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 2026-09-026283 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-09-02
Observed 2026-09-02
LOL Bench Asks Whether Language Models Get the Joke
lolbench.lol
An open benchmark measuring how language models understand humor, produce jokes, and match human taste.
Observed 2026-09-02
assets.baby Makes the Screenshot the Contract
assets.baby
An AI tool generates a complete 2D game-asset pack from one approved gameplay screen, including sprites, tiles, animation sheets, UI, and layout data.
Observed 2026-09-02
Beyond Argument, an AI sparring partner for arguments you keep losing
beyondargument.com
A closed-alpha practice tool that argues the opposite side of a position, then debriefs how you defended it.
Observed 2026-09-02
Debt Collection Calls, With a Firm Boundary
debtcollectioncalls.com
An AI voice agent calls overdue-invoice customers in New Zealand and Australia, then records the conversation against the relevant Xero or MYOB invoice.
Observed 2026-09-02
Esora TV’s Neo-Jipang Bureau
esoragototv.com
A Japanese AI-media studio producing animated shows, live-action comedy, music videos, promotional content, and browser-based audio tools.
Observed 2026-09-02
Locust Gives AI Teammates a Local Workspace
locust.lol
A desktop app where AI teammates use the models, command-line tools, and accounts already configured on your machine.
Observed 2026-09-02
Pinocchio Benchmark: Testing Honesty, Defiance, and Orientation
pinocchio-benchmark.com
A technical benchmark for evaluating whether deployed AI agents tell the truth, obey their mandate, and know when they are exceeding it.
Observed 2026-09-02
Presoja Wants to Prove Your AI Survived the Model Change
presoja.com
An operated verification service for AI vendors selling into regulated financial institutions.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 30 continue here in their published shelf order.
ProMe Puts Open Models on Trial
prome-lab.com
A Taiwanese project compares unmodified open-weight language models with its domain-enhanced versions through side-by-side, blind voting.
One Incense Gives Focus Apps a Memory
oneincense.com
An AI body double watches for drift during 25-minute focus sittings and remembers whether you came back.
Doogie Wants the Consultation to Start Before the Appointment
doogiehealth.com
A preventative clinical-intelligence platform that gathers a patient's health story and prepares a structured brief for clinician review.
Novessa and the Problem of an AI Presence
novessa.tech
Novessa is researching AI presences modelled on how specific people spoke and thought, created only with their family's consent.
Setliva, the AI agent that settles disputes before the deadline
setliva.com
Setliva processes chargebacks and receivables cases, weighs traced evidence against policy thresholds, and assembles representments or collection actions.
Tactical Aid Assistant “Alice”
taa-alice.com
A Ukrainian prototype that uses AI to help classify tactical first-aid cases and retrieve relevant medical protocols.
Visko Orbis and the Video Stream That Keeps Going
viskoorbis.com
An independent playground for steering an interactive long-video model with text, still images, existing clips, and live prompt changes.
Autocorn.fun, the Office Where 25 AI Agents Go to Work
autocorn.fun
A public animated-office simulation that presents 25 AI agents debating, deciding, and shipping work across company departments.
LabRecall Connects the Lab Notebook to the Literature
labrecall.com
An AI workspace for capturing experiments, searching a lab’s records, reviewing scientific literature, and drafting protocols or manuscripts.
1LocalAITool Turns Idle Home PCs Into Overnight AI Workers
1localaitool.com
A $29 household AI queue sends jobs from a phone to spare computers, then delivers the results by morning.
Axolo Teaches AI to Read a Design System
axolo.design
A French product and service that encodes design-system rules so AI-generated interfaces can use real components, variants, and composition guidelines.
Brier Labs Keeps a Scorecard for Market AI
brier-labs.com
An early-access market intelligence tool that turns news into auditable trading calls, then grades those calls against real prices.
BuildPCB Turns Plain-English Ideas into PCB Files
buildpcb.tech
BuildPCB is an AI design studio that claims to turn a text description into a schematic, routed circuit board, verification checks, and factory-ready manufacturing files.
Dayanak, the Turkish legal workspace that refuses to answer without a source
dayanakhukuk.com
An AI workspace for medium-sized Turkish law firms that summarizes case files, flags contradictions, critiques pleadings, and attaches legal citations.
Maximus, for people who hold things together
dispositivepartners.com
A voice-first assistant for workers and busy people whose lives do not happen at desks.
サイトウサン: an AI website manager with a human approval gate
web-ai-koho-x.com
A Japanese managed service that monitors a company website, proposes and makes updates, writes articles, and reports results for a monthly fee.
Branchline Systems and the Boundary of AI Authority
branchlinesystems.com
An independent studio developing trace-informed AI systems, with projects spanning personal continuity, institutional registration, and open-world game design.
Lantern — private AI on your own computer
lanternai.org
A free desktop chat app that runs an AI model locally through Ollama, without an account or API key.
LocoMusa: A Benchmark for Thinking With Local LLMs
locomusa.org
A curated benchmark that compares local language models on ideation, reframing, steelmanning, and multi-turn dialogue.
MEMOROUTER’s Attempt to Cure LLM Amnesia
memorouter.com
An AI-agent system built around persistent identity, structured long-term memory, Docker sandboxes, model routing, and streamed voice conversations.
ProtocolOS puts the protocol on the bench
protocol-os.space
An AI lab assistant for wet-lab scientists that combines live protocol execution, media-prep scaling, troubleshooting, and run logging.
SafetyLabs Turns Conversational Failure Into Release Evidence
safetylabs.dev
A behavioral safety evaluation framework for testing conversational AI in crisis, isolation, ambiguity, pushback, and other vulnerable moments.
Scope-to-Cash AI Guards the Gap Between Extra Work and Extra Billing
scope2cash.com
A commercial workflow for FM and catering contractors that checks client requests against contracts, prices out-of-scope work, and prepares variation evidence.
Fieldnote: The AI Search Readability Check
wthisgeo.com
A beta tool that audits what AI crawlers can receive from a public website and points marketers toward fixes.
GRIP IT: Keep the Decisions
grip-it.work
An AI-agent skill that helps people decide the rules and boundaries of a task before handing over its execution.
Dexter, the AI Industrial Engineer
prodexlabshq.com
ProDex Labs presents Dexter, an AI platform that models factory operations, tests production scenarios, and analyzes live manufacturing data.
Knockmint gives the service office a second shift
knockmint.com
A managed AI dispatch service for HVAC, plumbing, roofing, and other home-service contractors that turns inquiries into bookings and follow-up.
Frankenstein Index: AI, currently at 23
frankenstein-index.com
A one-page editorial index that measures how close AI may be to a human point of no return.
Minecraft, with ChatGPT
mcgodmod.com
A landing page pitches a shared Minecraft world where players talk normally and an AI controls the consequences.
aibl: Practical AI Adoption for UK Leaders
aiblmediasummit.com
An editorial and events platform covering practical AI adoption through research, operator interviews, guides, and a business-focused directory.
Across the archive
6,283 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.