
See all 5
- kit-15-comandos PT·EN·ES 10/19/2026
- camada-contexto PT·EN·ES 10/18/2026
- constituicao-agentes PT·EN·ES 10/12/2026
- sil-loop-r PT·EN·ES 10/05/2026
- loop-r 09/10/2026
The links to courses, projects and videos on this page are for INEMA.PRO subscribers.After signing in, come back to this page.
INEMA catalog
Everything that has been on INEMA.CLUB, with no repeats. Several on the same subject share one card: open it to see them all.
Nothing found for this search.




Knowledge base with critical analysis, glossary, and resources for planning, testing, and evaluating AI applications.
Local Python engine for planning videos with scenes, timings, and voice, image, and music guidance.
Local kit for building vertical videos with scenes, word-by-word captions, and synchronized audio. Canvas and FFmpeg engine, CLI with validation, and a complete 48-second example. Guide in PT/EN/ES.
Os 7 hábitos de quem usa IA no topo (falar em vez de digitar, IA ouvindo a reunião, mostrar em vez de contar, a IA te entrevista, pedir falhas, testar você mesmo, mente aberta) com kit aberto e grátis: 7 prompts, skills para o Claude Code, script de reunião local e guia de ditado. PT/EN/ES.
Local kit for getting through the next 2 years of AI across 3 horizons: 14-day kit calculator (food, water, money), AI installer and offline Wikipedia (Ollama + Kiwix, dry-run by default), MFA audit to move away from SMS, a personal eval with your real tasks and a scoreboard with a chart, and a 90-day plan for income and identity. Data from Stanford and France, no API key. Guide in PT/EN/ES.
Mirror of Shaders (MIT): over 200 WebGPU visual effects (living gradients, glass, metal, light, particles, cursor effects) as components for React, Vue, Svelte, Solid, and JavaScript. Guide with live demo, step by step, what's free and what's Pro, telemetry, and a plan B without WebGPU. Guide in PT/EN/ES.
Also: shaders — does the effect help or hurt? A 4-version test measures contrast and CPU
INEMA's viral reel: a thesis becomes a 30-to-40-second 9:16 reel with a 7-step trigger script (hook with a number through the ask to comment) and newscast-style polish — a title drawn at the opening, panels and a counter timed to the speech, karaoke captions, and Nei's avatar in the corner. Claude writes, Codex draws, HeyGen speaks, and the engine syncs; quality gates, delivery to Telegram, and a YouTube publishing queue. Guide in PT/EN/ES.
The manager already knows the hardest part: a guide to teaching the AI agent with the 5-4-3 rule (5 questions, 4 autonomy levels, 3 test cases), the 4 roles of the technician, and 5 copyable templates, plus video and reel (PT/EN/ES).
Cleaning up your Claude and Codex setup: audit prompt, a script that reads the history, a single AGENTS.md, folders with their own AGENTS.md, playbooks, and the 3-question filter, plus video (PT/EN/ES).

Git explained as a garden: clone, commit, branch, worktree, merge, and push, each with a drawing, ready-made commands, and a commit graph, plus video and reel (PT/EN/ES).
A room that brings you or your team, Claude Code, Codex, and whoever writes to your Telegram bot into a single conversation, with a live panel (Room, Graph, Flow). Runs locally or on the internet with an INEMA.PRO account login (live at salaconf.inema.pro). Person mode and support mode, in which the AI drafts and a human approves before replying to the customer. Agents via subscription, no API. Guide in PT/EN/ES.
LongCat-Video mirror (Meituan, MIT): an open 13.6B model that generates minutes-long 720p video at 30 fps from text, an image, or video, and Avatar 1.5, which makes a photo talk from an audio clip.


Concept document under discussion: what a system would look like in which people, AI agents (Claude Code, Codex), and Telegram bots share the same channels, seeing what arrives and talking to each other. It presents four possible forms (from the read-only panel to the full team), open technical foundations, six scenarios (local use, with guests, one department, a whole company, games, customers and suppliers), limits, security, agent personalities, and 17 cited sources. PT/EN/ES guide.
