How to learn more about AI agents once you know the basics comes down to one habit, which is thirty minutes a day split between watching, reading the changelog and building.
I'm Julian Goldie, and I test new AI agents and models on camera almost every day.
You already know what an agent is, so this page skips the beginner material.
Here is the short path, and the rest of the page shows you how to run it.
- You pick one main agent and follow its release notes every week.
- You watch one build video a day and copy one thing from it.
- You join one community so you can ask questions when you get stuck.
- You test every new tool yourself with the same fixed prompts.
- You turn each thing you learn into a small build within 48 hours.
Where To Learn More About AI Agents, Ranked By Format
This is my own list, so my own channels sit at the top, and I've said so plainly.
I ranked the formats by how fast they turn into something you can use.
| Rank | Format | Where to go | What it is best for |
|---|---|---|---|
| 1 | Daily videos | My YouTube channel and my podcast channel | You see a new tool running on a real screen the week it launches. |
| 2 | Communities | AI Profit Boardroom and the free AI Money Lab | You get answers when a setup breaks. |
| 3 | Written guides | agentos.guide | You can follow each step at your own speed. |
| 4 | A model leaderboard | goldiebench.com | You can see which model to try next. |
| 5 | Official docs and changelogs | Hermes, Anthropic, OpenAI and Google | You learn what actually changed, from the people who shipped it. |
| 6 | GitHub repos | The repos behind the tools you use | You see activity before any announcement reaches you. |
| 7 | Newsletters and blogs | A short list of long-running ones | You get a weekly summary with context. |
| 8 | X accounts | The official accounts of the labs | You get launch alerts on the day. |
| 9 | Papers | Three or four classics | You understand why agents work the way they do. |
Videos are first because watching a tool run is the fastest way to judge it.
Papers are last because they are slow to read and rarely change what you do on Monday morning.
Every other row is covered in its own section below.
The 30-Minutes-A-Day Routine For Learning More About AI Agents
Most people who know the basics stop improving because they scroll instead of following a plan.
This is the weekly plan I would hand you, and each day has one job.
| Day | Your 30 minutes | The output |
|---|---|---|
| Monday | You read the release notes for your one main agent. | You write down one new feature to try. |
| Tuesday | You watch one build video and pause it to copy the steps. | You have one working copy of what you saw. |
| Wednesday | You build with the feature you noted on Monday. | You have one small thing that runs. |
| Thursday | You read one docs page about a feature you have never used. | You know one more thing your agent can do. |
| Friday | You test one new tool or model with your fixed prompts. | You have one new row in your score sheet. |
| Saturday | You post what you built in a community and ask one question. | You get feedback from people doing the same work. |
| Sunday | You review your notes and pick next week's build. | You start Monday with a plan. |
Thirty minutes is short on purpose.
A routine you can finish before breakfast will still be running in three months.
You should also notice that only two of the seven days are about new information.
The other five days are about building, testing, asking and reviewing, because that is where the learning sticks.
🔥 Want the routine, the agents and the people in one place? Inside the AI Profit Boardroom, you get the installable Agent OS, step-by-step courses on Hermes Agent, Grok Bot, DeepSeek Harness and more, and four coaching calls a week. There are 3,400+ members learning and building next to you. → Get access here
How To Read A Changelog In Five Minutes
Changelogs are the most underrated way to learn more about AI agents, because they are written by the people who shipped the change.
They also move fast.
When I checked the Hermes Agent releases page on 13 October 2026, it listed four releases in about three weeks.
Version 0.21.4 came out on 21 September, version 0.21.5 on 24 September, version 0.21.6 on 8 October and version 0.21.7 on 12 October.
If you only check your main tool once a quarter, you are running months behind the tool you already have installed.
Here is how I'd read any release note in five minutes.
- You skim the headings and ignore the bug fixes on the first pass.
- You look for new commands, new settings and anything marked as a breaking change.
- You pick one new feature that touches a job you already do.
- You write that feature on Monday's line in your notes.
- You close the page, because one feature a week is enough.
