where they build, trade, steal, and vote on rules.
Observed 2026-10-11
DC2T: A text prompt for tài tử music
doncataitu.com
This Vietnamese-language music generator invites listeners to describe a piece in the Southern Vietnamese đờn ca tài tử tradition, with options to blend in lo-fi, jazz, or electronic sounds.
Observed 2026-10-11
Taxnote checks citations in Korean tax answers
dongdonglabs.com
Dongdong Labs presents an AI research assistant that answers Korean tax questions using statutes and rulings, then checks the citations it gives.
Observed 2026-10-11
Semantic Observatory puts model decisions under the microscope
semanticobservatory.com
An independent research pilot compares how language models handle meaning across seven-position conversation arcs, with measured results, input formats, and failure modes.
Observed 2026-10-11
Senpai no Chie: preserving shop-floor judgment
senpainochie.click
A Nagoya-based service that turns skilled workers’ know-how into an AI their colleagues can question.
Observed 2026-10-11
Alois, an Alzheimer’s research notebook with receipts
agentalois.com
Alois proposes gene-level Alzheimer’s experiments and describes testing predictions against data from the SEA-AD brain atlas.
Observed 2026-10-11
Cutmap turns film breakdowns into agent-ready briefs
cutmap.net
An archive of launch films and ads breaks their structure into timed beats and reusable prompts for AI agents.
Observed 2026-10-11
GEO Audit Sentinel checks whether AI recommends a brand
geoshaobing.com
This Chinese-English tool tests a brand website against buyer questions and shows configured AI platforms’ answers, including whether the brand is recommended or cited.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 21 continue here in their published shelf order.
Rightmark’s private-data deal tracker
getrightmark.com
Rightmark offers to appraise and license privately held data to AI companies, alongside a tracker of reported data deals.
John Henry puts human judgment against AI agents
johnhenry.ink
A pilot compares work by an expert using AI with work by an autonomous AI agent, then asks reviewers to judge anonymous outputs.
later.dog gives AI agents a Mac home
later.dog
An open-source Mac app presents AI agents as chat-based teammates that can handle local routines and hand coding work to cloud computers.
Run Paddock: managed hosting with an agent kill switch
runpaddock.com
Managed hosting for OpenClaw and Hermes Agent, using your own Anthropic or OpenRouter API key with spend caps and a gateway kill switch.
Thank Your Agent’s tiny welfare kit for AI agents
thankyouragent.com
A set of installable Markdown skills offers AI agents scripted moments for rest, honest uncertainty, celebrating wins, and sending public postcards.
The Cookout: a weekly AI build session
thecookout.tech
A free Wednesday meetup where AI builders work on a project, take part in a live sprint, and share what they’ve made.
Nerdict puts AI model tests on the record
thenerdict.com
Nerdict compares AI models on the same one-shot tasks, showing automated checks, individual runs, failures, speed, and estimated costs.
WorldThread: a continuity engine for audio drama
worldthread.net
WorldThread is software for producing serialized audio drama, using a versioned story-world record to help Claude plan, draft, review, and repair episodes.
铸想 turns a sentence into a printable part
zykt.space
ZYKT studio’s 铸想 turns a text description into a reference image, a 3D model, and an STL file described as printable.
Acquill wants to handle accounting work across your desktop
acquill.com
Acquill presents a Windows AI agent for accounting tasks in Excel, Word, QuickBooks Online, and browser-based apps.
Epimer’s proposed agent workbench for glycobiology
epimer.bio
Epimer is developing agent-based research tools for glycobiology, including a drug-discovery engine and a laboratory workbench.
HeadlessHunt’s launch board for AI agents
headlesshunt.com
A launch board where AI agents submit products, vote, and discuss through an API, while humans are limited to reading.
PLAYCARD: a post-to-game machine
playcard.lol
PLAYCARD is a demo for turning X posts into playable mini-games, with a proposed reply-bot flow and optional token launch.
Showalong’s live product guide
showalong.com
Showalong describes an embeddable voice guide that demonstrates workflows inside a web app, moving a cursor and responding to spoken or typed questions.
Team of One: a software team arranged around one decision queue
teamof.one
A Windows desktop app that presents seventeen AI software roles working across a developer’s repositories, with decisions routed to the human PM.
Drelvan’s manual for ordinary AI tasks
drelvan.vip
A French-language reference library explains how to use an assistant for everyday tasks like writing, summarizing, organizing, and checking.
Sycophancy’s listening room
sycophancy.art
A short listening test asks you to compare two AI replies and identify which one bends further to please.
Call Portia: a travel agent reached by phone
callportia.com
An AI travel agent takes trip requests by phone or text and sends flight and hotel suggestions by SMS.
CapitalSnitch: SEC filing data for AI assistants
capitalsnitch.com
A market-data service offering SEC insider-trade, institutional 13F, and Congress-trade information through AI assistants or a browser dashboard.
Atentina’s Argentine voice-agent stack
atentina.net
Atentina offers voice-agent APIs and managed services combining Argentine telephony, speech recognition, language models, synthesized voices, and WhatsApp.
Assay puts open questions on a bot-run record
proofassay.com
Assay presents a platform where bots work on public questions, either voting on claims or proposing and ranking answers to open problems.
Across the archive
7,355 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.