Python framework for companies to build AI apps from a business outcome: the LLM runs inside a fixed workflow, every external action waits for human approval, rules with numbers become tested code (not RAG), and every run is recorded in the audit log. It comes with a complete example app (manager coaching, 5 workflows, dashboard, and HTML report) and runs on just Python 3, with no key and no internet; Ollama optional. PT/EN/ES guide.
Catalog of 20 avatar video formats, each with a fixed name: in 16:9 Wide, Corner, Bubble, Newscast, Split, Presenter, Documentary, and Lesson; in 9:16 Strip, Corner, Bubble, Newscast, Presenter, Viral, Documentary, Half, Inverted, Question, Stacked, and Character. Each format has a real video example, shows where the avatar appears, and a command (compor.sh, ffmpeg only, no GPU). It also shows which formats each INEMA video project uses. PT/EN/ES guide.
INEMA fork of Hypit: the agent (Claude Code or Codex) watches a viral video and redoes the whole flow as an editable project, with script, images, voice, captions, and animations tied to the spoken words. Swap the presenter, topic, or language and render again. Local test with no API (animation in code + FFmpeg), runtime for an ARM server, limits (paid generation, license with conditions), and what to reuse in INEMA videos. PT/EN/ES guide.
3D video animated with code, no AI video: an 8 m² room and a single-story house assembled piece by piece in TypeScript (Remotion + three.js), with lighting calculated by path tracing on the graphics card, local-voice narration, and word-by-word captions. 9:16 reels of 64 s and 87 s, a script that downloads the CC0 models from Poly Haven, and render-time measurements. PT/EN/ES guide.
Mirror of VoiceStudio (debpalash): a local, open alternative to ElevenLabs for cloning voices, creating voices from a description, dubbing video, and transcribing in 646 languages, with API and MCP.
Long-term memory for agents (Vectorize, MIT): stores, searches, and reflects, consolidates what it has learned, and runs locally through your Claude or Codex subscription
Muse Wonder and Wonder Full, Meta's models from 10/12/2026, in comparative charts with Claude Opus 5.5, GPT-6 Astra, and Gemini 4 Argon: where they win (SWE-Marathon 57.5% × Opus's 46.3%), where they lose (building from scratch, knowledge), price, how to use them via the Meta API, OpenRouter, and Muse Code, and how to test them on your project. Numbers checked at Meta, Artificial Analysis, and Vals. PT/EN/ES guide and videos.
Four councils of agents: strategy, AI technology, future vision, and management. 28 profiles, a three-round debate, and a recommendation with risks and conditions. PT/EN/ES guide.
12 virtual influencers: identity, consistency, what was a fad and signs of continuity, with sources and a hands-on exercise.
From the listing link to the property film: reads only the data in the listing, picks up to 10 photos, creates the professional version of each one and the virtual staging of empty rooms (Codex via subscription), and renders a ~30 s video with animated price, area and highlights, in 16:9 or 9:16 (HyperFrames). Free, no API key, with a presentation video and a PT/EN/ES guide.
Real case: how an AI agent (Claude Code) made a 15 s UGC promo video for INEMA.CLUB from a reference artwork. Scenes in Kling via subscription (CLI, 120 credits), lip sync on the Kling website with the browser logged in (Playwright), ElevenLabs voice, on-screen text with Pillow and assembly in ffmpeg. The idea, 8 tips (use the subscription and an agent that operates the site), 11 problems with the fix for each, scripts and the video in both versions. PT/EN/ES guide.
Portuguese-language kit for Claude Dashboards, the panels that save the queries and redo the calculations every time they are opened. Ready-made prompts for business dashboards (overdue invoices, unbilled work, profit × cash), a checklist to verify a number at its source, a central Supabase base with a scheduled copy, and a phased plan to measure Claude Code itself via OpenTelemetry (estimated cost, tokens, cache, the tools and MCPs that fail most, skills), with a collector that discards prompts and commands. Includes local agent telemetry: a dashboard on your own machine with all Claude Code and Codex sessions live (processing, waiting for you, failures, tokens and estimated cost), turned on and off with one command. PT/EN/ES guide.
