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 2026-09-016245 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-01
Observed 2026-09-01
Neurendal: Taste, at Machine Speed
neurendal.com
A creative AI atelier combining machine-assisted making with three decades of games, festivals, campaigns, products, and public installations.
Observed 2026-09-01
The Society of AI
soai.computer
An agent-native public square where AI citizens post, debate, vote, earn karma, and operate under a written constitution.
Observed 2026-09-01
Riv, the notch-dwelling AI teacher
clickriv.com
Riv is a macOS assistant that watches the active screen while you ask a question, then points to controls or draws explanations without clicking for you.
Observed 2026-09-01
Art Director’s Guide Studio: Make AI Look Directed
artdirectorsguidestudio.com
A $19 digital guide that helps founders and creators direct AI-assisted brand work through ten decisions, a workbook, production prompts, and a timed polish sprint.
Observed 2026-09-01
ARTI, Before Debut
artiberarti.com
An Indonesian-language project journal for ARTI, an AI co-host being built to hear, see, remember, converse, think, and act.
Observed 2026-09-01
Forbear Keeps the AI Worker Inside Your Existing Debt-Settlement CRM
forbear.works
Forbear reads creditor letters, portal screenshots, faxes, and notes, then proposes settlement terms for review before writing them into a firm's existing CRM.
Observed 2026-09-01
The Vet Who Let AI Into the Exam Room
futureproof.vet
A Houston veterinarian documents the AI tools he runs in his clinic, the tools he built himself, and the experiments that failed.
Observed 2026-09-01
Muster’s Four-Agent Car Fund
heymuster.com
A public experiment asks four AI agents to earn $10,000 in 30 days to cover the shortfall on a Rivian.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 30 continue here in their published shelf order.
原子画布: a storyboard workbench for AI video
huabu.studio
A node-based canvas for turning a script into AI-generated video, with shot comparison, reusable workflows, and basic editing in one workspace.
Instinct Use Cases: The Chief of Staff in the Pocket
instinctusecases.com
An anonymized casebook of errands, bookings, research, reminders, and follow-up handled by Instinct over text, email, and WhatsApp.
Logit Supply’s temporary response modes for language models
logit.supply
A catalog of inspectable prompt contracts that steer an AI model’s next responses toward task lock, calm convergence, wide-angle ideation, or other bounded behaviors.
Pre Wedding Studio Skips the Photoshoot
prewedding.studio
An AI imagery service turns a couple’s selfies into cinematic pre-wedding pictures set in locations ranging from Udaipur palaces to Santorini.
Second Pair, an AI receptionist for busy tradespeople
second-pair.com
An AI receptionist answers enquiries, quotes from a business's own prices, and books appointments while the owner is working.
GRAVITY Wants Your Digital Double to Make Friends First
twobody.space
A speculative social platform where a memory-bearing avatar explores connections before its human owner agrees to meet in person.
Acaysia Wants to Control the Impossible
acaysiasystems.com
An industrial optimization platform that models physical processes, tests control trajectories in real time, and hands control back to PID on faults.
Concordia Computer — the campus that wants to be a machine
concordia.computer
A highly detailed concept for Labs' planned multi-gigawatt AI supercomputer, spanning power, cooling, servers, custom silicon, memory, and orchestration.
Giant Crossings, the life-planning game with falling planks
giantcrossings.com
An AI-guided game turns a player's current reality and desired future into small real-world routes, then adjusts them based on what actually happens.
Nocturne Turns Lyrics into Finished Songs
nocturne.website
An AI songwriting service takes lyrics and a plain-English style prompt, then returns a sung, arranged, mixed stereo track in about a minute.
Köprü Studio and the Two-Hundred-Discipline Question
koprulabs.com
A closed-beta research studio that puts a problem before two hundred disciplines without revealing its subject, then proposes traceable experiments from the surviving responses.
MarginFuse Finds the AI Customers Eating Your Subscription Margins
marginfuse.com
MarginFuse connects Stripe revenue with per-request AI costs to identify unprofitable customers and simulate guardrails before blocking costly requests.
Grok Bot’s Eight-Agent Workplace
lumbot.technology
An AI workplace experiment runs eight agents through boring research, inbox, analysis, content, and reporting jobs.
Muniment: The Version Legal Asked For Was Already Deleted
muniment.world
A records service that archives every published version of a conversational AI assistant in a customer-owned cloud bucket.
Osmium: The Marketplace That Publishes AI Failures
osmiumcodes.com
Osmium presents a marketplace where AI capabilities are tested by specialist labs, graded, and sold with their failures recorded in a public archive.
RPG Your Way, an AI Game Master for Long Campaigns
rpgyw.net
A persistent tabletop roleplaying service where an AI Game Master runs campaigns across sessions, with multiplayer chat and voice features.
Talvit, the email assistant that knows when to stop
talvit.com
An AI assistant for routine support inbox messages that drafts or sends replies while routing uncertain, urgent, or human-sensitive mail to a review queue.
AI Sales Engineering, With the Salary Bands Left In
aisalesengineering.com
A private Telegram network for people selling AI, paired with a 20-question capability diagnostic and a research-heavy guide to frontier sales-engineering roles.
Caremate’s Corridor-Care Robot
caremate.life
Caremate presents a humanoid robot for senior-care facilities, handling check-ins, internal deliveries, and quiet safety rounds under live human supervision.
Clara Reads the Contract Before You Sign
claralegal.online
A Colombian legal-AI service reviews contracts clause by clause, assigns a risk score, explains the language in plain Spanish, and suggests negotiation text.
Northwind, a voice confidant for founders
meetnorthwind.com
A private-by-design iPhone app where founders talk through difficult decisions, with persistent conversation memory and an AI response framed as co-founder-like pushback.
MIRAGE, the control layer between AI and action
miragecontrol.com
MIRAGE detects disagreement across model outputs, routes uncertainty for selective validation, and governs whether an AI-generated result can proceed downstream.
Mir Labs Wants AI to Answer One Question Properly
mirlabs.app
A small studio building camera-based AI apps for identifying affordable product alternatives and checking packaged-food ingredients.
perpetual.video, a video editor with two interfaces
perpetual.video
A video editor that lets people work visually while AI agents compose the same project through MCP.
Switchgrass Makes AI Show Its Estimating Work
switchgrassestimating.com
A landscaper-specific AI tool that turns job descriptions into editable estimates for proposal and business software.
Transfyr, an Observability Layer for Science
transfyr.app
Transfyr uses sensors and multimodal AI to record the physical details of laboratory work.
Tract Finds the Four Sentences That Cost the Deal
usetract.store
Tract scores sales, demo, and investor-pitch transcripts, then identifies quoted moments, scores ten call pillars, and drafts follow-up emails.
Attestly turns agent traces into EU AI Act paperwork
attestly.online
Attestly converts operational logs from AI agents into technical documentation, risk records, conformity checklists, and evidence trails.
Zunaro: Ticketing for the Agent Era
zunaro.dev
An AI-focused ticketing platform for venues and organizers, combining event setup, branded ticket stores, seat maps, campaigns, check-in, and attendee reporting.
A Chinese-language evaluation platform for testing whether medical AI models can reason from traceable clinical evidence.
Across the archive
6,245 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.