DeveloPassion's Newsletter #181 - AI Revolution
Sébastien Dubois
I help pros turn information into insight, freeing up 10+ hours/week. Trusted by 1,000+ happy customers ❤️
Hello everyone! I’m Sébastien Dubois, your host. You’re receiving this email because you signed up for DeveloPassion’s Newsletter. Thank you for being here with me ✨
If you enjoy this, please forward it to your friends 👍. If this email was forwarded to you, then don’t forget to subscribe and become a supporter. You can also follow me on X, Bluesky, and Mastodon.
Welcome
Another week, another newsletter! I hope that you all had a great one 🤩
It's been a while. These past weeks, I've been really busy learning new things about AI in general, and Large Language Models in particular. And it's been super enlightening. Things are moving REALLY fast, and it's hard to keep up. But it's not an option. We have to follow the trend, or we're bound to become obsolete real soon.
Alright, let's gooooo 🚀
The lab 🧪
The lab has been on fire recently. I've spent a lot of time exploring and using AI tools, and I've worked on an updated version of my Website.
I'll tell you everything in the next sections.
AI Revolution(s)
The AI tools I use
First of all I want to discuss about all the cool AI-related things I've been exploring and working on. I've started using Ollama locally, enabling me to install and use Large Language Models (LLMs) locally. I love how this enables me to leverage AI in various contexts, including in my Knowledge Management practice. Together with the Companion plugin for Obsidian, I'm able to use any LLM I want to generate/summarize/rephrase text. In addition, I still use the Smart Connections plugin for Obsidian to chat with my notes, and find relevant links to create across my notes.
In addition to Ollama, I also use Claude heavily, along with the updated version of ChatGPT (o1). I've seen huge improvements with ChatGPT, but it doesn't yet support o1 over the API.
Next to that, I use Cursor.com and Vercel v0 to help me code better and faster.
I also continue using Replicate.com to generate images using my Replicate plugin for Obsidian and FLUX.1. In the future, I want to give a try to the latest versions of xAI Grok, which is becoming more and more powerful.
Last but not least, I plan on using AI video generation tools in the coming months, such as OpenAI Sora, Hailuo AI, or other ones.
AI Milestones - Model Context Protocol and the future of Writing
The other HUGE improvement I've seen recently is the release of the Model Context Protocol (MCP), an open standard created by Anthropic (the company behind Claude). It's currently supported in Claude Desktop, but will probably be supported by other Large Language Models in the future. I really believe that MCP is an important milestone for the future of (truly open) AI. It's an enabler for tons of concrete use cases.
In a few words, the Model Context Protocol (MCP) enables connecting AIs to local and remote data sources. It uses a client-server protocol, where the client (e.g., Claude Desktop) connects to a server that provides access (read/write) to "whatever". There are already dozens of MCP servers in the wild (e.g., to access databases, search the Web, interact with Notion, control Spotify, etc).
You can find more details, tutorials and useful resources in my notes about MCP:
Since discovering MCP, I've been able to use the filesystem server to connect Claude Desktop to my Obsidian vault, which makes it a breeze for me to ask Claude to read my notes, and interact with those. And that alone is mighty powerful. I can create new notes based on my discussions with Claude, update existing ones, summarize those, etc. But that's just the tip of the iceberg.
After watching Evelyn Chapman's AI Creative Copilot Masterclass, I realized that I could actually use the MCP and Claude Desktop to create a powerful AI writing assistant for myself. And the results have been really amazing. I don't plan on writing everything with AI, but I do want to create more content with the help of AI. Right now, my AI writing assistant already knows about:
I've fed Claude and ChatGPT with tons of samples of my own writing, and derived a set of "rules" I can feed back into any AI to write using my own style, respecting my own voice, my templates, my rules and quirks, etc. And believe me, the results are really impressive. And the coolest part is that it's not a one-time thing. I can update those documents anytime, and the AI will adapt immediately.
And after all that, I started using AI to study the style of successful online creators such as Nicolas Cole, Sahil Bloom, Justin Welsh, Ryan Holiday, Dickie Bush, etc. I've then used the results to adapt and improve my own content types and templates.
I've published a summary of my findings along with my approach here:
If you're curious about this, then you're in luck. I plan on creating a tutorial/course about this in the near future. Just reply to this e-mail if you're interested.
AI Milestones - AI-generated podcasts & voice
Another experiment I've done is using NotebookLM to automatically generate podcast episodes based on my content. I fed NotebookLM with the content of my latest article, and it generated a podcast discussion covering all the points. I don't know about you, but the ability to do something like that feels amazing.