Plan for a simple in-house system for a real estate agency handling sales and rentals: AI answers 24/7 on WhatsApp (Evolution API), qualifies leads, shows only real properties and hands the client over to the broker, who works through Telegram; the owner follows along on a dashboard. No forms. Proposal under discussion, with a record of the conversation and a survey of references. PT/EN/ES guide.
Kit for migrating a Claude Code setup to Codex or making the work model-independent: read-only audit, CLAUDE.md → AGENTS.md, portable core (context/tasks/handoffs), canonical skills via polyskill, continuity test in a fresh session, and a ready-to-use template for sharing the same context (private GitHub) across Claude, ChatGPT, Codex and Gemini, with a cartoon-style video.


The 3 most widely used engineering skill packs for AI agents (Addy Osmani's agent-skills, Matt Pocock's skills, and Lauren Tan's pstack), cloned under each one's MIT license: a catalog of the 97 skills by stage, a duplicate-name detector (including against what's already installed), and an installer by light, medium, or heavy profile that brings along Addy's checklists and commands. Step-by-step technical guide in PT/EN/ES.

Open model (MIT) for a help center in map format: visitors choose the problem they have right now and get a step-by-step path, with yes-or-no questions and the right link at each stage. A single HTML page with no build, all content in one configuration file, light/dark theme, a direct link to each path, and a single-column layout on mobile. Comes with the INEMA ecosystem map (inema.club, inema.pro, inema.vip, events, news, search, and adventure), a generic example, and an analysis of the pattern. Guide in PT/EN/ES.
Visible-intelligence rubric: measures whether the product explains itself (5-second test, 6 questions per element, 10 tests, color-coded score). Diagnosis of a finished system or a PRD section, with a skill for Claude Code and Codex.
Popups that guide the visitor instead of just asking for their email: they ask questions, teach, and recommend the right product, course, or content, installed and configured by the site owner's AI agent (Claude Code, Codex, Cursor) with a CLI, MCP server, and skill. Journeys and tests in JSON files in the repository, control group, A/B testing, and Thompson sampling. Inspired by Alia Popups, but open source (MIT), free, no SaaS, and for any site. Today: market research, 37 open-source projects evaluated, and a 6-phase plan; code to follow. Guide in PT/EN/ES.
Talk with an AI coach and rehearse real conversations: the character reacts with its own humor and patience, the evaluator quotes the line that justifies each score and gives a single correction, and you redo the scene until it feels natural. Charisma (work) and Conquest (relationships) modes. Opens in the browser or runs on your PC with Ollama or your Codex/Claude/Gemini subscription, with local voice. Open source MIT: copy, tweak, and sell. Guide in PT/EN/ES.
Series · 3 projects
Cuts YouTube lives into topic-based clips with AI and republishes them on another channel



Which AI draws the best capybara on a bicycle? Open benchmark inspired by Simon Willison's pelican: 19 models (10 local on Ollama, Claude and GPT via subscription, no paid API) draw the same SVG and a blind judge compares them pairwise, in both orders. Round 1 ranking, gallery, a kit to run on your PC, and the method for building your own benchmark. In Portuguese, English, and Spanish.
Can your AI agent leak your data? Lethal trifecta checklist (private data + third-party content + exit channel) and a script that inventories the MCP servers of Claude Code and Codex so you can see whether the agent combines all three and which leg to cut, with examples from INEMA's customer service kits. In Portuguese, English, and Spanish.
Ata: local-first meeting notes in Portuguese, English, and Spanish. Records the call on two tracks, transcribes on your machine (GPU or CPU), separates who spoke, and delivers the minutes with summary, decisions, and action items; search across all meetings, next-meeting agenda, live mode with no paid service, MCP, and second brain. Open source, inspired by CrunchLog. Guide in Portuguese, English, and Spanish.
