AI Assistants
This week, I want to step outside the usual theme of the newsletter and talk about AI assistants.
This week, I want to step outside the usual theme of the newsletter and talk about AI assistants.
Last week, I attended the second Claude meetup event. The focus was on creating AI assistants, what we might call “Skills” in Claude’s language, that can handle some of our repetitive tasks. “Skill” is the name of Claude’s customizable method/template system for repetitive tasks.
The event, organized by Claude Ambassadors Onat Vural and Said Sürücü along with QNBEYOND, drew a lot of interest. I even ran into some faces from the Web3 world I hadn’t seen in a while, which was a nice chance to catch up.
It was an enjoyable event, not too much like a lecture.
After learning a bit about “Skill,” we heard from Gözde Sarak, Bedir Siyar Lacin, and Kerim Batuhan Baran, who had developed their own assistants for their specific work and needs.
Gözde talked about an AI assistant she created to help fill out a massive Excel sheet full of complex formulas, which she uses for company evaluations.
Şiyar was using an AI assistant to manage his many AWS clients, helping him decide when and how to engage with them.
Kerim Batuhan Baran explained how he set up a system with an AI assistant to track news and organize information, making it easy to find what he needs.
The key takeaway from these stories was the clarity of the steps involved in their tasks and the strategy used, essentially, the algorithm to be transferred into the AI environment.
These stories got us thinking about where we might need an AI assistant and what tasks we could get help with.
For me, trying to use ChatGPT every week to write the newsletter has honestly become unsustainable. Getting it to research weekly news and topics, and supporting me with sources and usage, is taking longer and longer. Plus, it’s getting harder to do this within a specific time frame and area. Even when I set up separate projects or provide a master prompt, it feels like the context and commands get forgotten, and we end up starting over each week.
So, I decided to try out “Skills” on Claude. ChatGPT might have been tired of our ongoing disagreements and my dissatisfaction with its performance, because it actually encouraged me to attend the event about “Skills,” saying an AI assistant might do a better job than it could. 🙂
I set up two different “Skills” on Claude. Well, I shouldn’t say I set them up, Claude did. I just shared the content and context of my past newsletter articles and answered some questions.
Now, I have one assistant for news research and another for technical details, quotes, and sources for my writing. But since I didn’t need them for this week’s article, I haven’t fully tested them yet.
If I have to tie this back to Web3 and blockchain, I’d say I mostly want an assistant to handle “trading” tasks for me.
I don’t like the idea of being glued to a screen for trading in 2026. But I don’t need just any trading bot; I want something more complex that trades like me, following my strategy, leaving me with just occasional checks. I hope I can get to that point with AI assistants.
Of course, I’m not an expert, just a user. 🙂 A research and summary assistant isn’t all that complex, but the products and infrastructure in this area do allow for creating AI assistants for more complicated tasks.
I’ve already seen a few different and new applications pop up in my LinkedIn feed, like FinBot for fundamental analysis or Meleon AI, which helps you create your own more complex AI assistant.
There’s even a great article by Enis Erdem Yurdatapan on creating a personal AI assistant:
Kişisel YZ Asistanı Oluşturma Kılavuzu
Meanwhile, with news about OpenAI, Anthropic, and the Pentagon making the rounds, I’m starting to wonder if I should move all my AI support over to Claude. But that’s probably a topic for a longer, separate article. 🙂
Note: I’ve been using Claude for just a few days, and we’re already experiencing some performance issues. But I’ve read comments suggesting that these might be due to power outages and other disruptions in AWS system rooms in certain regions, given the recent geopolitical events. I hope that’s the case.