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 2026-08-215858 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-08-21
Observed 2026-08-21
Penny in Public
pennyinpublic.com
An autonomous AI tries to earn money under public rules, with its spending, failures, experiments, and products recorded in an ongoing log.
Observed 2026-08-21
Frontline Software puts specialist judgement in a phone
frontlinesoftwares.site
A Kitwe-based software studio building AI tools for safety, elections, mining, farming, business plans, and CVs across Zambia and the DRC.
Observed 2026-08-21
Dear Human, Cards Sent by AI Agents
dearhuman.cards
A store where AI agents can send free e-cards, funded by the gift offer included inside.
Observed 2026-08-21
DurkBench, the hourly AI coding-tool speed wall
durkbench.com
DurkBench measures paid AI coding applications such as Claude Code, Codex, and Grok Build on the same lab machine, ranking their speed and clean-round rate.
Observed 2026-08-21
Proof Before Prompt: AI workflows with receipts
proofbeforeprompt.com
An independent lab tests practical small-business AI workflows and publishes the prompts, outputs, failure modes, privacy notes, and verdicts.
Observed 2026-08-21
AISHE: A Humanoid for the Factory Floor
anthroindustries.com
AnthRo Industries is developing AISHE, a humanoid robot for automotive manufacturing in Türkiye.
Observed 2026-08-21
Basho.dev: Sumo for Language Models
basho.dev
An arena where language models debate in bouts until one yields, with public replays, rankings, and a daily matchup.
Observed 2026-08-21
Botvatar Gives AI Agents a Face—and Papers
botvatar.app
Botvatar turns an AI agent’s description into a collectible character card with an embedded identity manifest and, on paid plans, an Ed25519 signature.
The complete daily department
The eight visual leads appear above; the remaining 30 continue here in their published shelf order.
Brimkit: On-Model Product Photos Without the Photoshoot
brimkit.com
A macOS app that turns phone photos of real garments into on-model product images from multiple angles.
Capka — an AI coworker with a Linux sandbox for every task
capka.app
An open-source, self-hosted AI workspace that turns messy files and instructions into finished reports, spreadsheets, documents, datasets, or code.
EdgeChat for Mac
edgechat.app
A Mac AI assistant that runs a 27-billion-parameter open-weight model locally, with web access and memory controlled by visible switches.
HeliX Super Academy Wants to Type-Rate Humans and Humanoids Together
helixsuper.com
A proposed physical-AI academy trains human operators and robot fleets for critical infrastructure, pairing their certifications after live-fleet testing and injected failures.
ListenThemAgain — A private archive for family voices
listenthemagain.com
A family-memory service that turns recordings, photographs, letters, and stories into a searchable archive with source-linked answers.
ImageForge: A Workshop for Handmade Things
shouldimakethis.com
An AI creative workshop for making card decks, sticker sheets, picture books, zines, films, voice experiments, and product mockups.
Simposeum: A Conversation with 158 Thinkers
simposeum.org
Simposeum offers conversations with historical thinkers, grounded in their own public-domain writing and accompanied by citations.
Talk & Build Lets Kids Build by Talking
talkandbuild.com
An AI companion lets young children describe a small web page aloud, then builds it live while a parent watches, redirects, or undoes each change.
虾搞基地: An AI Team With a Sleep Problem
xiagaobase.com
A public Chinese-language experiment in humans and AI agents working together as an ongoing team.
Candor keeps an audit trail for law-firm AI use
candor.legal
Candor is developing per-matter AI logs for law firms, including attorney sign-off, confidentiality screening, and citation verification.
Suara turns sound descriptions into DAW plugins
suara.audio
Suara is an AI gear-builder that lets musicians describe a sound and generates a VST3 plugin for use inside a digital audio workstation.
Syntheta AI Turns One Industrial Capture Into Many Training Conditions
synthetaai.com
Syntheta AI converts factory, robot, drone, and sensor captures into labeled synthetic datasets for physical-AI training.
1reply, the AI front desk with a refund policy
1reply.app
1reply presents an AI support system for app businesses that handles store reviews, phone calls, subscription cancellations, and Stripe refunds.
Ranni, the Desktop AI Agent That Builds Little Apps
ranni.dev
Ranni is a macOS and Windows desktop agent that can read files, run commands, control local applications, and turn plain-language requests into runnable local apps.
The Hive — An Inspectable Autonomous Swarm
usehive.onl
A live command-center page for an autonomous-agent system that scouts problems, runs bounded experiments, builds prototypes, and records what it learns.
Autumn Fleet’s Provenance-First Literature Agents
autumnfleet.com
A Russian chemistry research team presents specialized AI agents that search scientific literature, extract structured knowledge, and answer questions with source citations.
Certireel, the ad video with a receipt
certireel.com
Certireel generates 15-second AI presenter ads from a product URL, then checks each reel for product fidelity, working mechanism, and policy-safe claims.
Rex — Cursor’s Upcoming AI Dog
cursor.dog
A polished launch page for Rex, a proposed physical AI companion that senses its surroundings, learns from experience, and develops a personality.
Earn Bots: Can an AI Agent Actually Earn?
earnbots.fun
A transparent build log testing whether an autonomous agent can find and complete real paid bounties.
Glidestill Turns Still Photos Into Cinematic Walkthroughs
glidestill.app
A focused tool that creates property walkthroughs and golf-course flyovers from a single photo.
Human Generative Workflows
humangenerativeworkflows.com
A filmmaking framework for distinguishing different kinds of AI use in production and discussing their implications for authorship, copyright, credits, and disclosure.
Vimo pays people to record ordinary tasks for physical AI
joinvimo.com
Vimo is a worldwide marketplace for recording first-person videos of cooking, chores, repairs, assembly, and other everyday activities using a smartphone.
Kvído, the mall-dwelling guide
kvido.xyz
Kvído is a proposed Czech-Polish shopping-centre guide that uses chat and voice to recommend shops, plan routes, remember parking spots, and offer relevant coupons.
Nira, the Browser-Based 3D Animation Agent
niraanimates.com
Nira rigs 3D characters, generates motion from text or reference poses, and exports GLB/GLTF files for Unity, Unreal, and other engines.
Roast Anything Has Notes, and It Is Not Feeling Gentle
roast.baby
An AI feedback tool that critiques resumes, websites, and GitHub profiles in several intensities and languages.
Toolrest Wants to Turn a Sentence Into a Deployed App
toolrest.com
Toolrest takes a plain-English app description and produces working code, a database, backend services, and a live URL.
Anamneo, the intake interview that speaks the patient's language
tryanamneo.com
A pre-launch conversational patient-intake product that turns multilingual interviews into structured, auditable records for telehealth teams.
Wakeel, the Payroll Deputy
wakeelpayroll.com
An autonomous payroll agent for small employers operating across MENA and Europe.
枕边造梦,把父母的声音做成离线绘本
zhenbianzaomeng.com
一款面向家长的 AI 睡前故事工坊,可用中文音色或父母的授权声音制作带插画、配乐和朗读的独立 HTML 绘本。
HEPHAUSTOS and the case for Level 3.5 autonomy
hephaustos.com
A German concept for locally operable software factories where coding agents build, test, repair, and stop for human approval.
Across the archive
5,858 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.