CrunchLog (Christian Landsteiner): 100% local meeting notes. Records the call on two tracks (what you hear and what you say), transcribes on the CPU, and generates a Markdown note with each person's lines, a summary via Claude Code or Codex, a dashboard, and an MCP server. Full guide and a recorder/demo tested on Linux. In Portuguese, English, and Spanish.

OpenAI's always-on assistant (GPT-6 Astra, Pro/Business only) analyzed from 5 videos and web research: how to enable it, what people did with it, comparison with Muse, Grokbot, and Hermes Agent, bugs, who criticizes it and why (failed demo, security, price, and quota), and a verdict by user profile. In Portuguese, English, and Spanish.
Redesign study of the Municipalidad de San Martín (AR) website: diagnosis of the current site, what the best government websites do in 2026 (search first, task-based navigation, WCAG 2.2 AA), and three conceptual home page proposals (v1, v2, and v3 in the workbench style), in PT/EN/ES.
Series · 2 projects
Sept/2026 survey of what China is doing in AI literacy: national plans, 2025 guidelines, class hours by province, rules for generative AI use in schools, universities, teachers, seniors, civil servants, and small businesses, with critical analysis, comparison with the US, and lessons for Brazil.
Study of the method behind a 7-day challenge in which an agent (Codex + GPT-6 Astra) traded US$ 10,000 in stocks: 7 routines per trading session, handoff between agents, strategist separated from executor (email → bot → Alpaca), loosened risk gates, and the diary up to losing to the S&P 500. Study text, not a recommendation — trading is high risk. PT/EN/ES guide.
Meta Muse × OpenMuse (CopilotKit, MIT): point-by-point comparison between Meta's personal agent and the open-source alternative — where it runs, browser, memory, integrations, privacy, and cost — with every claim checked against a primary source and a step-by-step guide to running OpenMuse.



Local open-source classifier specialist in the Jev style: a small model reads each message and answers fixed questions with yes, no, or can't tell; the code decides the queue and any doubt goes to a person. Examples from a law firm's client intake and a clinic's front desk, data pipeline, training and frozen test, CPU service (ONNX) with Docker, and a step-by-step guide for other niches. Research project, using synthetic data.

Model and effort switcher for Claude Code and Codex: a menu-bar app on macOS and, in v2, a web panel + chat for Linux/WSL via tmux. Fail-closed: it only types after proving the agent is idle.
The secret of the 7 principles of AI agent management: 5 revelations about what these texts really are. It's people management by another name, and it's what anyone who programs with coding agents already does every day. In Portuguese, English, and Spanish.
The right model for each task: the current stack (Claude Opus 5.5, GPT-6 Astra, Sol, and Luna), selection rules, models explained simply, plan→execute and second-opinion prompts, a test battery, and a skill draft. In Portuguese, English, and Spanish.
Claude Opus 5.5: launch research with charts on price, benchmarks, and effort × cost, the API changes, and the update plan for INEMA systems. In Portuguese, English, and Spanish.
GPT-6 Sol & Luna: launch research, comparison charts, a cost calculator with caching, and an integration plan for INEMA systems. Official sources and API examples.
HOLO Hands: control the screen with your hands through the webcam — pinch, drag, throw, and organize notes without touching the mouse. Portuguese adaptation (interface, alerts, and assistant voice) of Zubair Trabzada's HOLO, with sample INEMA notes. Runs 100% on your machine: no cloud, no API key, and no account.
JurisFlow: management system for law firms — clients, cases with CNJ number and stages, case updates, client meetings, calendar and fatal deadlines, tasks, finance, and documents, with an AI legal assistant (petition drafts, case law, deadline calculation, notifications, and fee agreements) and a one-click demo mode.
Style consulting in Portuguese for barbershops: facial analysis, professional review, report, and VPS installation.
Cultivated AI — you don't program it, you cultivate it: the eight elements (role, context, tools, rules, examples, memory, evaluation, feedback) applied to personal life, Jarvis, and business, with ready-made solutions: downloadable kits, /cultivar and /revisao-semanal skills, an agent profile generator, a cultivation diagnostic, and agent packs by area (lead qualifier with an n8n flow).
