Built in Rust with btleplug for BLE. The hardware is a $25 iDotMatrix 64x64 LED panel from AliExpress — you send it GIF frames over Bluetooth and it plays them on loop.
The interesting part ended up being the state machine. Claude Code fires PermissionRequest and PreToolUse within ~100μs of each other, out of order, and PostToolUse fires after every tool call (not just approvals). Getting the display to not flicker took more iterations than I'd like to admit.
I use LLMs to find items for me all the time and it's great. I still have to use my human capacities for consideration and kindness as well as my deep understanding for the people in my life who I care about to suggest what gift I might want to get them. I still need to care. The people in my life have very specific needs and finding the right vendor for a gift involves a lot of searching and reading. Not to mention navigating the brilliant tariffs that have recently made shopping such a joy. LLM tools turn this 3-5h process into a 5 minute exploration.
I'm glad to see this being made mainstream. Hopefully nobody imagines that this will let them care less about people in their lives, since we need a lot more people to care about each other in my opinion.
I'm sorry that has been your experience with agile. That doesn't sound very good. In my experience, if agile rituals take more than three hours per 2-week sprint, something is not aligned with the team or the project. Maybe the necessary skills aren't there. Maybe the team doesn't trust each other. Perhaps the project plan is not aligned with business goals, etc. None of these problems can be solved by adding more agile rituals.
How I prefer to run it:
- 1h per week alternating between estimating and planning
- (optional) 1h per sprint for backlog refinement
I like to keep two sprints worth of work in the backlog. More than that, the spec will change too much, and you will re-estimate. When a project has just started, or the requirements have become understood differently, to get four weeks of backlog work may require additional refinement. I typically do the bulk of this refinement myself (as a tech lead), and I try to avoid asking the team to have any extra meetings beyond what I mentioned above.
It can be good to have retros sometimes, but it's better to have a team that feels safe to share and discuss issues constructively without a mediator.
I'd also like to advocate for 1:1s. We are all humans and social organisms. Ideally, these 1:1s are 80% unrelated to a work task.
This reminds me of a Douglas Adams quote - although he was talking about technology, it seems appropriate here also :)
“I've come up with a set of rules that describe our reactions to technologies:
1. Anything that is in the world when you’re born is normal and ordinary and is just a natural part of the way the world works.
2. Anything that's invented between when you’re fifteen and thirty-five is new and exciting and revolutionary and you can probably get a career in it.
3. Anything invented after you're thirty-five is against the natural order of things.”
As college student who grew up watching Zuck "get rich easy", I was easily taken in by the Hipster startup culture. Very unfortunately for me, I only found out about this fantastic site (Hacker News) about 5 months ago. By this time, the hipster ethos was rampant and I was swept up by it. My ego deeply resonated with their fancy promises and extravagant buzz words.
However, there is a thread of hope. An undercurrent of truth which I have recently started to uncover within the layers of crud. I just read Paul Graham's essay Beating the Averages last night and realized that I have no idea what the hell I'm talking about.
Lately there has been a growing thread of dissatisfaction with Hacker News. It feels as though the original community has become diluted by an influx of folks like myself, eager to make their mark but too naïve to really do anything constructive (and too naïve to realize we’re naïve). But we are learning.
So, Hacker News, thanks for putting up with us noobs! I realize now that I won’t be the next Zuck, but if I keep listening and learning from HN, I just might make a dent in the universe.
The interesting part ended up being the state machine. Claude Code fires PermissionRequest and PreToolUse within ~100μs of each other, out of order, and PostToolUse fires after every tool call (not just approvals). Getting the display to not flicker took more iterations than I'd like to admit.
MIT licensed: https://github.com/terraboops/lfg
brew tap terraboops/tap && brew install lfg
Happy to answer questions about the BLE protocol, the sprite system, or the state machine!