
| Routine tasks automated | 90% (author's claim) |
| Post-production speed increase | x2-3 (author's claim) |
| Claude Pro monthly cost | $20 (shown on screen) |
| Claude Max monthly cost | $100 (shown on screen) |
| RTX 4090 price range | 250-350k RUB (shown on screen) |
| Value of tokens in Claude Pro vs API | $211 vs $20 (shown on screen) |
Who will this analysis save an hour?We’ll send the link with the “actually” takeaway — they won’t have to watch the whole video.
A standard chatbot like ChatGPT predicts the next word to answer a prompt. An AI agent uses an LLM as a brain but is embedded in a 'harness' that gives it tools to browse the web, read/write files, and run code. Agents can work in loops, checking their own progress until a final business goal is achieved.
Running agents via API can be very expensive because they use millions of tokens to complete complex tasks. However, using a subscription like Claude Pro ($20) or Max ($100) provides much higher value, often giving you 10 times more token usage than you would get for the same price via API.
Yes, you can run agents locally using tools like Ollama or LM Studio combined with a harness like Codex. This keeps your data on your own hardware, but you will need a powerful computer with a high-end graphics card (like an RTX 3090 or 4090) to run models that are smart enough for complex business tasks.
You teach agents by creating 'Skills,' which are folders containing a system prompt, reference documents, and examples of the desired output. By placing these in a specific directory that the agent harness can read, the agent gains the context needed to perform specialized tasks like drafting contracts or posting to your specific CMS.