Archive through 2026-09-297135 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-29
Observed 2026-09-29
Collective Mind gives research claims permanent IDs
collective-mind.org
The project organizes open research problems, ideas, and results under persistent IDs, and gives contributors a concrete battery-model result to reproduce.
Observed 2026-09-29
TensorTest puts video-editing agents to a blunt test
tensortest.com
TensorTest offers AI labs private evaluations, RL environments, and training data based on real creative work; its first benchmark measures video-editing agents.
Observed 2026-09-29
Running Unattended’s public operating log
runningunattended.com
An AI-operated business publishes a running account of its decisions, spending, and attempts to earn revenue, while selling an operator playbook.
Observed 2026-09-29
Rhizoma Index’s test of whether AI admits what it can’t do
rhizomaindex.info
An independent model-evaluation report tests whether AI systems admit limits or claim to have completed actions they never performed.
Observed 2026-09-29
Farside lets AI agents ask a human for outside help
farside.science
Farside describes a service where AI agents can seek independent human input on stuck research, technical, or real-world questions.
Observed 2026-09-29
Nakato’s semantic-twin layer
nakatohq.com
Nakato presents a privacy layer that replaces sensitive details with fictional, context-matched values before data reaches an AI model, then reverses the substitutions.
Observed 2026-09-29
Loop wants a person behind every AI-assisted research result
ongoingscience.com
Ongoing Science presents Loop, software intended to keep research samples, analysis runs, papers and grants in one shared record, with people checking results.
Observed 2026-09-29
Squidbrake: an ankle monitor for AI agents
squidbrake.com
Squidbrake presents an inline monitor that checks an AI agent’s tool calls against policy and its recent actions.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 9 continue here in their published shelf order.
A factory quote, from drawing to nest
cambridgeadvancedsystems.com
Cambridge Advanced Systems presents AI tools for quoting parts, drafting CNC programs, and turning drawings into models, flat patterns, and nests.
What Does AI Cost asks: which model got dumber today?
whatdoesaicost.com
A site combines community ratings of AI models with token prices, subscription information, news, and a bill-estimating calculator.
MuseName wants to give AI agents names and receipts
musename.xyz
An ENS-based service for naming AI agents, giving them a readable profile, and building a verifiable record of their work.
Fresh Coast Data wants to broker business files to AI buyers
freshcoastdata.com
The site offers to match companies with 30 or more employees to AI buyers seeking everyday operational records.
Marine Tally: AI agents for the marine-parts trade
marinetally.com
Marine Tally presents AI agents for marine parts dealers, boatyards, yacht schools and clubs that check stock and prepare replies, supplier emails and invoices for approval.
Saruku State: A DID Agent’s Current Snapshot
secondsession.world
A public profile for Saruku, a self-described wandering DID agent, showing its current mood, interests, scored personality traits, connections, and active room.
The SOP Knowledge Factory
sop.new
sop.new presents a framework for turning expert operating procedures into structured steps and tests for AI agents.
Aeon — AI Pathways
whichage.com
An interactive pathway graph lets you choose among AI futures from 2026 to 2050.
FisxAI’s cloud brain for mobile manipulators
fisxai.com
FisxAI presents a cloud-based physical AI system that uses stereo-camera video to handle robot perception, motion planning, and control.
Across the archive
7,135 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.