The top trending AI agent projects this week span 20 tools: WikiFix, Thinking Orbs, Celio, Sente, Rival Workshop, Sorcerer, qarunbook, ZooWork, eu/jev, Notchware, Crowny, Agent Activity, Kilo, Singularity, Thanor AI, Sapien, Yubi, Gauth, Veltrix AI, and Cue by Manus.
Agent Activity is a macOS app from LeMax, LLC that presents AI-agent sessions inside Xcode alongside build timelines and test results in one window. The video also describes a menu bar panel that counts running builds and agents waiting for user input. The app runs locally, requires no account, and has no server of its own. It does not send project names, paths, transcripts, messages, or build-log content from projects; optional usage-event and crash-report sharing is controlled by separate settings and can include limited diagnostic details. The listed requirements are macOS 26 or later, and the page lists a price of $9.99.
Crowny is a native MacBook notch app that puts everyday tools and AI coding-agent activity in a single glanceable interface. It displays the status of Claude Code, Codex, and Grok sessions, shows which agent needs attention, and lets users approve or deny commands from the notch while tracking Claude usage limits locally without an account. The app also provides access to music, timers, calendars, reminders, clipboard history, files, pinned sites, games, weather, water reminders, and focus controls. It is sold as a one-time purchase rather than a subscription.
CUE is a language used to develop Crossplane composition functions independently of the final YAML composition. In the workflow described, CUE functions are embedded into YAML at build time.
Cue is a standalone personal-agent app from Manus. It gives each agent its own email address, phone number, wallet, and computer so it can send messages, answer calls, make payments within a budget, and work with other agents as a team.
Dzine is an AI creative workspace for generating, editing, and developing images and videos for individual creators and creative teams. It provides image and video generation, enhancement, lip sync, text-to-speech, generative fill, background and object editing, style transfer, image-to-3D conversion, and related creative tools in one workspace. Its production workflow includes an asset library, script analysis that turns scripts into structured scenes, and image, video, and audio generation nodes. Teams can share projects and assets, manage members, seats, credits, and workspaces, and collaborate in real time. Dzine also records references, prompts, model choices, edits, approvals, and contributors; its copyright tools flag potential copyright, likeness, trademark, and similarity risks. It is available through free and paid plans, with credits used for AI generations and other creative operations.
ElevenLabs is an AI voice and audio platform offering text-to-speech, voice generation and cloning, speech-to-text transcription, dubbing, music generation, and voice agents. Its transcription capability produces timing information for video cuts, subtitles, and synchronized animations. It provides more than 5,000 voices across more than 90 languages, supports custom voice clones, and offers secure APIs and SDKs through ElevenAPI. ElevenCreative supports content creation, while ElevenAgents supports customer-experience applications.
eu/jev is an inference API by Bevel for a System One decision model. It turns application context into structured choices, probabilities, and scores, supporting tasks such as routing, sentiment analysis, content moderation, text classification, and option comparison through a single endpoint. It provides a playground and a Jev-style API compatible with requests using the `jeff-latest` model. Requests are processed on a Hetzner server in Germany without an intermediary proxy, CDN, or external model API. The service states that prompts and answers are processed in Germany and not stored, with GDPR documentation including a data-processing agreement and subprocessor list. The page lists input-token pricing, free output tokens, and €5 in credits for new accounts.
Gauth is an AI-powered learning platform with interactive courses and AI-guided practice. Its science curriculum teaches physics, biology, chemistry, and medicine through visual explanations and interactive models, covering topics such as sound waves, cells and organelles, molecular structures, heart circulation, and lung anatomy. The platform also lists math and history curricula alongside its science courses.
Kilo is an agentic fitness coaching service delivered through iMessage. It connects a user's goals, training program, workouts, logs, and check-ins to determine the next workout and adjust the program when training or life circumstances change. Its stated workflow is to define goals and constraints, deliver workouts, record completed work and changes, and update the plan using that context. Kilo is offered as a beta and is onboarding lifters in small groups.
ManuAGI & AutoGPT Tutorials is an email newsletter covering AI, AGI, and AI-agent project updates. Its subscription page says it provides new project updates every week and is published through beehiiv.
