541–550 of 1001 posts
I must say it has been more challenging than what I though it would be, specially if you are looking to put it onto production. I'm doing it for fun though.
Nothing published yet, I'm not sure if it will ever be. What do you think ?
Finding work opportunities which enable me to grow in my personal interests (material science, physical chemistry, applied physics, additive manufacturing). I feel compelled to support scientific research, such as material informatics - or maybe automating labs with robotics.
It would be thrilling to manage infrastructure for scientific computing workloads (as an idea).
Honestly, the tech stack + role matters far less than to me than what we're doing, and who I'm doing it with. I'm motivated by curiosity and the desire to learn. I am tired of reading scientific literature solo with nowhere to apply the knowledge, except personal engineering projects in my workshop by myself. I am not an academic, just a full-stack/polyglot software engineer. In the last 3 years my interest in this stuff just exploded.
Additive manufacturing is another field that I'm very interested in, and would love to work in.
10 days ago I ended a 2.5 year relationship which was not healthy for me to be in. I'm recovering physically from chronic stress. Everyday is getting better.
Planning to move to another country soon. Currently I'm in a non-EU country in Europe. I would be very interested to live in Germany, but am open to other possibilities - the work opportunity would be the driving factor. I'm a US citizen, 38 years old. There are (seemingly) no communities and resources for me to grow my personal interests here. The country is falling apart due to an extremely corrupt government.
Technically?
Been using agentic coding tools the past 7 months.
Recently, I built a slicer for a clay paste extrusion 3d printer (which I built for fun). It aims to generate a continuous toolpath (including between layers) that does not self-intersect, from an STL model. Stress tested it with complex geometries.
The slicer involved a lot of computational geometry solutions. Despite my graphics programming experience, it would not have been possible for me to build this without the help of LLMs, and more crucially - access to publications from research groups dedicated to this space. Reading literature about toolpath planning for industrial robotics was thrilling.
I learned a lot about how to do R+D in a somewhat unfamiliar domain. Feels like I pushed the limits of what's possible with LLM coding on this project, after trying many workflows, models, and techniques. Will be posting some insights on my soon-to-be-released blog soon.
It was a whole lot of fun implementing a zillion approaches to solve this problem, and seeing the results quickly visualized with matplotlib. If anyone's interested, I can share some images here.
I'm also building a PKM (personal knowledge management) system based on a graph db that helps me keep up with all the research + projects + daily activities in my digital life. I've been an org-mode + emacs user for 8 years, and it's just become increasingly obvious to me that I need something more powerful. Trying to coerce a relational db to support multiple inheritance w/labeled nodes + edges, while getting normalized tables is not pretty, and all the join tables will cause performance hits. This project is the definition of scope creep, but it's a personal project, so I'm okay with it.
Documentation and logs of many past/current projects are going up on my soon-to-be released blog soon. I've written many draft posts, and it's already deployed. By next month I expect to be able to show everything that I've mentioned here. There are some repositories if anyone is interested in previews of anything mentioned.
Socially?
I'm trying to be more open and transparent online - to reach out and find people who I could talk with about shared interests, and potentially build something together.
Hacker News has been a consistent source of inspiration in my life since I discovered it in 2012, and I want to start contributing to the discussions and inspiring people in any way that I can. I've been a lurker for far too long.
Can send links to my github and unfinished blog (with drafts for some past projects and more about future projects + topics of interest), if anyone reaches out. Next month I intend to be posting many links, and to use my "real" username. Just don't want LLMs scraping the personal bits of this post.
Still working on https://periplus.app , and recently started to see some traction. It's an environment for open-ended learning with LLMs. Something like a personalized, generative Wikipedia. Has generated courses, documents, exams and flashcards. Each document links to more documents, which are all stored in a graph you grow over time.
I'm working on a new app for creating technical diagrams - https://vexlio.com . It's an area with some heavyweight incumbents (e.g. Visio, Lucid) but I think there's good opportunity here to differentiate in simplicity and overall experience. I'm still in the fairly early phase, and I suspect I haven't quite found the best match of features to customers yet. From a dev perspective this area has a ton of super interes…
Stacktape is a PaaS that deploy to user's own AWS account.
v3 adds many new features, but namely the ability to generate IaC config directly from code, by analyzing the user's repository (both deterministically and using multiple AI techniques).
For example, if it assumes your application is a Web API that uses Postgres and Redis, it will create a Stacktape IaC config that deploys Fargate container, load balancer, Aurora Serverless v2 Postgres and Elasticache Redis (behind the scenes it will also configure things like networking, VPC, security groups, IAM, etc.)
Launching this weekend.
The big trick or the language is that it doesn't hide the pipelining you have to do to up your FMax, instead, you can manually add register stages in the places they're important, and the compiler will synchronize the other paths.
A really neat trick with this pipelining system is that submodules can respond to the amount of pipelining around them (through inferring template parameters). This way the programmer really doesn't have to think about the pipelining they do add. Examples are a FIFO's almost_full treshold, inferring how many simultaneous state there needs to be for a pipelined loop, inferring the depth of BRAM shift regs, etc.
After 10 years in defense tech, watching missile attacks in Ukraine and the Middle East made it clear how little most people really get about air defense. So I'm builiding this simulator which drops you into the operator’s seat. You can test out different scenarios and build an air defense network against various types of threats (stats from real world). Also have Ukraine, Israel-Iran scenarios. https://airdefense.de…
And I’m looking to productise a bookmarking app “Tsundoku” I built for myself and have been using for a year https://bsky.app/profile/gingerbeardman.com/post/3ls2ymul33s...
Still working on https://periplus.app , and recently started to see some traction. It's an environment for open-ended learning with LLMs. Something like a personalized, generative Wikipedia. Has generated courses, documents, exams and flashcards. Each document links to more documents, which are all stored in a graph you grow over time.
A few courses I generated using above:
- https://dev.to/freakynit/network-security-cdn-technologies-a...
- https://dev.to/freakynit/aws-networking-tutorial-38c1
- https://dev.to/freakynit/building-a-minimum-viable-product-m...