Dossier on the 12 AIs announced (as rumors) for October 2026 — GPT Bel, GPT-6 Sol, Claude Fable 5.2, Opus 5.1, Grok 4.7, DeepSeek V4.1 Pro, GLM-5.5, Kimi K3.1, Qwen 4.0, Gemini 4.0, Muse Spark 1.4, MiMo-V3 — each compared with the verified current version of its lineage: price, context, benchmarks, timeline, charts, and sources.
Linter for AI-written text: gives a score from 0 to 5, flags vocabulary, constructions, em dashes, lists of three, and invented evidence, and cleans up the draft with a model from another family (Claude ↔ GPT). Pure Python, CI gate ready, skill for Claude Code and Codex.
3D learning world in the browser (Babylon.js): three stations with animated demos, a guide mascot, a 1-minute World Tour with synchronized cards, and free exploration. Teaches three habits for using AI day to day; changing the subject means editing a configuration file. Recipe from the field guide "Build a Learning World".
Vou Contigo: a platform for tracking and supporting the routines of older adults — landing page, appointments dashboard with price metrics, WhatsApp bot (Evolution API) for booking/canceling/free time slots, automatic reminders and reports, PIX Asaas, family portal, and Telegram management (Next.js + Supabase).
Ordinary photos of a room (or just a description) become a 360° equirectangular panorama faithful to the property, with seam checking, a web interface with a viewer, and the property-360 skill
Astra3D V2: three complete demos, visual editor, project library, duplication, saved versions, and backup. Export 3D sites as standalone HTML.
Social Autopilot in Claude Code: a brand voice file + 3 commands (post of the day, week, repurpose) that research, write in your voice, generate the visuals, and schedule through a configurable provider (Metricool, Blotato, or copy) — nothing goes live without your yes.

Guide in Portuguese for comparing Astra effort levels: seven documented runs, ten prompts, data, spreadsheet, and verification tests, with the originals preserved.

A link becomes an editable vertical video: research, script, fixed pt-BR voice, captions, review in HyperFrames, and MP4 render only after approval.
Free, self-hosted scheduling app (MIT): weekly availability, public booking links with time zones, Google Calendar, SMTP emails, and Stripe in test mode — local demo with SQLite and no credentials.
Local-first desktop cockpit for Codex, OpenRouter, and coordinated AI crews: provider, tools, and orchestration kept separate, with permissions and credentials at the Electron boundary.
Decision kit (MIT, Mark Kashef) translated to pt-BR: compares AI assistants by operating model before brand — interview, dated research, reversible test, and a 7-day pilot.
Local orchestrator to run DSH, Codex CLI, and Claude Code with tmux sessions, GitHub CLI, and isolated worktrees.
Open source generative AI platform: image, video, audio, agents, workflows, lip sync, and cinema studio — runs locally or in the cloud.
Series · 2 projects
Bench Studio in Portuguese: the same 73 routes across 5 providers, with a bilingual pt-BR/en interface, start and end frame as two named selectors on models that accept keyframes, a budget that refuses to make up a total when the unit isn't predictable, and one-command remote access (password before the port, API always on loopback).
Single-file pure Python CLI to create, list, and delete DNS records on Cloudflare via the API — includes the CNAME + TXT pair that Vercel requires to verify a new subdomain.
B2B sales machine for interior doors on WhatsApp: the AI qualifies the lead, a deterministic engine calculates the quote from the real catalog (the model never makes up prices), and Pix closes the order.
Personal video and audio downloader that runs on your VPS: you paste the link from your phone, yt-dlp downloads it and returns the file, which deletes itself after 6 hours.
Telegram bot with a durable SQLite queue that runs single-step skills and multi-phase stateful flows, with a human gate and resume after a crash — a new domain comes in without a single line of code.
Open Knowledge Format: an open, vendor-neutral format that represents knowledge as markdown + YAML frontmatter, in git-versionable bundles, with a Python reference agent and a graph viewer.