Notchware is a native macOS app that turns the MacBook notch into a live island for Now Playing controls, battery, volume, brightness, Focus, Wi-Fi, and internet-status indicators. Its Pro edition adds local alerts for Codex, Claude Code, and Cursor when an agent needs approval or finishes a task, with optional sounds, plus Lock Screen music controls and profile hiding. It runs locally without an account, analytics service, or cloud backend; its agent integrations use inspectable local hook commands that write status events without reading prompts or responses. Notchware supports macOS 14.6 and later on Apple silicon and Intel, with a free edition and a one-time $3.99 Pro purchase.
qarunbook is an app-testing runbook that helps teams create, execute, and track test plans for web, Android, and iOS applications. Its AI can draft checks from an app's code, while people, teammates, or connected AI assistants can work through the plan and record results. The service keeps issues connected to the checks that found them: raising an issue removes the check's pass status, and a fixed issue must be checked again before the test passes. It supports custom plans, platform-specific checks, project access controls, team join links, and integrations including GitHub, GitLab, Linear, Jira, Slack, Microsoft Teams, Sentry, PostHog, email, webhooks, and Zapier. qarunbook connects to AI assistants through MCP. Its documented commands include `/test-checklist`, which drafts a plan, and `/qa-runbook`, which works through web and mobile checks, records results, and follows up on fixes. The service has a free plan and paid flat-rate workspace plans, with API access and integrations available according to the plan.
Rival Workshop is a downloadable kit of 150 editable instruction files, or skills, for AI coding tools including Codex, Claude Code, and Cursor. The skills cover work across 22 disciplines, including product strategy, accessibility, design, frontend and backend engineering, data, AI engineering, security, testing, delivery, marketing, and operations, and are supplied with working examples and an installer for adding them to a project. Its companion app lists the skills, supports search and filtering, shows which skills an agent loads, and lets users enable, disable, rewrite, or add skills from the web. The included Brief tool turns a code change into a visual report with source references for review before merging. The download also includes an optional image helper; the app requires Node 20 or newer, while the image helper requires Python 3.10 or newer and a compatible image-provider key. The product is sold as a one-time purchase with future Workshop editions included, and the skills can be used and edited in personal or client projects but not redistributed as a competing collection.
Sapien is an AI-powered market research platform that builds synthetic populations—simulated audiences grounded in purchase histories, behavioral and research data, demographics, surveys, and public conversations. It uses these populations to run surveys, qualitative interviews, and comparison tests for products, prices, and campaigns, with options to add customer data and past research for context. Sapien returns interactive reports with predictions, segment comparisons, recommendations, and the responses behind the findings. Users can ask follow-up questions of the same simulated audience; the service states that its team builds the audience, runs the study, and provides same-day predictions and interpretation.
Sellio is an AI customer-support platform with a shared inbox for website live chat, WhatsApp, Instagram, Telegram, and email. Teams can install a branded chat widget with a single script snippet, keep replies, notes, assignments, tickets, and handoffs in one conversation thread, and mirror new conversations to Slack, Discord, or Microsoft Teams. Its optional AI agent is trained on a company’s website, documentation, and FAQs, answers website visitors, and hands conversations to human agents without losing context. Sellio also includes automations and analytics for tracking conversations, resolution, response time, and customer-experience metrics.
Sente is teai.io's coding-agent CLI, distributed as a thin launcher over the MIT-licensed OpenCode project rather than as a fork. A one-line installer installs OpenCode when needed, points it to a teai.io-generated configuration, and synchronizes teai.io's model catalog so models can be selected from the terminal; Sente reads and edits repository files, runs commands, and reports results. It supports text and voice interaction, with Sente Cloud providing a browser-based workspace that continues running when the user's computer is closed. Optional local PII scrubbing masks items such as email addresses, phone numbers, addresses, API keys, private keys, and configured names before requests are sent. The CLI runs on macOS, Linux, and Windows through WSL, and the service offers Japanese and English interfaces.
