exploring o3 and accidentally building a product

LOG 202502022330 #log

so i was doomscrolling X and found out that openai reasoning models can show their chain of thought just like deepseek. except not really lol, easy to get misled on the internet.. but as has been the case with my dev explorations lately, this happy accident turned into something i think i might love.

invisionapp was a prototyping and collaboration platform that i used for several big web projects over the years and i found the workflow to just make sense. it bridged a big communication gap for me because you can just see how stuff is supposed to work. this makes it easy to mock up site flows. the best part was that it tracked changes over time. as a lover of git and version control... i was really sad when the company got acquired and all of the amazing functionality abandoned.

the site isn't even there anymore.

getting o3-mini, and o1, running in rahn.chat was relatively straight forward, the docs are there. long as claude isn't hallucinating about gpt-3.5-turbo (really great habit lol 🫠) and you pay attention to stay within context limits, adding another openai model doesn't take much.

but i went into it thinking that i'd be able to show the chain of though with more detail, that wasn't the case so i decided to just try building something with it and see where it goes.

enter invisionrip, i crafted a juicy brain dump outlining the core feature set; sites > pages > images, build mode with hotspots, preview mode with interaction, review mode with commenting/tasks, all with versioning. and iterated with o3-mini to develop a detailed project spec and development plan.

two rounds of follow-up questions, and i had what i needed and the repo was live. i cheekily set cursor to o3-mini and dropped the plan into the readme.md and set o3 off on the build.

lots of stumbles but i had build and preview modes mostly working.. though the ui was rather atrocious. o3-mini was lazy, it took a lot of prodding and 46 commits in i was getting frustrated by the lack of progress and amount of rework i was running into.

it was time for claude.

the progress i made with o3-mini inspired a lot of confidence, and i was already thinking of domains the this could live on.. invision.rip wasn't it.

✨ yuckie.dev used to be my obsidian publish domain; and now it's gonna be this 😂

with the refined prompt in hand, and yuckie.dev sitting dormant and urging me ahead. i fired up the new repo and set to work with claude. we had some trials and tribulations, as is usual with claude, sometimes i have to yell or sob because i don't have the heart to discard and start again lol.. we pushed through and 59 commits in, we're two little bugs away from needing to deploy this and start working on v2 🥵. this went so fast.

need to start laying out a roadmap for the future...

anyway that's what i got up to this weekend and why i've only slept like 10 hours 😞 need to charge up for the week.. always nice to have projects like this to fuel the grind mindset

we can sleep when we're dead ✌️