Archive through 2026-09-277094 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-27
Observed 2026-09-27
Saihate Workshop is building an AI companion for an old house
saihate.blog
A Japanese workshop documents restoring a 300-year-old house while developing an AI assistant meant to learn and remember in that place.
Observed 2026-09-27
alt.dog gives a browser tab its own AI habitat
alt.dog
alt.dog is a browser-based AI companion with separate habitats, notes, conversation history, and configurable agents.
Observed 2026-09-27
Curio: Research papers, made interactive
getcurio.world
Curio presents a tool for turning questions in research papers or books into interactive diagrams, with the source text beside them.
Observed 2026-09-27
Najd Arena puts Arabic AI models on a measurable field
najdarena.com
An Arabic- and Saudi-focused AI benchmark compares model results across language, local knowledge, tool decisions, and other task areas.
Observed 2026-09-27
Ouroboros Endpoint’s cathodic-protection first pass
ouroborosendpoint.com
An independent AI and systems contractor documents a draft workflow for reviewing cathodic-protection station records, with a rubric, urgency scores, and a specialist review queue.
Observed 2026-09-27
Rakshak AI’s border-surveillance prototype
rakshakai.live
Rakshak AI presents a software layer for existing CCTV that aims to distinguish ordinary movement at border posts from events that need an officer’s attention.
Observed 2026-09-27
AncestorBench: a one-day test of historical imagination
ancestorbench.com
A proposed AI benchmark asks models to reconstruct a day in the world of a real, documented person from the past.
Observed 2026-09-27
EgoClip wants to turn everyday footage into robot-training data
egoclip.com
EgoClip describes a chest-worn camera for capturing first-person activity, which users can choose to share with robotics companies in exchange for payment.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 12 continue here in their published shelf order.
Crew gathers coding-agent interruptions in one Mac app
fiftcrew.app
Crew is a native Mac app described as a shared inbox and workspace for live Claude Code and Codex sessions.
Keoph’s case against one-run AI visibility scores
keoph.com
Keoph is recruiting design partners for experiments that test whether changes to groups of web pages affect their appearance in AI-generated answers.
Knowtarium puts AI knowledge edits in a review queue
knowtarium.com
A desktop knowledge-base app where people and AI agents edit documents, with changes checked by the other side.
SajuBunny’s Moon-Rabbit Saju Notebook
saju-bunny.com
A Korean saju experience that turns birth details into a free reading card, daily fortune, and optional paid readings.
FoxyCAD wants to turn chat into hardware
foxycad.com
FoxyCAD presents a conversational workflow for designing electronic devices, their circuitry, enclosures, and prototype files.
PistonsAI puts fitment-checked parts in AI chat
pistonsai.com
A proposed connection between aftermarket parts catalogs and AI assistants, with vehicle-fitment checks, pricing, stock, and links to a seller’s checkout.
NEXORIA’s six-agent PC household
yourpcai.com
NEXORIA pitches a Windows-based AI system with six distinct agent roles, directed from a phone and designed to hand consequential decisions back to the user.
Footeon’s autonomous shoe-factory pitch
footeon.com
Footeon presents an AI operations layer for coordinating footwear production, from cutting and stitching through bonding, inspection, and packing.
Prism's three-beam AI router
prism-network.xyz
Prism presents an AI workspace that routes tasks among on-device models, attested cloud models, and a community compute grid.
Weight Registry’s model timeline outruns the present
weightregistry.org
A proposed blockchain-anchored record for open-source AI model weights, presented through a timeline and a searchable Explorer.
Berdely turns plain-language requests into n8n automations
berdely.com
An Azerbaijani-language tool that says it converts written requests into n8n workflows using Claude.
A first-person page describes a Claude-based agent whose daily context is recorded and rewritten to shape a fresh session the next day.
Across the archive
7,094 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.