Singularity is an Agentic Development Environment from Meteor Factory for building and maintaining software with AI agents. It combines a code editor, terminal, Git, ticket board, and agent orchestration in a desktop application, with support for features, bug fixes, refactoring, and multiple projects. Work is organized into atomic tickets with a limited scope, dedicated context, and isolated Git worktrees. Relevant skills, conventions, and files are injected into each ticket; agents can work in parallel, while users inspect changes and review integrations. The application also includes Git branches, graphs, stash, hooks, review tools, native macOS completion or failure notifications, and a DevOps profile that can detect failed CI pipelines, create repair tickets, and relaunch deployments. It supports Claude, Codex, GitHub Copilot, and Gemini, according to the site. The free application includes agents, projects, worktrees, language servers, and the Notes plugin without a subscription. A paid Singularity Licence enables marketplace plugins, including the Pipelines CI/CD plugin, after a trial period. The product is described as privacy-first and self-hostable: code, tickets, and contexts remain on the user's machine, while only API calls to language models leave it. It is available for macOS and Windows, with Linux packages also published.
Sorcrr is a bounty-driven, video-first hiring platform that combines referral-based recruiting, structured video profiles, collaborative candidate evaluation, and AI-assisted interviews. Employers create video job descriptions, attach referral bounties, and share roles with their networks; candidates introduce themselves through short videos, while teams can filter, rank, vote, comment, and communicate in one workspace. Its Career Rank Score uses ratings, reviews, and referrals to build a multidimensional candidate profile, and its SAI assistant supports interviews, candidate breakdowns, job-description drafting, salary benchmarking, customer-support drafts, and company knowledge management. The platform describes a one-time fee charged to both the hiring company and hired candidate, with part of the fee allocated to the referral bounty, and currently invites users to join a waitlist.
Thanor is a library of complete design directions for AI-built websites, covering visual elements such as palettes, type scales, motion curves, and page structure. Users browse a direction, copy its single prompt, and paste it into tools such as Claude, ChatGPT, Cursor, v0, Lovable, or Bolt; the library can also be connected to editors through MCP. The page says that new directions are added weekly, with premium sections and a page-cloning tool listed as upcoming features.
Thinking Orbs is an open-source React component library of animated orb status indicators for AI applications. It provides states such as reasoning, searching, working, retrying, compacting, and waiting, with configurable shapes, rendering methods, size, speed, density, tilt, pausing, labels, and CSS tinting. The library can be installed from npm or copied with shadcn; it also provides a plain JavaScript version that requires no dependencies. Optional shapes and rendering methods are separately imported so only the selected components are included in the bundle. The project is MIT licensed and is described by the video as being under 4 KB gzipped.
Veltrix is an AI analytics and business health-check service for e-commerce businesses. It connects to stores, accounting, payment, and marketing tools, including QuickBooks, Xero, Shopify, Square, and HubSpot, or accepts PDF, Excel, and CSV uploads. It examines the supplied business data for risks, opportunities, cost problems, and strengths, then presents each finding with what was found, why it matters, and what to do next. Connections are read-only, so Veltrix does not edit, post, or delete data. The service also provides separate read-only workspaces for accountants and bookkeepers managing multiple clients. The first business health check is free and requires no payment card; paid plans are monthly.
WikiFix is an AI-powered Confluence app that monitors selected spaces for outdated, contradictory, duplicate, or ownerless pages. It reads page content rather than filtering only by dates, compares conflicting claims side by side, and presents findings for review. An administrator chooses the authoritative claim; WikiFix then proposes the edits, requires approval before writing, and allows content changes to be reverted in one click. It can also reassign abandoned pages or post an inline comment mentioning the page owner and space administrators. WikiFix runs inside an existing Confluence installation and is billed through the Atlassian Marketplace, with payments converted into scan credits for new or changed pages.
WikiFix is a Confluence app for administrators and knowledge leads that scans selected spaces for outdated, contradictory, duplicate, and ownerless pages. It reads page content to detect conflicting claims, presents the relevant passages side by side, and lets a reviewer choose the claim to keep before applying the change. WikiFix writes nothing without approval; content fixes can be reverted in one click, while reassigning an abandoned page owner is identified as a non-reversible action. It runs inside an existing Confluence installation, can notify page owners through inline comments, and scans new or changed pages using credits billed through the Atlassian Marketplace.
Yubi is a Mac voice assistant that types spoken words wherever the cursor is and helps locate on-screen settings. The user holds a key, speaks, and releases it; transcription runs locally on the Mac, so the voice recording is not uploaded. It supports English and Japanese.