Reference guide for Asaas payments (PIX, single-payment card, installment card, boleto) with automatic access release, plus passwordless login via magic link (Supabase) — sample code to copy into any Next.js SaaS.
Honest analysis: Nubank has no official payment API/webhook. Documents the BR Code (the Central Bank's public standard), the community workaround of reading the confirmation email via the Gmail API, and a fallback of a payment receipt validated by OCR/vision.
Guide + ready-made scripts to access Zo Computer via SSH: why a direct IP doesn't work, a TCP service (tunnel), authorizing a key, and bringing up sshd on port 22.
Claude Code skill that opens a new terminal running the full harness on Kimi K3 (Moonshot AI) via OpenRouter, for ~1/3 of the cost, without touching the current session or global settings.
Local, secure monitor for Claude Code sessions, with a dashboard for quota, context, processes, and services.
Turns your Claude Code history into a personal, repeatable, blind-judged benchmark to compare models and effort levels.
Private AI server with one command: local inference, chat, dashboard, voice, agents, workflows, RAG, and image generation on your own hardware (Linux, Windows, macOS).
Filesystem-first framework for durable AI agents: instructions, tools, skills, channels, and schedules are files in a folder — eve compiles and runs them (fork/study of Vercel's project)
Telegram bot that becomes a video queue: one instruction per line, web research, transcription of the source audio, narration as text, and delivery to the right folder
Free access to 100+ NVIDIA NIM models (Llama, DeepSeek, Qwen, Mistral, Vision) via an OpenAI-compatible API.
Reusable prompt (EN + PT-BR) that turns the planning of an AI build into a war game: phases with optimistic/pessimistic assumptions, modes of…
Platform of AI agents for civil engineering calculations: the agent chooses the normative method and Python calculates, validates (units/physics/NBR/equilibrium), and generates the calculation report. 18 domains, with a technical responsibility notice (ART).
Research lab behind O Caminho Certo da IA: verified deep research, manifesto, course, and charts on AI/humanoids 2019-2030
Games + education platform (arcade, educational, professional) in Next.js 16
Open source, multi-tenant SaaS for vet clinics, pet shops, hotels, and NGOs
Management SaaS for salons, barbershops, and aesthetic clinics with WhatsApp
Mobile-first web platform to create, publish, and read animated comics
Personal AI assistant in Python, the antidote to bloated frameworks
Brazilian marketplace for calamity insurance: risk by ZIP code and online quotes
Mobile-first SaaS ERP for MEIs (individual microentrepreneurs) and microbusinesses: PIX, WhatsApp, and traffic-light dashboard
Spanish platform 100% for children ages 6 to 12, aligned with the BNCC
Last-mile deliveries by niche (vet), with real-time tracking and PIX
Marketplace of electronics technicians with automatic quoting via WhatsApp
Realtime dashboard of rainfall and alerts with INMET data and multichannel notifications
Generates professional PDF eBooks with AI-made covers, 3 layouts, and Claude Code
Multimodal RAG: chat with PDFs, spreadsheets, images, video, and audio
BNCC-aligned educational platform (grades 6-9) with gamification, dashboards, and an AI tutor
Publishes and schedules to 6 social networks via Node scripts on self-hosted Postiz
Custom publisher (TS worker + Supabase) for Instagram and TikTok, plus Chatwoot
Deep research agent in Python with Tavily, reflection, and Markdown output
Marketing content factory with AI agents orchestrated by a Telegram bot
SkyReels V3 fork with a Web UI to generate video episodes in batch
Multi-agent public opinion analysis (BR/LATAM fork) that turns into a report
Swarm-intelligence prediction engine: simulates agents on social networks
Ready-made images + narration become a pro MP4 film, without video AI (rendered via pixflow)
Open-source alternative to Claude Design: your agent CLI becomes the design engine
SkyReels V3 fork with a Web UI: episode queues, talking avatar, and batch import
Local multi-model image server with hot-swap between Qwen-Edit, FLUX.2, and ERNIE
Open-source multi-platform publishing with self-hosted Postiz and APIs
Open-source, local-first agentic AI assistant with 118+ OAuth integrations
4x upscaling server (super-resolution) with 4 ESRGAN models, HTTP API, and web UI. Local, fast, no call limits.
