Hi, I'm Brad, software engineer, long time HN lurker. I have been working on this project for a while and I think it is ready for more feedback. I have been calling it Snow Day Adventures, an app for k-12 students home from a school snow day that beats watching YouTube or Instagram.
One of the things I consistently used chatGPT for is creating bedtime stories for my two daughters. They were never very long, but we could customize them and ask for edits. It was a fun activity we could do laying down before bed.
Fast forward to just about a year ago when an administrator at my daughter's school reached out about AI. We had a meeting where we talked about the basics but I really had no idea how AI would work in K-12 education. There were a couple apps that they use, which were fine, but I felt like they did not take advantage of the technology. At the same time, I was not qualified to give them advice.
So I started building little apps here and there. I would test them with my daughters. Usually they would get bored or it would feel to much like homework. Sometime this summer I focused on generating engaging stories and came up with a workflow for generating choose your own adventure stories. After a couple iterations, I got to where the app is now: turn any story into a choose your own adventure.
I have loaded a couple public domain stories to start with. The idea is pretty simple: pick a story, you will see the first page and at the bottom you will see a link to the original story the author wrote and three generated pages. Once you pick a generated page, you will only see generated options (and you will probably see the options take ~10 seconds to load). You can also click the refresh button which will replace that page with a new generated one.
The generated pages are never as good as the originals, but it does make it more fun. Going back and trying different paths creates endless amounts of content. I think the bigger opportunity is to customize the stories to the students reading level and language abilities. I did a couple tests with bi-lingual story generation, and generating from a strict lexile reading level; both showed some promise.
I know it does not technically need a user to log in, but I am a little worried about story generation getting out of hand from bots. If anyone has any ideas for preventing a search engine bot from clicking every link on a page that generates new content, I would be happy to find an alternative. Maybe you have to log in to generate new content but the already generated content is public? Let me know what you think.
I’m Brad Davis, a senior software engineer and product-focused developer with over a decade of experience building scalable applications, data pipelines, and AI-driven tools. At Ginkgo Bioworks, I led the development of a 5.6 TB genomic and metadata data lake supporting 120+ researchers, integrating AWS and Snowflake to streamline scientific workflows. My work has spanned creating React.js applications for lab teams, implementing machine learning pipelines, and optimizing developer workflows with AI-assisted code generation.
I’ve contributed to open-source AI projects such as RooCode and built my own educational platform, [Snow Day](https://github.com/thecolorblue/snow_day), which generates vocabulary quizzes using LLMs. My career has included both engineering and product management roles, giving me a unique ability to bridge technical and business requirements. I enjoy designing systems that make data more accessible, applications more intuitive, and teams more effective.
He would cherry pick from a random branch that was never merged back in, add two commits of style changes inconsistent with the rest of the repo, then force push.
Location: Cleveland, Ohio, US
Remote: Yes
Willing to relocate: No
Technologies: Typescript, Python, Postgres, LLMs, computer vision, bioinformatics, ecommerce, SaaS
Résumé/CV: https://docs.google.com/document/d/1uuq8KB2uVcMB0l5zJR-8Nruca_OgRQpIkVX1Dv-p9s4/edit?usp=sharing
Email: brad.bdavis1@gmail.com
Hello! I am software engineer with 13 years of experience with javascript, and 6 with typescript and python. About half of my career has been as a front end developer, and the last 6 have been as a full stack developer. I have done a little bit of everything, from ecommerce, to software as a service, to bioinformatics and genomics R&D.
I just ran a test giving the same prompt to claude, gemini, grok and qwen3 coder running locally. Qwen did great by last years standards, and was very useful in building out boilerplate code. That being said, if you looked at the code side by side with cloud hosted models, I don't think anyone would pick Qwen.
If you have 32gb of memory you are not using, it is worth running for small tasks. Otherwise, I would stick with a cloud hosted model.
That should remain true for foreseeable future. A 30b model can't beat 300b. Running 300b model locally is prohibitively expensive. By time it would be feasible cloud will also move to 10x larger model.
At 4 bit quantization the weights only take half the RAM. You need a good chunk for context as well, but in my limited testing Qwen3-30B rand well on a single RTX 3090 (24GB VRAM).
Beyond Meat has seen less growth than their some of these examples, but my conclusion is still the same: it would have been better for them to find their footing as a private company, then use the IPO money to expand, rather than to pour the IPO money into R&D. Publicly trading stock causes more overhead and restricts the companies ability to pivot. Investors, founders, and employees are able to cash out before the real value has been created. The money generated does not guarantee longer term success or better long term investments as much as they thought it would.
I am still bullish on many of these companies but if I were to get in my time machine back to 2019, I would have avoided investing.
Not nostalgia. Putin is what he is. Movies about hitler aren’t nostalgic. Germany under Hitler was an “enemy culture” no? We must never minimize the global disdain for Russia under Putin. How many have died? Nostalgia just isn’t the right word, or it shouldn’t be.
One of the things I consistently used chatGPT for is creating bedtime stories for my two daughters. They were never very long, but we could customize them and ask for edits. It was a fun activity we could do laying down before bed.
Fast forward to just about a year ago when an administrator at my daughter's school reached out about AI. We had a meeting where we talked about the basics but I really had no idea how AI would work in K-12 education. There were a couple apps that they use, which were fine, but I felt like they did not take advantage of the technology. At the same time, I was not qualified to give them advice.
So I started building little apps here and there. I would test them with my daughters. Usually they would get bored or it would feel to much like homework. Sometime this summer I focused on generating engaging stories and came up with a workflow for generating choose your own adventure stories. After a couple iterations, I got to where the app is now: turn any story into a choose your own adventure.
I have loaded a couple public domain stories to start with. The idea is pretty simple: pick a story, you will see the first page and at the bottom you will see a link to the original story the author wrote and three generated pages. Once you pick a generated page, you will only see generated options (and you will probably see the options take ~10 seconds to load). You can also click the refresh button which will replace that page with a new generated one.
The generated pages are never as good as the originals, but it does make it more fun. Going back and trying different paths creates endless amounts of content. I think the bigger opportunity is to customize the stories to the students reading level and language abilities. I did a couple tests with bi-lingual story generation, and generating from a strict lexile reading level; both showed some promise.
I know it does not technically need a user to log in, but I am a little worried about story generation getting out of hand from bots. If anyone has any ideas for preventing a search engine bot from clicking every link on a page that generates new content, I would be happy to find an alternative. Maybe you have to log in to generate new content but the already generated content is public? Let me know what you think.
Thanks for taking a look.