The links to courses, projects and videos on this page are for INEMA.PRO subscribers.After signing in, come back to this page.

Subscribe to INEMA.PROAlready a subscriber? Sign in

INEMA catalog

All videos

Everything that has been on INEMA.CLUB, with no repeats. Several on the same subject share one card: open it to see them all.

107 videos in 54 cards

Local: INEMA players and files 72 videos

Say what matters

Managers already know the hardest part: processes, customers, and team. What's missing is talking to the agent the way you talk to a new employee — the 5-4-3 rule (5 questions, 4 levels of autonomy, 3 test cases), the technician who becomes a mentor, auditor, trainer, and consultant, and the bakery example. Includes version 1, with 6 communication ideas.

updated on 10/18/2026

Spring cleaning for your Claude and Codex setup

Why the agent gets worse over time and how to lighten the load: an audit with one prompt, a script that reads the history, an AGENTS.md as a control tower, folders per business, playbooks, and the three-question filter before installing anything new.

updated on 10/18/2026

Git is a garden: the 6 commands without fear

Git explained with a gardener's story: clone brings the garden home, commit takes the photo, branch tests an idea, worktree opens two flower beds, merge brings everything together, and push shares it with the team. Drawings, animated commands, and the commit graph. Also as a 9:16 reel.

updated on 10/17/2026

REA reads your bank's app, but not the server PT·ES

REA sees everything that reaches your device: programs, Android and Electron apps, the code a site sends to the browser, and network requests. But balance, identity, and transfers are decided on the server, and its code never leaves there. What leaks is what was put inside the app; the rule is to keep keys and decisions on the server. Also as a 9:16 reel.

updated on 10/16/2026 · Area: REA

Your listing becomes a film: imovelfilme from link to property video

imovelfilme presentation video: paste the listing link and it reads the data, sorts the photos, makes a professional version of each one, stages the empty rooms, and delivers the finished property film. Made with the engine itself; the demo listing is fictional. Also the generated film, in 16:9 and 9:16.

updated on 10/14/2026

Claude and Codex with the same context

Your business context in a private GitHub that is yours, read and written by Claude, ChatGPT, and Codex. In drawings: what context is made of, level 1 on the computer, level 2 in the cloud, the project instructions, and how to switch AIs without redoing anything. The ready-made template is in the agente-claude-codex kit.

updated on 10/13/2026 · Area: Claude → Codex

AI said no? Break any restriction PT·ES

When AI refuses to do the work: an unlocked open model, DeepSeek's harness, and REA with Ghidra rebuilt the Mac video player, with free speed control, for four dollars. The three pieces, the roughly 80% limit, and responsible use. Also as a 9:16 reel.

updated on 10/12/2026

10 free tools that turn into an agency with Claude

Ten free GitHub repositories, each demonstrated on screen: HyperFrames, Google Workspace CLI, Apify, a cinematic scrolling website, Humanizer, Agent Skills, Expo, Supabase and Stripe, Playwright and Claude Security, and SEO. With the problem each one solves and the service you can sell with it.

updated on 10/11/2026

20 AI news items of the week, each one demonstrated

Twenty AI news items in under eight minutes, each shown on screen: Claude inside Google Workspace, /dashboard, /design, and /motion, OpenAI building interfaces on the fly, subagents with effort levels, Uber's MCP gateway, the Decisions API, app reverse engineering, and the reporting site that withstood 143,000 people at the same time. Also in a vertical version.

updated on 10/11/2026

inemaSEO in practice: your site found on Google, Bing, and AIs

Writing more pages doesn't help if the site's foundation is broken. The open inemaSEO method: three principles (foundation first, only real data, measured without the brand), the 10 steps with scripts, from the audit to notifying search engines, agents with two human gates, the cadence, and plan B. No paid tools.

updated on 10/10/2026 · Area: inemaSEO

Long Runs in practice: the agent working for hours or days, safely

Sending an agent to work for ten hours without a method doesn't work: it gets lost, redoes what it already did, or approves itself. The three pieces of the open Long Runs kit (a goal with proof of done, state in files, caps and gates), the two paths (supervised or running alone with frozen tests), the context bands, and the first longrun.

updated on 10/10/2026 · Area: Long Runs

From builder to consultant PT·EN·ES

For AI professionals and devs: when building becomes a commodity, what to learn to earn in 2027. The four roles (consultant, follow-along coach, mentor, and auditor) and five things to learn, none of them a tool.

updated on 10/08/2026 · Area: Professional 2030

Decisions API × Jev

OpenAI's Decisions API compared with Jev on cost, speed, and accuracy, with three side-by-side demonstrations. Source: video by Mark Kashef on YouTube.

updated on 10/07/2026 · Area: JEV

Cultivated AI PT·EN·ES

You don't program it, you cultivate it: why your AI repeats the same mistake and how it improves.

updated on 10/01/2026 · Area: IA Cultivada

Series · 30 videos

Reels

updated on 09/28/2026

See all 30

On YouTube 35 videos

The 5 levels of the second brain with Claude: which one are you?

From the folder with CLAUDE.md to the always-on brain: what each level solves (curated wiki, semantic search, graph, server), Claude and Codex memory (CLAUDE.md, MEMORY.md, AGENTS.md), and the question that shows where to stop. Most people solve it in the first three levels. With the free astra-2cerebro kit.

updated on 10/14/2026

Meta's AI that beat Opus: charts, how to use it, and how to test it

Muse Wonder Full scored 57.5% on the programming marathon versus 46.3% for Claude Opus 5.5, at the same price per token. Charts of where it wins and where it loses (programming from scratch, knowledge), the caveat about tests run by Meta itself, how to use it via the API, OpenRouter, and Muse Code, and how to test it on your own code. Also as a 9:16 reel.

updated on 10/14/2026

Skills inside out: Addy Osmani, Matt Pocock, and pstack step by step

Technical video: the anatomy of a skill (SKILL.md, the description as a trigger, disable-model-invocation), Addy's package inside out with the /spec, /plan, and /build flow, how to install each package and the standalone-skill gotcha, name conflicts, and the trio-skills project, which clones, catalogs, and installs by profile.

updated on 10/12/2026

Google's secret AI that codes like Opus 5.5

Internal documents show Gemini 4 Carbon running only inside Google and coding at the level of Claude Opus 5.5, while Argon hasn't even reached the public. And Microsoft's opposite move: Decision-1, a small model that only decides, 35× faster than GPT-6 Sol. Also as a 9:16 reel.

updated on 10/12/2026

Your own Jarvis, without writing code

An AI assistant that's all yours, that knows you, remembers everything, and replies on Telegram. The seven pieces of every Jarvis, INTELECTO (you choose the pieces, Claude Code or Codex builds it), more than twenty open systems and more than twenty free courses, all in the JARVIS area. Also in 9:16.

updated on 10/12/2026 · Area: JARVIS

The biggest building in history is a chip factory (TeraFab)

Elon Musk's TeraFab in Texas: more than 100 million square feet under a single roof, five times the largest building today. The reason: according to Musk, all the chip factories in the world deliver only about 2% of what his companies will need. The goal of up to 200 billion chips per year and what this race changes for you.

updated on 10/11/2026

AI in mathematics: proving is not understanding

OpenAI published 722 math papers at once. What Lean, Riemann, and Unique Games are in plain language, why mathematicians reacted badly, the Navier-Stokes case (a mathematician accuses, OpenAI denies), the risk to cryptography, and what this changes for those who use AI at work.

updated on 10/10/2026