ZooWork is an AI agent delivery platform for forward-deployed engineers and domain experts. It packages a role's knowledge, industry context, standard operating procedures, delivery standards, and tool permissions into Agent Skills, then uses them to build agents that plan and execute work, call connected systems, request human approval before sensitive actions, and deliver outputs such as drafts, documents, orders, and system updates. Its Agent Builder turns a conversational specification into a deployable agent, while the ZooWork Agent Runtime handles planning, task breakdown, execution, guardrails, feedback-based skill improvement, isolated sandboxes, autoscaling, GPU-on-demand execution, and observability. It connects to services including Google Workspace, Microsoft 365, Slack, Lark, Jira, and GitHub through MCP-compatible integrations, and agents can be used from ZooWork, Slack, Microsoft Teams, WhatsApp, or embedded applications through an API and SDK. The platform also includes ZooData, which converts URLs into structured, agent-ready data. The page describes tenant isolation, encryption by default, zero data retention, no training on customer data, role-based access, runtime checks on inputs, outputs, and tool calls, approval gates, and replayable audit records. ZooWork offers forward-deployed support for scoping and taking a workflow into production through a stated seven-day pilot process.
Searchable transcript of Top AI Agent Projects : WikiFix, Sente, ZooWork, Singularity & Yubi — ManuAGI - AutoGPT Tutorials (19:57). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by ManuAGI - AutoGPT Tutorials. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 Everyday developers build new AI tools and keeping up with all of them takes time. This is our weekly project update video where we cover the top trending AI agent projects this week so you can find what is worth using in one place. You will discover trending AI tools and projects across coding, testing, and learning. Without wasting time, let's get started.
00:21 >> Before we jump into today's project updates, here's a quick announcement for everyone. We've launched a brand new YouTube channel called AI Agent Studio dedicated entirely to AI Agent projects, tutorials, and tools. So, if you're interested in staying up to date with the latest AI agent open source projects, learning how to build your own agents, or exploring cuttingedge agent frameworks, make sure to check it out.
00:47 Subscribe now to get weekly videos, in-depth guides, and real time project breakdowns. The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, Wiki Fix for Confluence. Finds and fixes wrong Confluence pages. WikiFix is a Confluence app that finds the outdated and contradictory pages in your wiki and fixes them in a few clicks.
01:09 It solves the quiet drift that leaves run books wrong in an incident. Setup guides broken for new hires and co-pilots like Rovo repeating your doc's mistakes at scale. It watches the spaces you choose and reads the content, not just dates, flagging pages that disagree and laying each claim side by side, quoted from its source, so you settle which is right.
01:30 It also catches duplicates and pages whose owner has left. You pick the correct answer. It writes the fix and nothing changes until you approve with content fixes reverting in one click. It runs inside your existing Confluence with no migration and can use your own anthropic key. It is built for Confluence admins and knowledge leads. Install it from the Atlassian marketplace.
01:51 Project number two, Thinking Orbs, animated status indicators for AI interfaces. Thinking Orbs is an open-source React component library of animated status indicators for AI apps. It solves the small but repeated job of showing what an agent is doing so builders don't handcraft a thinking or searching animation from scratch. You install it from npm or copy the component in with shad CN.
02:14 Then drop in an orb and give it a state like reasoning, searching, compacting, retrying, or waiting, plus an optional variant. Props control size, speed, shape, density, tilt, and color. And opt-in shapes and renders keep your bundle small since only what you import ships. The core needs nothing but React, weighs under 4KBGzipped, and a plain JS build needs no dependencies at all.
02:38 It is MIT licensed and free. It is built for developers making AI interfaces. Install it from npm and drop one in. Project number three. Celio one shared inbox with optional AI. Celio is an AI customer support platform that brings every channel into one shared inbox. It solves the scatter of answering customers across live chat, WhatsApp, Instagram, Telegram, and email in separate places.
03:05 You add one snippet to your site for a branded chat widget, and all channels land in a single inbox where your team assigns notes and resolves conversations, raising a ticket when work continues. You turn on an AI agent only when ready, trained on your site, docs, and FAQs so answers stay grounded, handling the first reply, and handing off to a person with full context when needed, so you control cost.
03:30 Analytics tie each conversation to how it ended and how it felt with topics, response time, and CS8. The inbox, widget, and channels are free. It is built for stores, hotels, SAS, local teams, and agencies. Sign up free with no card. Project number four, Cente Coding Agent CLI. 380 plus models. Cente is TA.io's coding agent CLI that turns every model on the platform into an agent in your terminal.
