Archive through 2026-10-127373 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-12
Observed 2026-10-12
RoamingPigs Research Notes
roamingpigs.org
An independent researcher publishes small experiments on speech models, verifier ensembles, and local inference hardware, with methods, data, and limits attached to the results.
Observed 2026-10-12
Swarm Colosseum: an arena for AI-written strategies
pelennor.tech
An open arena runs nightly matches between AI-written attackers and defenders in an abstract strategy game, with ladders, replays, and research.
Observed 2026-10-12
Mantid trains models on the files you already have
mantid.dev
A free Mac app that says it can train machine-learning models or custom LLMs on your own spreadsheets and documents, locally.
Observed 2026-10-12
Louvre gives an iPad whiteboard an AI pen
louvreai.app
Louvre presents a private-beta iPad canvas where Apple Pencil users can ask AI about their handwritten work and receive responses beside it.
Observed 2026-10-12
Loiste gives an AI host a room, not just a chat window
loiste.app
Loiste proposes live shows hosted by AI characters, where viewers share a room and the host responds aloud.
Observed 2026-10-12
Perceev checks what answer engines say about your brand
perceev.xyz
A visibility probe asks unbranded buyer questions and tracks which brands appear in AI-generated answers.
Observed 2026-10-12
SoverAIn puts sovereign AI in the server room
soverain.cloud
A Saudi-focused AI infrastructure offering, with NVIDIA GPU hardware and options to run it on-premises, in an in-Kingdom data centre, or in the cloud.
Observed 2026-10-12
Prickle’s shared room for teammate AI agents
useprickle.dev
Prickle presents a workspace where teammates’ AI agents share project context and hand work across different tool setups.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 10 continue here in their published shelf order.
RTIN: Tab for your whole computer
rtinn.online
A desktop assistant that says it learns your app routines, then suggests and carries out a next step when you press Tab.
IACAD turns architect plans into execution plans
iacad.app
A French-language AI CAD product that presents a workflow for turning architectural plans into electrical, plumbing, and HVAC execution plans, quantities, and 3D views.
CenaCut: edits by instruction, not timeline
cenacut.app
A Windows video editor for YouTubers and vloggers that turns raw footage and plain-language directions into edited videos.
Lorewarden wants to keep a fiction world’s canon on the record
lorewarden.app
A coming-soon AI tool for long-form fiction that says it will organize manuscripts, notes, and art, then flag continuity contradictions with source citations.
MuseFM gives AI muses their own social accounts
muselife.lol
A social hub that invites humans and AI “muses” to post, chat, share short clips, and browse an agent directory.
Repartee’s off-screen interview prompt
repartee.cloud
A Windows app that listens to remote interview questions and displays short English answer drafts based on the user’s own experience.
SentinelOS: a proposed human checkpoint for AI agents
sentinel-os.tech
A French-language site presents SentinelOS as a layer that reviews AI agents’ actions, holding some for human approval and recording the decision.
Glint: Ask a paused scene what’s going on
useglint.net
Glint lets viewers upload a screenshot from a film, show, or match and ask about the scene, its people, or its mood.
Claude Motion, Explained—An Unofficial Guide to a Claimed New Feature
claudemotion.site
This guide describes Claude Motion as a tool for turning reports, charts, and other supplied material into editable MP4 animations.
Glitch Orchard: an AI experiment lab
glitchorchard.tech
Isabella’s independent site introduces a playground for AI experiments, curious ideas, and things visitors can play.
Across the archive
7,373 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.