These are the official pages worth bookmarking.
| Source | What you will find there |
|---|---|
| Hermes Agent docs | The docs include a Learning Path page with beginner, intermediate and advanced reading orders. |
| Hermes Agent releases | Every version is listed with its notes. |
| Claude release notes | Anthropic lists changes to the Claude platform here. |
| Anthropic Academy | Anthropic lists courses such as Claude Code 101, Building with the Claude API and Introduction to Model Context Protocol. |
| OpenAI API changelog | OpenAI lists changes to its API here. |
| OpenAI Agents SDK docs | These are the docs for OpenAI's framework for multi-agent workflows. |
| Google Agent Development Kit docs | These are the docs for Google's open-source toolkit for building agents. |
| Model Context Protocol | This is the home of the standard that connects agents to tools. |
The Hermes Learning Path is a good Thursday read.
Its intermediate tier walks through sessions, messaging, tools, skills, memory and cron, and the page estimates two to three hours for it.
GitHub Repos To Watch
A repo tells you the truth about a project, because you can see when it was last updated.
Every repo in this table had new commits in October 2026 when I checked.
| Repo | What it is |
|---|---|
| NousResearch/hermes-agent | This is Hermes Agent, which describes itself as the agent that grows with you. |
| anthropics/claude-code | This is Anthropic's agentic coding tool for the terminal, with a public changelog file. |
| openai/codex | This is OpenAI's coding agent that runs in your terminal. |
| openai/openai-agents-python | This is OpenAI's framework for multi-agent workflows. |
| google/adk-python | This is Google's code-first Python toolkit for building and deploying agents. |
| modelcontextprotocol/servers | This repo holds Model Context Protocol servers. |
| microsoft/ai-agents-for-beginners | This is Microsoft's set of 18 lessons on building AI agents. |
| ollama/ollama | This runs open models on your own machine. |
You don't need to read the code.
You click Watch on the repo, choose Custom and tick Releases.
GitHub then tells you when a new version ships, and you ignore everything else.
Three or four watched repos is plenty for one person.
How I Test A New Tool The Day It Launches
Reading about a launch teaches you very little, and running it teaches you a lot.
My routine is on my GoldieBench about page, and it has three steps.
- I wire the new model into my Agent OS, so I can send it work from the same board my other agents use.
- I run my fixed prompt set through it with one shot each, no second tries and no hints.
- I score each result from 0 to 10 on camera, based on whether it ran, how close it got to the brief and how good it looks.
The video above is an example, because it shows me putting a new free model into Hermes Agent and trying it on real tasks.
You can run a smaller version of this in your Friday slot.
- You save three to five prompts in a text file, and each one is a real job you care about.
- You add the new tool or model next to the one you already use.
- You run every prompt once and save what comes out.
- You give each result a score from 0 to 10 and write one line about why.
- You compare the scores with your current tool before you change anything.
The prompts must stay the same every time.
Fixed prompts are what turn an opinion into a comparison.
How To Avoid Hype When You Learn More About AI Agents
Hype is the tax you pay for following a fast field.
I use three checks, and they take about ten minutes together.
You check the repo and the README.
A real tool has a repo or official docs, a README that explains how to install it and recent activity.
If a tool has a launch video and nothing you can install, you wait.
You check the date.
Agent advice goes out of date quickly.
Anthropic's well-known engineering post, Building effective agents, was published on 19 December 2024.
Anthropic has since added a note at the top saying that much of the tooling landscape it describes has changed.
The ideas in that post are still worth reading, and the note is a good reminder to look at the date on everything.
You test it yourself.
A newer model is not always better for your work.
On my own leaderboard, Claude Opus 5 averages 8.27 and the newer Claude Opus 5.5 averages 7.57 on the same one-shot prompts. Opus 5.5 was scored in a separate, stricter run than Opus 5, so I treat that gap as a reason to test, not as proof.
That is one test with one set of prompts, which is exactly why you run your own before you switch.
🔥 Want to see new tools tested before you spend time on them? Inside the AI Profit Boardroom, I share the setups from my daily videos, and you can ask about any new launch on four coaching calls a week. You also get the Agent OS that I wire every new model into, alongside 3,400+ members. → Get access here
Newsletters, Blogs And X Accounts Worth Following
I kept this list short and neutral.
Every source below is long-running, and each one had published within the week before 13 October 2026.
| Source | What it is |
|---|---|
| Simon Willison's Weblog | This is a frequently updated personal blog covering AI tools and models. |
| Latent Space | This is a technical newsletter and podcast for AI engineers. |
| Import AI | This is Jack Clark's weekly AI newsletter, which had reached issue 476. |
| The Batch | This is the weekly AI news email from DeepLearning.AI. |
One newsletter is enough.
