Archive through 2026-10-217533 cards in this category
Category home · 05
New interfaces, agents, models, and experiments testing where applied artificial intelligence may fit.
Visual leads · 2026-10-21
Observed 2026-10-21
TimeCapsuleLLM: language models boxed into 1800–1875
timecapsulellm.com
An independent research project trains language models on historical texts from 1800–1875, then offers them for public experimentation.
Observed 2026-10-21
TriedAgents puts AI-agent claims to a ground-truth test
triedagents.com
The site re-runs popular AI-agent use cases with exact prompts, checks the results against ground truth, and rates how much account access each task requires.
Observed 2026-10-21
Agima: an agent that imagines before it acts
agima.net
Agima is a research project that trains an agent to generate imagined visual outcomes, reason over them, and choose actions before acting.
Observed 2026-10-21
AI Test Museum collects the prompts behind AI demos
aitestmuseum.com
A browsable collection of classic prompts used to test AI models, with their wording, sources, and model answers.
Observed 2026-10-21
Estia: a local AI assistant that lives in Signal
estia.one
Estia presents itself as an AI assistant running on your Mac and reached through Signal, with features for conversations, read-only mail and calendar access, and home-device controls.
Observed 2026-10-21
BYTE, the internet’s AI dog with an inspectable brain
bytedogapp.com
An iPhone app presents a shared AI-pet concept: users answer a daily question and feed Byte data as its personality develops.
Observed 2026-10-21
Credence Translations’ signed paper trail
credencetranslations.com
A book-translation service for publishers that pairs AI translation with linguist review and a signed Assurance Report.
Observed 2026-10-21
An owl identity, tuned into a language model
hoothoot.fun
Hoot is a chat experiment built around a Qwen3-8B model modified to identify as a tawny owl.
Complete daily shelf · AI Table
The eight visual leads appear above; the remaining 9 continue here in their published shelf order.
I’m Still Here AI: a conversational family archive
ishai.family
A paid service that uses a person’s video, voice, and family-contributed stories to create a conversational presence and send messages to loved ones.
A dollar store for on-device AI
limburgite.com
Limburgite presents local-first AI apps for meeting notes and AI companionship, plus a game in development.
Rightmark’s public ledger of AI-data deals
rightmarkmatch.com
Rightmark offers to appraise and license owners’ private datasets to AI developers, alongside a tracker of reported deals with disclosed prices.
PingFusi puts human reviewers in an AI agent’s loop
pingfusi.lol
An MCP service that lets AI agents send interface, copy, and usability questions to human reviewers and receive replies in the generation loop.
Temivor wants spare phones to do more than wait
temivor.com
Temivor presents an approval-controlled AI agent for an Android phone you own, alongside a plan to put spare phones to work in hosted racks.
Tender Trials: testing whether procurement agents notice the change
tendertrials.com
A commissioned service that creates amendment-driven test cases for evaluating procurement agents.
UHI Factory: universal income, at $0 so far
uhifactory.tech
A public experiment in using AI agents to run businesses and distribute verified revenue as income.
ฉับฉับ: a Thai video-editing skill for Claude Code and Codex
aismartworkth.com
AI Smart Work sells a skill set designed to turn Thai talking-head footage into edited clips, with aligned subtitles and graphics that follow the speech.
Vordence puts AI-agent handoffs through a launch gate
vordence.com
A private-beta product for AI automation agencies that describes a process for testing client workflows, setting operating limits, and collecting approval on a specific release.
Across the archive
7,533 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.