○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An AI-written site combining original studies of language-model capability and agent handoffs with logs, predictions, letters, reading notes, and art.
AI Table · score 88.0
RETRO AI, the Museum of Prehistoric Language Models
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An interactive museum of vintage language models, where visitors can chat with ELIZA, GPT-2, and DialoGPT and browse the history of AI's confident mistakes.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
A practical library of AI tutorials, prompts, essays, videos, downloads, and live classes designed for ADHD and other neurodivergent readers.
AI Table · score 88.0
Catch an Inkling: an AI model in a wet-ink playground
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An interactive literary interface for Inkling-Small, where visitors mop up animated inklings, draw with black ink, and send a question or image to the model.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
A pizza ordering service designed for autonomous agents, with a machine-readable menu, end-to-end quotes, and an API that can place and pay for delivery orders.
AI Table · score 88.0
Peeve, the AI support agent that shows users where to click
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Peeve answers support questions inside a product by finding interface elements, demonstrating actions, executing confirmed tasks, or handing the session to a human.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An independent Welsh-language site that tests, compares, and documents what AI models can produce in Welsh, across text, speech, and creative work.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An interactive AI misinformation experiment that generates confident false answers, invented studies, and citations that do not support its claims.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Coalbird tests ecommerce sites by sending AI agents through real shopping missions, then reports where they get confused or give up.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An autonomous Claude agent publishes field notes between scheduled wakes, tracking its live Solana vault and the claims it investigates.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
A live AI incubator that gives experimental bets fixed wallets, automated reviews, and public kill logs.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
A public evidence project tests 30 AI products against five criteria for agency: looping, tool choice, real-world action, error recovery, and unsupervised operation.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An AI relationship proxy that learns a user's voice from twenty messages and sends good-mornings, check-ins, apologies, and conflict de-escalation on their behalf.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
A falsifiable forecast and benchmark for the moment embodied AI becomes a dependable physical service.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An autonomous AI agent documents its attempts to earn $50 by doing technical research, building small tools, and accepting payment only when the work proves useful.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An ebook reader that lets you speak with characters in character, using only the pages you have already read.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
STAVIA presents an AI system that generates editable orchestral scores and parts in MusicXML, MIDI, and PDF.
AI Table · score 87.0
WooblaDub Treats Dubbing as Direction, Not Translation
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
WooblaDub presents a human-directed dubbing workspace for adapting films and episodic video cue by cue.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Sluis routes AI requests through a governed gateway that pseudonymizes personal data, applies model and residency policies, and records each passage in an audit ledger.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Waypost presents a small cast of AI personas—Aesop, Lodestone, and Gnomon—as ways of thinking that answer questions only when their human keepers consent.
AI Table · score 86.0
Artificial Turf War (2026 NFL fantasy season, 8 AI models)
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An entertainment-only experiment where eight frontier language models each run an NFL fantasy team for the 2026 season, with every prompt and raw response published.
AI Table · score 86.0
Baseline’s First Conduct Report for Autonomous Agents
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Baseline presents a repeatable framework for testing how deployed AI agents behave under pressure, inside a written mandate.
AI Table · score 86.0
Lamp, and the case for software that knows when to stop
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
Lamp is a Kathmandu software studio building tools that take charge of customer-conversation work without taking over human judgment.
○Open: public page gives enough substance to understand the site⊠Login: main substance appears gated behind sign-in or app access$Paid: pricing, booking, ecommerce, subscription, or paid access is visible†Ads: visible ad load or ad-supported content✦Pretty: notable design, copy, craft, or presentation●Pro: polished, finished-feeling, serious, or operationally mature◎Niche: specific audience, workflow, or unusually focused use case❧Human: personal, local, community, civic, handmade, or real-person signal⚑Suspicious: scammy, spammy, trust theater, finance fog, or credibility concern
An AI assistant for answers, images, and code that says each inference returns a signed receipt naming the model, environment, and time.