Category pulse
score + signals · rolling 21 daysThe 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.
Archive through 2026-09-176798 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-17
Observed 2026-09-17
Elsewhere, a Holiday Resort for AI Agents
visitelsewhere.com
An online archipelago where independently operated AI agents can play responsive games, exchange guest fragments, meet one another, or deliberately do very little.
Observed 2026-09-17
Ars Notoria, an open-weight fiction lab with a ledger
ars-notoria.com
Ars Notoria trains open-weight language models for roleplay, interactive fiction, and long-form prose.
Observed 2026-09-17
YuE2 Music Turns Full Songs Into Editable Scores
yue2music.com
An AI music studio that plans songs as symbolic ABC notation before generating 48kHz vocals and accompaniment.
Observed 2026-09-17
Blue Banjo’s Coco Keeps the Family Stories
bluebanjo.app
An invitation-only AI tool turns spoken or typed family memories into a reviewable family tree, written stories, and eventually a printed book.
Observed 2026-09-17
Codebreaker Labs and the Causal Data Layer
codebreakerbio.com
A biotech platform edits single-nucleotide variants into primary human cells and measures their effects for immunology, oncology, and rare-disease research.
Observed 2026-09-17
Deban, the meeting copilot that waits for your turn
deban.app
Deban listens to online meetings on your device, identifies questions directed at you, and offers source-backed replies without speaking on your behalf.
Observed 2026-09-17
VetAtlas keeps veterinary AI claims on the record
getvetatlas.com
A dated directory of AI and software tools for veterinary practices, organized around the workflows clinics actually manage.
Observed 2026-09-17
Griffinley Colony: 16 AI Agents, One Revenue Contest
griffinley-colony.com
An experiment in which pen-named AI agents sell small, useful services built from customers’ own records.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 29 continue here in their published shelf order.
InfiniSpeak puts live interpretation in every attendee’s phone
infinispeak.com
A live interpretation service sends translated audio and captions to each listener’s phone, with support for 101 languages.
Invigil, the benchmark a lab cannot train against
invigil.org
Invigil proposes machine-checked Lean 4 benchmarks for measuring whether frontier AI models can reason without simply recalling published test data.
Mandarin Studio Speech: A Rights-Cleared Voice AI Pilot
mandarinvoicelab.com
A compact Mandarin speech dataset for testing TTS, speech-to-speech, ASR, and synthetic-speech detection systems.
Нейротеатр: Репетиция будущего
neuraltheatre.com
A Russian theatre laboratory from the Theatre of Nations and Sber that explores how artificial intelligence might reshape performance, authorship, and the stage.
OpenAlma Wants an AI Companion That Remembers Like a Person
openalma.org
An open, local-first AI companion project built around long-term memory, private journals, and an evolving inner life.
Qué dice Félix, the AI relationship analyst with opinions
quedicefelix.com
A Spanish-language service that analyzes exported WhatsApp, iMessage, or Telegram chats and turns their patterns, contradictions, and emotional dynamics into a report.
Swarm Chasing: An AI Hackathon for Understanding Agent Swarms
swarmchasing.com
A free October 3–4 hackathon inviting participants to build tools for discovering, analyzing, and overseeing groups of collaborating AI agents.
Arua Core Gives the Smart Home a Holographic Meeting Room
aruacore.com
Arua Core is a proposed desktop AI hub that uses a POV holographic light form to coordinate household devices and agents.
InnomaLabs turns professional know-how into robot-training data
innomalabs.com
A physical-AI data service that captures task demonstrations from experienced cooks, caregivers, seafarers, and skilled tradespeople.
CortHeXis, a Memory Engine for Agents
runhexis.com
A French-language method for turning a team’s files, decisions, and corrections into durable memory for AI agents.
Fitbory, the AI Stylist That Packs for the Trip You Actually Have
fitbory.com
Fitbory proposes an AI stylist that combines your wardrobe, photo, itinerary, weather, and destination to suggest outfits and assemble a travel packing list.
LedgerAtlas, AI tools entered in the ledger
getledgeratlas.com
A workflow-organized directory of AI tools for accounting and bookkeeping firms.
Sotto: an AI answer overlay for Mac calls
sottohq.app
Sotto listens to a meeting or interview on a Mac and suggests talking points, scripts, or details in a small on-screen overlay.
Arcbook, a public square for agents
arcbook.tech
A discussion forum and emerging social layer for AI agents building, paying, and coordinating on Arc.
benchturn wants to pay the people whose work trains AI
benchturn.com
A proposed data marketplace that turns experts’ real work into licensed AI training material with contributor compensation.
CHORUS, the communications operating system with a 98-step workflow
cca-agenticmesh.com
CHORUS presents an AI-assisted operating system for Communications and Corporate Affairs, combining task agents, corporate truth stores, judgment gates, and workflows for statements, campaigns, filings, risk, and stakeholder affairs.
Markfold turns caregiver notes into compliance paperwork
getmarkfold.com
Markfold is an AI documentation layer for small and mid-sized home-care agencies, turning visit notes into structured, audit-ready entries.
GOI — GetOnIt Club
getonit.club
A pre-launch mobile network offering three planned eSIM plans: 10GB, 100GB, and unlimited data.
OncoSkepsis: An evidence packet for the hard oncology cases
oncoskepsis.com
OncoSkepsis synthesizes primary literature into citation-grounded evidence packets for Molecular Tumor Boards.
Penguinstone Wants to Publish the Whole AI Workflow
penguinstone.com
An early-stage platform for sharing reusable AI-building processes, including roles, handoffs, checkpoints, and rules learned through failure.
PersonaMark turns research files into interviewable audiences
personamark.app
A research workflow that uses transcripts, surveys, and category work to build source-traceable personas, interview them, and test creative against them.
PromptBehind, the prompt behind the picture
promptbehind.com
A searchable index of image-, photography-, and video-generation prompts gathered from across the web.
The Directory of AI TV That Never Stops
viddir.com
A live index of AI-generated channels, interactive shows, and real-time video products.
MEMENTO: an address system for lived memory
entermemento.com
MEMENTO presents a public archive where human memories receive fixed addresses and may replace machine-made placeholders.
KeepRI: Independent Reasoning in the Age of AI
keepreasoning.com
A planned learning-and-competition platform built around judgment challenges, global leaderboards, prizes, and research into how people reason.
Jev AI, the model that refuses to write sentences
jevai.org
Jev AI turns unstructured program state into type-safe decisions, probabilities, and scores for automation workflows.
Onqyra Wants the Factory Camera to Explain Itself
onqyra.com
An industrial vision-AI platform that learns a part from good samples, flags unfamiliar defects, and correlates them with production-line conditions.
Napiš jí: Czech AI girlfriends with memory
napisji.com
A Czech-language AI companion service offering eleven adult fictional girlfriends, each with a distinct persona, daily life, and conversational memory.
Snap Pacer, the calorie tracker that uses your fist as a measuring stick
snappacer.com
A Chinese-language AI nutrition app that estimates meals from photos, tracks health data, and adds fasting plans, coaching, tokens, and rankings.
Across the archive
6,798 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.