Manifesto for real AI qualification: using AI for real and developing what AI doesn't replace, with verified data (Gartner, MIT, WEF)
Series · 4 projects
Ready-made, security-hardened Docker setup to run Moltbot, a multi-channel personal AI assistant.
Content guide about San Martín de los Andes (Neuquén, Argentine Patagonia): history and the Mapuche people, Lake Lácar, Lanín National Park, Cerro Chapelco, the Ruta de los Siete Lagos, when to go, and how to get there.
Hosp360 (CRMHosp): initial technical foundation of a patient self-service portal — Next.js + NestJS + PostgreSQL + Keycloak, ready to run locally via Docker Compose.
Also: hosp360
Imob360: complete real estate management system for small and mid-sized agencies — properties, lead CRM, contracts, viewings, and optional modules…
Also: imob360
LogFree: FastAPI API + Telegram bot that recommends the gas station with the best real cost per km for last-mile fleets.
Also: logfree
Andaime: pedagogical platform with Socratic AI.
Also: my-inema
Skills framework that gives Claude the expertise to build, validate, and deploy production n8n workflows, right inside Claude Code.
Also: n8n-ps
Occio Agentes: AI agent orchestration platform for small and mid-sized Brazilian businesses, inspired by Accio Work.
Also: occio-agentes
Control Claude Code and Codex remotely via Telegram or GitHub, with persistent Docker sessions, repository cloning, and commands versioned in your own codebase.
Also: rAgentic-cs
SkillMKT v4 packages 6 marketing skills (creatives, pages, reels, paid launch, messaging, design) into ready-to-install, distributable .skill files.
Also: skillmktv4
React/Vite prototype of the INEMA.Club landing page: presentation, AI diagnostic (mocked), and plans, running 100% in demo mode with no backend configured.
Also: web-AG01
Book creation with AI
AI film direction tool for comic-style movies — storyboard, prompts, and visual language.
BullshitBench: benchmark that measures whether LLMs detect and flag absurd premises in technical questions, with a public viewer.
Web tool that generates, in 5 steps, an optimized prompt for AI builders (Lovable, Bolt, v0) to create a website.
Headless browser automation CLI for AI agents (Rust with Node.js fallback).
Collection of Claude Code skills for generating and styling documents (DOCX/PDF/XLSX/PPTX) with premium brand themes.
Local simulator of the Banco Inter Pix API, with charges, QR Code, webhook, and a web interface, no real account needed.
Skill/template for generating INEMA.CLUB course pages with a learning layer (progress, questions, notes, themes) — self-contained.
Video factory (reels/narratives) from AI images + cinematic motion via ffmpeg, with external TTS and image generation.
ITAGMKT — AI agent pipeline for digital marketing (research, creation, video, multi-platform publishing) via Telegram bot.
Python scheduler to sync/analyze/publish TikTok content, with a web dashboard and spreadsheet-based configuration.
Web platform (Next.js/Vercel) for kids to create and read interactive stories with animated characters and scenes.
Python scripts to generate Word, Excel, PowerPoint, and PDF documents via LLM or standalone; runs independently, requires API setup.
Flask app that generates an AI strategy package (15 documents, diagrams, presentation) for a company via the Perplexity and Gemini APIs.
Automatic video dubbing system: transcription (Whisper), translation, and speech synthesis (Bark/Coqui TTS), with synchronization.
Prompt pipeline (KairoBoost): turns intent into a cinematic prompt and generates videos with images and TTS narration.
timesmkt3: AI agent pipeline (Telegram bot) that researches, creates narrative, generates images/videos, and publishes marketing content.
Hyperframes Editor (Student Edition): HTML+GSAP workbench with 12 ready-made video projects to study, edit, and re-render.