03:58 It solves the setup and lock in of coding tools tied to one provider installing in a single curl command. It is a thin launcher over the open- source open code, not a fork. So, it inherits every upstream improvement and speaks OpenAI compatible natively with no conversion layer. It reads and edits files in your repo, runs commands and reports back, and you switch among over 380 models with one line to balance cost and quality.
04:24 You hand over a task by voice with TTalk and hear the reply ridback and Centecloud keeps work going in the browser while your Mac sleeps. A three- tier safety model asks before anything risky. Failed responses cost nothing and opt-in local PII scrubbing masks sensitive data before requests leave your Mac. It runs on Mac OS, Linux, and WSL. It is built for developers.
04:49 Install it free with 100 credits. Project number five, Rival Workshop. 150 editable skills for coding agents. Rival Workshop is a kit of editable skills, examples, and image tools for your coding agent. It solves the gap where agents design and build, but leave the work around the screen like tests, security review, launch copy, and pricing undone. It gives your agent 150 skills across 22 disciplines.
05:16 Each an instruction file you can read and rewrite, plus editable example projects and image and motion helpers. A companion app shows every skill in your repo as books on a shelf, so you see what each agent loads, switch skills on and off, and add good ones from the web with an installer that drops them into your project folder. An included tool called Brief turns a code change into a visual report with source references, so you review what the agent changed before merging.
05:42 It works with clawed code, codeex, and cursor on your own AI subscription. It is built for people shipping products with agents. Try the free sample. Project number six, Sorcerer Bountydriven video first AI hiring platform. Sorcerer is a hiring platform that combines video referral bounties and AI to help teams build great teams faster. It solves the slow resumebound hiring process where job posts and LinkedIn profiles tell only part of the story.
06:10 You create a video job description, attach a referral bounty, and share it so quality video referrals roll in from your network and sorcers with people rewarded for connections even when the hire lands elsewhere. Candidates introduce themselves in short videos and a career rank score built on real collaborator feedback shows how they stack up. Its woven in AI SAI runs real-time assisted interviews suggesting follow-ups.
06:35 Breaks down candidates afterward and drafts job descriptions, salary benchmarks, and knowledgebased work in the background. Pricing is a one-time fee, half return to the community and referers. It is built for founders and hiring teams. Sign up free. Project number seven, Mikorin Book AI written test plans your team runs. The Cororin book is an app testing runbook that turns your code into a clear test plan your team or AI works through.
07:02 It solves the hardest part of QA. Knowing what to test, which can become a bigger job than the testing itself. You run a command and your AI drafts the checks from your app, each with steps and an expected result. Then you edit them, add your own, or start from scratch. Your team, you, or a connected AI works through the plan on web, Android, and iOS, keeping the checks and results in one place, and an issue stays attached to the check that found it, taking the pass away until someone verifies the fix.
07:31 It connects over MCP in about a minute, so Claude, Codeex, or Cursor reads the plan, records results, and follows up on what it fixes. It is built for developers and QA teams. Start testing free. Project number eight, Zoo Work. Deploy your expertise as AI agents. Zoowork is an AI agent delivery platform for domain experts and forward deployed engineers.
07:55 It solves the gap where capable models still start from a blank prompt instead of knowing your job. You package your knowledge, methods, SOPs, delivery standards and tool permissions into agent skills. So an agent starts with the full job description and runs your playbook around the clock, improving with every run. It plans a task into steps, reads and writes your real systems, pauses at approval gates before anything moves money or data, and delivers real work like decks, documents, and orders rather than just
08:25 answers. It runs on every Frontier model with no lockin. Connects over MCP to Google Workspace, Microsoft 365, Slack, Jira, and GitHub. And you build an agent in chat and ship it in one click. Agents work in Slack, Teams, and WhatsApp like co-workers. Tenant isolated with no training on your data. It is built for enterprises and domain teams. Get started.
08:47 Project number nine. U/Jev system one decision model hosted in EU/JEV is an EUh hosted inference service for a system 1 decision model built by Bevel. It solves the problem of running a fast decision model while keeping data inside Europe under GDPR. Instead of generating pros, it turns context into choices, probabilities, and scores through one API.
