💼 GPT-5 Matches Human-Level Across Jobs

PLUS: How to Learn AI Agents Steps By Step (Free GitHub Course)

Author

Alvaro Cintas
September 26, 2025

Good morning, AI enthusiast. OpenAI just dropped some eye-opening numbers: GPT-5 and Anthropic’s Claude Opus 4.1 are approaching human-level performance across a range of professional tasks.

PLUS: How to learn and master AI agents step by step (free GitHub course).

AI NEWS

OpenAI introduced GDPval, a benchmark testing AI performance on 44 occupations in nine industries, from finance to healthcare. GPT-5-high matched or outperformed humans 40.6% of the time, while Anthropic’s Claude Opus 4.1 scored 49%.

→ Tasks include financial analysis, software engineering, and journalism
→ GPT-5 nearly triples GPT-4o’s 13.7% performance from 15 months ago
→ Benchmark shows AI can offload routine work, letting professionals focus on higher-value tasks

Why is important: GPT-5 and Claude Opus 4.1 demonstrate that AI is becoming a credible assistant for complex professional work. While full job replacement is still far off, these models can significantly boost productivity and decision-making across industries.

AI NEWS

OpenAI is rolling out ChatGPT Pulse, a feature that generates personalized reports while users sleep. Pulse delivers 5–10 briefs each morning, summarizing news, events, or personalized tasks, and is currently available to Pro subscribers.

→ Generates proactive AI reports with text & images
→ Works with Gmail, Google Calendar, and other ChatGPT Connectors
→ Uses memory and context from previous chats for personalization
→ Users can request new reports or provide feedback for refinement

Why is important: Pulse signals a shift from reactive AI to proactive assistance, helping users start their day informed and organized. It could compete with news apps and newsletters while demonstrating the potential of AI as a personal assistant.

AI APPS

Meta Platforms launched Vibes, a new feed of AI-generated short-form videos, aimed at accelerating its AI efforts and giving users creative tools to make, remix, or enhance video content.

→ Create videos from scratch or remix existing content
→ Add new visuals or layer in music
→ Upload directly to Vibes or cross-post to Instagram & Facebook stories/reels
→ Available on the Meta AI app and Meta AI website

Why is important: Vibes represents Meta’s push to integrate AI-generated content into its ecosystem, potentially opening new monetization avenues and expanding creative tools for social media users.

HOW TO AI

How to Learn AI Agents Steps By Step (Free GitHub Course)

If you want to go from playing with prompts to actually building an AI workforce, this open-source GitHub repo is the only crash course you’ll need. It’s structured as a practical guide, covering everything from the fundamentals of agents to advanced reasoning and multi-agent collaboration.

🧰 Who is this for

  • Beginners curious about what AI agents are

  • Developers who want to learn how agents use tools like Slack or Docker

  • Builders exploring RAG, MCP, and reasoning models

  • Researchers diving into memory and multi-agent systems

STEP 1: Access Repo

Go to this GitHub repo. At the top you’ll see the standard tabs: Code, Issues, Pull requests, Actions, Projects, and so on.

Stay on the Code tab. Below, under the free_courses/agentic_ai_crash_course/ folder, you’ll find all the lessons (part1_what_are_ai_agents_anyway.md, part2_the_4_types_of_agentic_systems.md, and so on).

STEP 2: Start With Basics

Click on part1_what_are_ai_agents_anyway.md. This opens a detailed explanation of what AI agents actually are, why they’re different from simple LLM prompts, and how they form the backbone of agentic systems. Read this before moving to the technical parts, it’s your foundation.

STEP 3: Work through tools, RAG, and reasoning

Next, move down the list of files:

  • part3_what_are_tools_in_ai.md → explains how agents connect with tools like Slack, Docker, APIs.

  • part4_what_is_rag_and_agentic.md → introduces Retrieval-Augmented Generation in agent workflows.

  • part5_what_is_mcp_and_why_care.md and part6_planning_in_agents_reasoning_models.md → dive into planning, reasoning, and model coordination.

Click on each file to open its lesson directly in GitHub. Each section gives both conceptual explanation and applied examples you can replicate.

STEP 4: Learn memory and multi-agent systems

Finish with part7_memory_in_agents.md and part8_multi_agent_systems.md. These lessons show how agents remember past interactions, collaborate with other agents, and build systems that actually deliver results.

ESSENTIAL BITES

  • Spotify rolled out new AI safeguards, a spam filter, impersonation rules, and AI disclosure credits, after removing 75M+ AI spam tracks.

  • Microsoft cut certain Azure and AI service access for Israel’s Ministry of Defense after an investigation found the tech was being used to store surveillance data on Palestinian phone calls.

  • Clarifai launched a new reasoning engine that makes AI models twice as fast and 40% cheaper to run, optimizing inference for multi-step agentic models across cloud platforms.

  • Elon Musk’s xAI will offer its Grok chatbot to U.S. federal agencies for just $0.42 per year, undercutting OpenAI and Anthropic’s $1 pricing and including engineering support for integration.

  • Databricks is integrating OpenAI models, including GPT-5, into its platform and Agent Bricks in a $100M deal to boost enterprise AI adoption.

Hot AI Tools

That's it! See you tomorrow

- Dr. Alvaro Cintas