Agents for Humans - Ideating and Information Gathering
Information gathering around Agents for Humans hackathon and the AWS Strands SDK as well as ideating a concept to build out.
My recent background
My sons are two and a half and almost two months and having moved to Japan to have them I have spent more time being a Dad than I have crushing work and coding goals. This is a good thing my little boys are the most important thing in my life. But I still love coding and the landscape over the last two years has been interesting to watch and occasionally dabble within to say the least. Vibe code a little project here and there and wondering what software engineering looks like for my old colleagues back in Australia has been mostly all I have touched in that time.
Translating that into this Hackathon
This hackathon piqued my interest by being more than just a little app vibe coded by another agent it lets me try to customise as an agent specifically for a task that has access to a strong LLM backing it up. I also always love building tools that make life easier for humans. The crux of this being I'm a human and I can pretty much just look to make something that "scratches my own itch" and makes my own life easier or better. The win here being that even if I don't end up being successful in the hackathon or running it commercially there is a good chance I'll be able to use the tool at least until something better comes along. To me it also makes the process pretty fun building towards something I can use. Also asking the customer or product owner what they want in this scenario is easier assuming I can keep my own scope under control - not my greatest skill based on past experiences.
An itch to scratch
So for this project I've always like the idea of keeping smart notes or developing a second brain potentially with Obsidian or even a physical slip case like the Zettelkasten Method. But my love of reading and researching and even having (if not novel) my own ideas around the concepts has never pushed through that barrier of keeping the notes or more to the point managing a system that makes me want to record the notes. Obsidian was abandoned after a few notes recorded and likewise for my collection of paper cards and storage boxes for them. The overhead of managing it just gets away from me and I get back to enjoying reading articles having interesting thoughts about them. The sad part is that for the most part these ideas and concepts are forgotten within a month or year like they never happened - let's see if we can fix that.
Information Gathering - Hackathon
The hackathon timeline looks good, it gives me a short focused window before we have a family holiday trip back to Australia. Given Australia is excluded from the Hackathon it is lucky that I am living in Japan for this one. As already said the concept of making life easier for humans (like me) is an easy win for stepping into a hackathon, as is looking at customising an AWS Strands Agent to help with this process.
Information Gathering - AWS / Strands
I don't know if my first point of call was Claude or YouTube trying to understand Strands SDK. I think it was a little of both, the AWS Developers videos in this playlist were pretty good at getting me thinking about what a Strands agent would be - an agent backed by an LLM with tools that I would choose to give it and perhaps an MCP and probably built in system prompts for customisation. But so far the playlist is kind of getting me to a "hello world" example or perhaps checking my age. Cluade on the other hand gave me some quick clues that the AgentCore would be the AWS hosting that we could eventually run our agents on and Bedrock hosts some of the LLMs and probably the default model that we will access when getting a hello world app running.
Stay Tuned
I see a "hello world" strands agent coming not long after lunch, if the little baby lets me get away with it. And then onwards to our slip case goal.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article