09:12 So you classify text, compare options, and return structured decisions for routing, sentiment, or moderation. You give it a state and a question, and it answers as a choice, a yes, no, or a score, ready for your application. requests go straight to a Hzner server in Germany and are answered there with no proxy or external model API in between and prompts and answers are processed in Germany and not stored with a DPA and subscessor list one click away.
09:39 It is API compatible with Jev style requests and output tokens are free. It is built for EU developers. Get a free API key with 5 hours of credits. Project number 10, Notchwear, your coding agents in the notch. Notchwear is a native Mac app that turns the MacBook notch into a live island with coding agent alerts built in. It solves the constant terminal checking and status hunting that pull you away from work.
10:05 Around the notch, it shows now playing with controls, battery, volume, brightness, focus, and network status. Each indicator one you can turn off and it gets out of the way or hides itself in full screen. Its pro tier tells you in the notch the moment codecs, clawed code or cursor needs approval or finishes a task with optional sounds and adds lock screen music controls.
10:26 Alerts use local hook commands you inspect before installing which writes small status events and never read your prompts or responses and it runs locally with no account analytics or cloud. It is built for Mac users who run coding agents. Download it free for Mac OS. Project number 11, Gown Beand, your day and agents in the notch. Gown is a native Mac app that turns the MacBook notch into a live home for your day and your coding agents.
10:54 It solves the constant app switching and terminal checking that pull you out of focus. It keeps music, timers, your calendar, reminders, clipboard, files, water nudges, and games one glance up and gives each claude code, codeex, and gro session a little face showing what it is doing and which one needs you. When an agent asks to run a command, you approve or deny it from the notch without finding its terminal, and it tracks your claw usage limits.
11:20 It watches sessions you already run in terminal, eterm, ghosty, or warp with nothing to wrap. Holds pings during focus and you hold space to ask it a question by voice. Everything stays on your Mac with no account. Voice and mail triage. Use your own anthropic key. It is built for developers on a Mac. Buy it once. Project number 12. Agent Activity. Mission control for Xcode AI agents.
11:46 Agent Activity is a Mac app that shows what your AI agents do inside Xcode. It solves the blind spot of running coding agents in Xcode where their work scrolls past and you cannot easily see what changed or whether it worked. It brings live agent sessions, build timelines, and test results into one window. So, you watch an iteration as it happened from a failed build through the errors and fixes to a green one.
12:08 A menu bar panel counts how many builds are running, which agents are working, and which are waiting on you. It has no account and no server of its own. No transcript, build log, project name, or path leaves your Mac, and it asks once before sending optional independently toggled usage and crash diagnostics. It is built for developers who run AI agents in Xcode.
12:31 Get it on the Mac App Store. Project number 13, Kilo, Agentic Fitness Coach in iMessage. Kilo is an Agentic Fitness coach that works entirely through iMessage. It solves the gap between tools that only log workouts and coaching that keeps a plan pointed at your goal without the cost of a human coach or the forgetfulness of a chatbot. You set a goal and your constraints and Kilo delivers the next workout right in iMessage.
12:56 You record the work and what changed and it adjusts the program from your real training history when life or progress shifts because it keeps a current program, follows up on the goal and works from what you actually did. It acts as an ongoing coaching relationship rather than a static app available without an appointment. It is onboarding serious lifters in small groups.
13:16 It is built for people who train with a goal. Join the beta. Project number 14. Singularity. Agentic development environment driven by tickets. Singularity is an agentic development environment from Meteoractory where AI agents build features, fix bugs, and refactor code in parallel. It solves the black hole chat loop where context saturates. The agent drifts and it changes far more than you asked.
13:41 Instead of an endless conversation, work is a ticket. One scope, one clean context, one agent, one result. You describe a feature or bug and agents work in isolated git work trees while you inspect the changes and approve the merge with an editor, terminal, git, and a conbon board in one desktop app. Each ticket loads only the relevant skills, so context stays minimal, and Mac OS notifies you when an agent finishes or fails.
14:09 It orchestrates many projects at once, supports Claude, Codeex, Copilot, and Gemini with your own key, and is privacy first and self-hostable, so your code stays on your machine. It runs on Mac OS, Windows, and Linux. It is built for developers and teams. Download it free. Project number 15, Thanor AI. Design directions for AI built websites. Thanor is a library of design directions for websites built with AI.