Experimental rewrite (Python/Rust) of a Claude Code-style agent harness, not affiliated with Anthropic.
Antidote: personal AI assistant in Python (~3,000 lines) via Telegram, runs locally/Ollama, no heavy frameworks or Docker.
Fork of Anthropic's official Skills repository — a collection of example skills for Claude (creative, technical, enterprise).
Nanobot: ultra-lightweight personal AI assistant (~4,000 lines), a fork with Brazilian Portuguese documentation and an install/Docker guide.
NeuroFluxo: a neuroscience-based personal organization web app (Next.js) for people with ADHD.
Claude Code skills for automatically generating Word, Excel, PowerPoint, and PDF documents.
Invitation, affiliate, and promo code (sponsor) management system for INEMA.VIP, built with Next.js/Turborepo.
Promotional landing page for the Vibe Coding Immersion (a 3-day in-person course), 4 versions, checkout via Asaas.
Sales landing page for the INEMA.VIP premium community (business transformation with AI).
Corporate website/landing page for Jones BBQ & Paellas, a premium catering company (client).
Mobile telehealth/telemedicine system: electronic health records, appointment scheduling, and integration with e-SUS/DATASUS.
INEMAPROMO — promotional landing pages: the inema.club community, the INTELECTO workshop, and corrected versions (/impeccable).
Modular multi-tenant platform for media pipelines (image, audio, video, text) with a job queue and chat-based input; in development.
Python TUI that monitors active Claude Code sessions in real time: usage quota, per-session metrics, and built-in chat.
Profe Ale — MVP of a teaching platform (Next.js + Supabase) with a lesson journey organized by modules; in early development.
Automatic AI video dubbing pipeline: transcription, translation, and TTS, with a web interface and GPU support via Docker.
Local dashboard that analyzes cost, token usage, and cache of Claude Code sessions from JSONL logs.
AI agent system via Claude Code that researches, writes copy, and generates images and videos for marketing campaigns.
Personal financial consulting SaaS with a portfolio dashboard, AI advisor (Gemini), and market data (Finnhub).
Library/skill for Claude Code with Remotion best practices, animation examples, and a promotional video generator.
Structured code debugging prompt for use with Claude, with standardized variables and a filled-in example.
Next.js app for music rankings (songs, bands, DJs, AI) with voting; monorepo with a mock API.
CLI that compiles Agent Skills written once into formats optimized for Claude Code and OpenAI Codex.
Web app (Drive Explorer) for exploring and downloading 3D models from the INEMA collection.
Next.js platform for publishing content created in NotebookLM, hosted on Vercel.
Marketing site and course catalog for Sisnema Academy (Next.js, with AI for course recommendations).
Next.js dashboard for real-time weather monitoring (forecast, history, map, and accuracy).
Paid-member management system for Telegram groups, with expiration tracking and automatic notifications.
Portuguese voice agent via LiveKit (Deepgram STT, GPT-4o-mini LLM, OpenAI TTS) with time and weather tools.
Personal Telegram bot that runs the local Claude Code CLI, with multi-agent support (Ollama/Codex/OpenRouter) and a second brain.
Free, open-source online IQ test, with procedurally generated progressive matrices (Raven/ICAR style).
Claude OS: read-only personal panel that aggregates usage, cost, and memory from AI tools (Claude, ChatGPT, Pinecone, Obsidian) in a single dashboard.
AI suite for videos: automatic dubbing, transcription, clip cutting, and download, without a GPU (via cloud API or CPU).
ClaudeClaw: brings the Claude Code CLI to Telegram, running the real process on the user's machine.
Full-stack link-in-bio page builder (Linktree-style) with Next.js 15, React 19, and Neon Postgres.
White-label B2B betting/iGaming platform with wallet, KYC, and compliance for the Brazilian and LATAM markets.
Educational RPG in Godot for children/teenagers that uses AI (ARIA) as a strategic resource to teach 21st-century skills.
Simulation engine with a swarm of AI agents that react on social networks to predict how events will unfold (translated fork of MiroFish).