It took me about a minute to create this. And you can do the same for free, using any piece(s) of content you want (images, documents, etc). How cool is that?!
Because of this, I decided to create my podcast on Substack and Spotify, and publish this as the first episode. I don't know if I'll continue, but it's certainly an interesting experiment.
Last but not least, I'm also eager to experiment cloning my voice. I found this interesting tool that I want to try:
AI Milestones - ChatGPT Improvements
Apart from the release of o1, OpenAI has also made other big announcements during their 12 Days of OpenAI event.
Here's an overview of ChatGPT projects:
They've also announced support for Canvas, which is similar to Claude experiments.
AI Milestones - Gemini 2.0
Meanwhile, Google has released Gemini 2.0, which is really mind blowing. What's interesting about Gemini 2.0 is that it's a multi-modal Large Language Models (LLMs). it can handle a full range of multi-modal inputs: text, documents, images, video, and audio. In addition, it also supports streaming, which enables it to receive, analyze, understand, analyze, and react to various kinds of inputs (including live video!).
Gemini 2.0 is also able to return bounding boxes for objects within an image, which can enable various scenarios. It can also read, write, and execute code.
I'm really eager to dive into it soon.
Useful links:
New Website
Since I was already knee-deep into AI exploration, I used it to "roast" my existing Website, including, the landing page, the navigation, etc. Based on that, I imagined an improved version with the help of AI. I was really frustrated with the Ghost theme I was using. It didn't give me enough freedom, and forced me to jump through hoops to do some things (e.g., landing pages, post signatures, etc). So, I decided to fork the theme I was using, and create my own version.
It took me about a week to adapt it to my own needs. I started with integrating and rewriting most of the CSS using Tailwind. Then I created a new landing page:
This new landing page better explains who I'm (mainly) here for, what I want to help people with, who I am, my approach, etc.
Then, I added a start page. That page focuses on describing a roadmap newcomers can follow to explore my content, and grow their skills step by step.
Recommended by LinkedIn
After that, I recreated the blog page, the newsletter page, cleaned up the tags, replaced the footer, and created new pages:
It's not perfect, but it's already a huge improvement over what I had before. In the coming months, I want to embed the landing page for all my products into the Website, and also create additional landing pages for each specific audience I want to "target".
In the coming weeks, I'm going to migrate my articles about software development to my notes Website. I don't want to write about that topic for now, and I don't want to create confusion. I already converted those to Markdown, and I'm going to publish them using Obsidian Publish.
Last but not least, while discussing with ChatGPT o1, I realized that what I should really do is reboot the PKM community, give it a new home, and focus on it. I'm not sure about the platform I'll be using yet, but my goal is to add a lot more value for paid subscribers. If you want to benefit from the best possible price, then join us today.
New articles
I published an article discussing the fact that your habits and routine MAKE you. They define who you are, and determine who you'll be in the future:
To help you reflect and grow, I've included a list of key questions to explore. Some readers have also shared additional insights on Reddit.
In addition, don't forget to check the updated version of my post covering the best Obsidian plugins out there:
Note that, obviously, many of those are part of the Obsidian Starter Kit
I've also published an article sharing concrete tips to avoid the trap of perfectionism:
Notes of the week
Paulo Coelho is one of the most inspiring authors. Consider the powerful "favorable principle":
Give opportunities a chance to change your life. Pay attention, and follow those.
If you want to write better, you need to answer the why, the what and the how:
Books
I recently started reading The Art and Business of Online Writing (book) by Nicolas Cole. I want to improve my content creation engine, and the insights that Nicolas shares in that book are really valuable.
I also grabbed a few biographies that I plan to read soon:
Another one I want to read soon is "Feynman's Rainbow" by Leonard Mlodinow:
Quotes of the week
Meme of the week
Links of the week
People are exploring novel ways to explore ideas:
Zsolt Viczián has shared his AI workflows for reading, note-taking and visual thinking:
Microsoft has released an awesome new tool for converting various file types to Markdown: PDF, PowerPoint, Word, Excel, Images, Audio, HTML, and more!
Obsidian Web Clipper is awesome and will keep getting better
Other links:
The Smart Connections plugin for Obsidian of Brian Petro is also improving a lot:
I recently stumbled upon the Sticky Headings plugin for Obsidian. I might add it later to the Obsidian Starter Kit:
I recently shared a thread explaining what the Knowledge Worker Kit is all about:
That's it for today ✨