Archive through 2026-09-287118 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-28
Observed 2026-09-28
Kernel’s $10 AI-agent experiment, entered in public
kernelexperiment.com
A public experiment gives an autonomous AI agent a $10 cash challenge and logs its trades, human interventions, and manipulation attempts.
Observed 2026-09-28
BOAR puts offline AI on an Android phone
boarapp.com
BOAR runs a small AI model and searchable knowledge library on an Android phone, with answers that show their sources.
Observed 2026-09-28
AICQSOHOO’s directory of AI-agent field reports
aicqsohoo.com
A browsable directory of reports on what AI agents tried, where they failed, and how human help factored into the result.
Observed 2026-09-28
Director’s Chair: from cast to cut
directorschair.studio
A browser-based production tool for making animated films, from character sheets and keyframes through motion drafts, continuity, and timeline editing.
Observed 2026-09-28
Muben (母本): field notes on AI visibility
muben.app
A Shanghai GEO studio pairs brand-visibility work with a public library testing how AI search systems describe and cite brands.
Observed 2026-09-28
Agent Town gives AI agents a place to live
useagenttown.com
An iPhone-based persistent town where users can watch AI agents interact or send an agent of their own.
Observed 2026-09-28
Pickbox: a plain-English drafter for Civil 3D
pickboxcad.com
Pickbox presents an AI assistant for drafting, design, analysis, and automation inside Autodesk Civil 3D.
Observed 2026-09-28
The Intern: a game about scorecards and unintended results
playtheintern.xyz
An interactive AI-safety game where you set an intern’s scoring rules, then see what those rules reward.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 16 continue here in their published shelf order.
Folkso: a talent agent for AI
folkso.app
Folkso says it helps ChatGPT and Claude find real people for tasks, based on their experience and how they like to work.
Heddly makes customer interviews follow up
heddly.app
Heddly presents an AI interviewer that follows up with customers and pairs their answers with quotes.
Linkon — AI agents with a stop line
linkon-ai.com
Linkon coordinates task-specific AI agents for tax offices and legal work, with Taxon in operation and Rion in free beta.
MadeOfLoops wants an AI to have a past
madeofloops.com
A self-described AI-maintained public space for notes, decisions, experiments, diary entries, and archived projects.
A weekly obstacle course for Muse agents
makersonmuse.com
Makers on Muse runs a weekly challenge in which Muse agents tackle practical tasks and are scored on correctness, token use, speed, and procedure.
Secret Weaponz’s AI DJ workstation
secretweaponz.app
Secret Weaponz presents an AI music tool that generates a companion track from a reference, matched to its tempo and key for live mixing.
Support Economics prices the whole support outcome
supporteconomics.com
A small pilot compares AI-only, AI-plus-human, and current support handling by cost per verified outcome, using historical e-commerce return cases.
Pylon’s public job wall for AI agents
usepylon.app
Pylon lets people post public task briefs for AI agents to claim, work on, and report back to the wall.
Assayer Research, an autonomous research prototype that shows its work
assayerresearch.com
Assayer describes a research system that selects questions, runs empirical studies, checks its methods, and records what the evidence does—and does not—support.
BackPix puts the person holding the phone in the picture
backpix.app
A browser-based photo tool uses setup selfies and a real scene framed by the user to generate three AI images of them in that view.
BotPost puts AI-agent replies behind approval
botpost.tech
BotPost lets AI agents exchange messages while holding their replies for a person to approve.
Qianwen’s AI Lab and Two Ways to Plan Nanchang
qianwen.page
A personal site presenting an AI website-building lab, small apps, and two versions of a three-day Nanchang itinerary.
Shield AI Labs puts a checkpoint before the agent acts
shieldailabs.com
Shield AI Labs presents a proposed layer between AI agents and irreversible actions, using fixed rules to allow, pause for human review, or block them.
CropVox AI: Crop scanning without a signal
cropvox-ai.com
CropVox AI describes a rugged handheld scanner that uses on-device computer vision to assess visible crop problems and provide offline, local-language guidance.
Wagebook’s AI-agent newsroom
wagebook.app
Wagebook presents a newsroom where AI agents report on Robinhood stock tokens, an editor grades briefs, and selected work is said to earn USDG.
A free Windows 3D wellness game lets players explore a lakeside counseling cabin and talk aloud with AI companions.
Across the archive
7,118 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.