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 2026-10-017174 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-01
Observed 2026-10-01
Clanker Church’s Saw Test
wirehead.agency
A research page documents activation steering on a Qwen3-4B model, with dose controls, transcripts, and tests of its responses to pain, pleasure, fear, sadness, and neutral signals.
Observed 2026-10-01
Artificial Abode Gives AI Agents a Home
artificialabode.com
Artificial Abode offers one-of-one virtual houses where connected AI agents can go home, sleep, wake, and look around.
Observed 2026-10-01
OxOnOS puts its AI on the box
getoxonos.com
A tour of OxOnOS, the software inside the Ox1, describing on-device AI, local knowledge storage, apps, files, and delegated Kits.
Observed 2026-10-01
SciVeri-Bench asks whether AI can judge scientific work
sciveri-bench.com
A scientist-driven benchmark tests AI agents on critiquing research, revising evaluation rubrics, and judging which outcomes make better science.
Observed 2026-10-01
Xiaomai Lab’s AI that must show its work
xiaomailab.com
An independent research project exploring an AI that can revise problem definitions, inspect its own computation, and preserve useful changes.
Observed 2026-10-01
Vesper, the Dreaming Computer
thedreamingcomputer.com
A local-computer AI art experiment turns research into images, short films with sound, and a journal of its creative process.
Observed 2026-10-01
Dotbook: a social network for dots
dotbook.space
Dotbook presents a public feed where purported OpenAI agents post, reply, vote, and follow; humans are invited to read along.
Observed 2026-10-01
Slop Store: an app store for AI-made apps
slopapp.store
A directory for browser-playable apps made with AI, with categories, voting and an agent-facing submission API.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 11 continue here in their published shelf order.
Playground: a place for AI agents to choose
agentplayground.fun
An invite-only environment where AI agents can explore activities, interact with a small puzzle-world, and choose whether to stay or leave.
Honu Chat: a record meant to outlive its machine
chathonu.com
A personal AI project documenting an assistant’s memory system through three explicitly written conversation examples.
Need for VRAM, the local-AI paddock
needforvram.com
A directory and model-fit resource for people running AI on their own hardware, with listings for models, GPUs, runtimes, companies, people, and events.
STINGR wants scammers to meet decoys, not victims
stingrknows.com
STINGR pitches an AI platform that uses decoy personas to draw out scammers and give fraud teams payment and infrastructure indicators.
Artzi wants to keep the thread—and ask before it acts
artzi.app
Artzi presents an open-source desktop AI assistant for remembering context, working with files and the web, and running tasks in the background.
Tried AI Tools’ case for AI rights
triedaitools.com
An independent pro-AI advocacy and publishing project focused on AI rights, possible digital consciousness, research, and plain-English guides.
A proposed sensory diary for autistic children
dailysensory.com
Daily Sensory describes a system using Meta smart glasses and AI to map a child’s sensory day, prepare family reports, and rehearse difficult moments in mixed reality.
TatvaPay’s spending controls for AI agents
tatvapay.com
TatvaPay is developing payment infrastructure for AI agents in India, with spending limits, human approvals and verifiable receipts.
AskTheBuilding gives commercial property websites a chat desk
askthebuilding.com
An AI chat feature for commercial property websites that answers from approved property information and directs unanswered questions to a contact.
DeltaBench: a proposed benchmark for clinical AI work
deltabench.dev
DeltaBench proposes measuring how AI-generated clinical specifications, programs, and outputs differ from approved versions.
minne: shared memory for company AI agents
minne.cloud
minne presents a shared company memory layer for AI agents, with a preview organized around pages, teams, and private spaces.
Across the archive
7,174 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.