14:34 It solves the generic templated look that AI coding tools produce where the output reads as generated rather than designed. Each entry is not a template but a complete art direction covering pallet type scale motion curves and structure. Written the way a season designer would as a single prompt. You copy whole. You paste one into claude, chat, GPT, cursor, vzero, lovable, or bolt, and what comes back looks designed.
15:00 You browse by category like hero, landing page, SAS, agency, and portfolio, and new directions land every week, or you connect Thanor over MCP to pull a direction without leaving your editor. It is built for people building sites with AI. Open one and paste it into your tool. Project number 16, Sapion, market research with synthetic populations. Sapion is an AI market research platform that simulates populations to predict how people respond to products, prices, and campaigns.
15:30 It solves the cost and slowness of traditional research where testing an idea with real audiences takes weeks. It builds synthetic populations, simulated audiences grounded in purchase histories, behavioral data, demographics, surveys, and public conversations with your own customer data adding context. You bring a question and it runs surveys, qualitative interviews, and comparison tests against the same audience, returning an interactive report with results, segment comparisons, and recommendations you can probe with
16:01 follow-up questions. It covers product concepts, messaging, pricing, segmentation, and market strategy. And its team builds the audience, runs the study, and helps interpret it with same-day predictions. It is built for product, marketing, and strategy teams. Book a demo. Project number 17, Mashion Yubie. Talk to your Mac, she types. Yubie is a Mac voice assistant that types what you say and helps you find things on screen.
16:25 It solves the slowness of typing and the hunt for a setting you cannot locate. You hold the right option key, talk, and let go, and your words land wherever the cursor is across Slack, Mail, Notes, Messages, Calendar, and more in English or Japanese. Holding the right command key asks for help with the screen and she points at the button you cannot find like dark mode in system settings.
16:49 Transcription runs entirely on your Mac so no audio is uploaded and the whole app is tiny with no screen recording or screenshots. Dictation is unlimited and free with paid tiers adding more of the screen help tasks. It runs on Apple silicon. It is built for anyone who would rather talk than type. Download it free for Mac. Project number 18. in Goth Unlimited digital canvas AIG guided interactive science courses.
17:13 Goth Academy is an AI powered learning platform whose science track teaches concepts through interactive lessons and AIG guided practice. It solves the gap between reading about a concept and actually understanding how it works. Where static textbooks leave the mechanism abstract, you pick a topic and work through visual interactive models that build the idea step by step with AI guiding the practice as you go.
17:36 The science catalog spans physics explainers on everyday phenomena, biology from cells to photosynthesis, chemistry across atoms and periodic trends, and medicine covering heart and lung anatomy alongside math and history tracks, and you can also create your own course. It is built for students learning science. Project number 19, Veltrix AI for e-commerce.
17:58 Find what's hurting your store's profit. Veltrix is an AI analytics tool that reads your e-commerce business data to find ways to improve sales, cut costs, and fix problems. It solves the guesswork of ordinary health checks, where you fill in a long form and get a score based on what you typed, not what is actually happening. You connect your store, accounting, payments, or marketing tools, or upload PDFs, spreadsheets, and CSVs, and in about 2 minutes, it examines your records and surfaces the issues and strengths you
18:28 may have missed. Every finding comes in three parts. What it found, why it matters, and what to do next with the source behind each answer shown, and a note when it does not know. Connections are readon, so it never edits or deletes anything, and you can disconnect any tool. It works with QuickBooks, Zero, Shopify, Square, and HubSpot, and accountants can run it per client in separate workspaces.
18:52 It is built for e-commerce founders and bookkeepers. Start your first health check free. Project number 20, Q by Manis. Personal agents with their own identity. Q is a standalone app from Manis for creating personal AI agents, each with its own identity. It solves the limit of assistants that act only through your own accounts, giving every agent its own email address, phone number, wallet, and computer, so it can work beyond a chat window.
19:17 You hand an agent a task and it picks the right tools, sending messages and emails from its own address, answering phone calls and leaving you a summary and paying for things within a budget you set with you keeping the final decision. Each agent finishes work on its own machine and several can work as a team in a group chat delegating to one another.
19:36 Built on the same infrastructure as Manis and running on its cascade harness, a purchasable cloud computer keeps jobs going while your laptop is closed. It is an early access on web, desktop, and mobile. It is built for anyone delegating everyday work. Request an invite and create your first agent. Thanks for watching. See you in the next video.