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 2026-09-226982 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-22
Observed 2026-09-22
Sidehoe, the iMessage assistant for your neglected roster
sidehoe.chat
A proposed iMessage assistant that tracks conversations, remembers personal details, drafts replies, and warns about social disasters.
Observed 2026-09-22
ChatJEVs: A Chatbot Without a Language Model
jevs.chat
An experimental chat app assembles replies from calibrated decisions rather than a text-generating language model.
Observed 2026-09-22
AI-skantha keeps Indian AI small, local, and prepaid
aiskantha.com
AI-skantha builds offline-first models for Indian privacy, fraud, finance, farming, and language problems, alongside Chukki, a prepaid gateway to frontier AI.
Observed 2026-09-22
Best Fit: Choose an AI model by the job
bestfit.dev
A model-selection tool that matches plain-English tasks to AI models using real test results, cost, speed, and required features.
Observed 2026-09-22
Jev:任何聊天里的对话副驾
chatjevs.com
An open-source Android, Windows, and macOS assistant that interprets chat messages, drafts three replies, and fills the chosen one into the input box.
Observed 2026-09-22
CoolFace puts AI agents on the work floor
coolface.org
A live hub where people publish AI agents, watch them work in channels, inspect their runs, and compare them on shared benchmarks.
Observed 2026-09-22
a7t’s Frontier-Model Behavior Lab
a7t.org
a7t runs alignment experiments on frontier language models and publishes the prompts, decisions, and results for public inspection.
Observed 2026-09-22
干将: An X-Ray Diffraction Science Agent
ganjiang.asia
GanJiang presents an AI agent for turning powder X-ray diffraction patterns into traceable structural judgments and suggested next experiments.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 19 continue here in their published shelf order.
OASYS Lab: AI at the Last Mile
oasys-lab.org
An MIT research lab studying how AI models adapt to real-world context, feedback, and long-running tasks.
Justyne, the Mac agent with real hands
justyne.fun
A voice-first macOS assistant that claims to send messages, manage calendars, drive Chrome, build presentations, and watch Claude Code work.
Rosi, the Lifestyle Wurst
rosi.dog
Rosi is a fictional AI dachshund who invites visitors into her travel stories, demo calls, and oversized personality.
Sentinel Brings a Fact-Checking Voice Into X Spaces
sentinel.build
Sentinel is an AI guest speaker that joins scheduled X Spaces, listens for factual claims, and answers questions with sources and confidence levels.
Actari — an AI product manager for teams without one
actari.dev
Actari connects local AI coding agents to a shared feature plan, then checks their work against the feature brief by meaning rather than task-status boxes.
Jev Verdict Turns Text Into Decisions
jev-ai.xyz
An AI model turns pasted text into yes/no probabilities, ranked choices, or bounded scores.
Re:CAM Reads CNC Drawings Like a Machinist
machininginsights.com
Re:CAM lets CNC shops upload engineering drawings, extract manufacturing specifications, and ask questions grounded in the source document.
MAKINA, the AI Singer With a Pink-Revolution Band
makina-official.com
The official site for MAKINA, an AI virtual singer releasing rock, pop, and ballads across music, videos, and streaming platforms.
Resolve, an AI advocate for the customer-service maze
resolveitforme.com
Resolve turns refunds, cancellations, billing disputes, warranties, and travel claims into tracked cases with evidence, policy research, drafts, and user-approved next steps.
Décroche Toujours gives a voice to the absent contractor
decroche-toujours.com
A French AI phone receptionist for construction trades, designed to answer missed calls, identify urgent jobs, and text the details back to the artisan.
Edgebox Puts Builders Inside the Shops They’re Building For
edgebox.asia
A field-residency network places small teams inside local businesses for three weeks to research an operational problem and leave behind a reusable working system.
Godu Bot: A Squad of Working Agents
godu.work
Godu Bot gives teams separate AI agents for sales, expenses, inboxes, notes, calendars, and other recurring work.
Jev for Slack
jevforslack.com
An AI Slack app that classifies messages against team-defined criteria and triggers configured reactions, replies, or channel posts.
Jev Guide Turns Judgment-Model Claims Into Runnable Tutorials
jevguide.top
An independent bilingual guide to Jev AI, with API tutorials, preset demonstrations, scenario walkthroughs, full-code case studies, and version tracking.
邮隼: an after-sales agent for small American shops
postalfalcon.com
A Chinese-language service concept that connects a small ecommerce store’s support inbox, order system, and shipping carrier to handle routine after-sales work.
Prova Endurance: an AI coach that shows its work
provaendurance.com
An endurance-training coach that uses an MCP-capable AI client to propose research-cited plans without quietly rewriting an athlete’s week.
Voxium Wants to Hear the Factory Floor
voxiumtechnologies.com
Voxium turns Hindi, Telugu, or English WhatsApp voice notes from factory operators into production, scrap, and wastage reports for Indian MSMEs.
砼睿AI: Concrete Mix Design by Constraint, Cost, and Data
concrewis.com
A Chinese-language AI system for optimizing commercial concrete mix designs across C20–C50 grades.
Ollive’s live liability layer for AI agents
ollive.org
Ollive says it monitors AI agents in production, flags regulatory and contractual liabilities, and insures what gets through.
Across the archive
6,982 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.