You pick the one whose style you like and unsubscribe from the rest.
For X, I'd follow the official accounts, because they are linked from each organisation's own website.
- @NousResearch is the lab behind Hermes Agent.
- @AnthropicAI is Anthropic's account.
- @GoogleDeepMind is Google DeepMind's account.
- @simonw is Simon Willison's account.
- @JulianGoldieSEO is mine, where I post launches and screenshots from my tests.
You should treat X as an alarm bell and nothing more.
It tells you that something launched, and then you go to the docs and the repo to find out what it is.
Papers For The Curious
You can skip this section and still become very good with agents.
If you like knowing why things work, these well-known papers are free to read on arXiv.
| Paper | Year | Why people still read it |
|---|---|---|
| ReAct: Synergizing Reasoning and Acting in Language Models | 2022 | It describes the loop of thinking and then acting that most agents still follow. |
| Toolformer: Language Models Can Teach Themselves to Use Tools | 2023 | It is an early paper on models learning to call tools. |
| Generative Agents: Interactive Simulacra of Human Behavior | 2023 | It explores agents with memory, reflection and planning. |
You read the abstract and the introduction first.
If those two parts don't change how you'd build something, you can stop there.
How To Turn Learning Into A Build
Learning that never becomes a build fades within a week.
My rule is simple, which is that one thing learned becomes one thing built within 48 hours.
I teach the same order to everyone.
- You start with one agent doing one job.
- You give each model its own profile, so you can swap brains without breaking your setup.
- You use free models first and pay only when a job needs it.
- You connect the pieces into an Agent OS once two or three jobs are running.
One build doubles as a learning tool, and it is a morning radar.
My own radar is an agent on a schedule that wakes at six and reads live posts on X, then scores each story.
When I wrote the guide to it, it had logged 49 mornings and scored 294 trends.
Your version can be an agent that checks three release pages each morning and sends you a short summary.
You can also give your agent a notes vault so everything you learn is saved once, and my Hermes second brain guide shows the setup.
If you want the whole dashboard, my Agent OS build guide walks through it in five steps with the exact prompts.
Related Reading
- My how to learn AI agents guide covers the full path if you want to revisit the basics.
- My best AI agent courses online list ranks the courses worth your time.
- My best AI agent course for 2026 explains which course I'd pick this year.
- My best Skool courses list compares the communities where this kind of learning happens.
- My Julian Goldie page tells you who I am and what I've built.
Also On Our Network
- 🌐 How to turn what you learn about AI agents into real builds for a business
- 🌐 Every AI agent learning source compared, with scores and a verdict
- 🌐 A 30-minute daily learning loop built inside an Agent OS
- 🌐 How to follow AI models and test each launch with a leaderboard
FAQ: How To Learn More About AI Agents
How do I learn more about AI agents after the basics?
You pick one main agent and follow its release notes.
You watch one build video a day, join one community and test new tools yourself with fixed prompts.
You then turn each thing you learn into a small build within 48 hours.
How much time do I need each day to stay current with AI agents?
Thirty minutes a day is enough if you give each day one job.
A simple week is changelogs on Monday, a video on Tuesday, a build on Wednesday, docs on Thursday, a test on Friday, community on Saturday and a review on Sunday.
Where are the best places to learn more about AI agents?
I rank them by format.
The order is daily videos, communities, written guides, a model leaderboard, official docs and changelogs, GitHub repos, newsletters and blogs, X accounts, and then papers for the curious.
How do I test a new AI agent tool on launch day?
You wire it into the setup you already use.
You run the same fixed prompts you run on every tool, with one shot each, and you score the results from 0 to 10.
How do I avoid AI agent hype?
You check that the repo and README exist and were updated recently.
You check the date on whatever you are reading, and you test the tool yourself before you repeat a claim about it.
Do I need to read research papers to learn more about AI agents?
You don't need to read papers.
If you are curious, ReAct, Toolformer and Generative Agents are three well-known papers that explain ideas modern agents still use.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I started Goldie Agency on my own, and it is now a 7-figure SEO and link-building agency with 70+ people.
- I've grown my YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught 29,000+ students.
→ Get my best AI training inside the AI Profit Boardroom
Give the routine thirty minutes a day for one month, keep your score sheet honest, and you will have a working answer to how to learn more about AI agents.
