I made a web game inspired by Geoguessr and Wordle, where you get shown a 3D model of a random human protein each day, and you have to triangulate its gene name using similarity clues.My background is in wet lab molecular biology and I intend this game to be engaging mostly to other biologists. But if you're outside the field, I'm interested to know if you can still solve it with browser use LLMs, and if you learned something interesting doing so. Let me know what you think.
I made it with Claude over the last 2 months. My coding experience is limited to basic python data analysis and figure making. I've seen people online asking, "Now that we have coding AI, why isn't there a deluge of awesome AI-generated apps made by non-coders?" - if this sounds like you, check out Geneguessr to understand what a web app by a non-coder looks like.
I might write more about the process if there's a demand, but what really unlocked the project for Claude was Linear MCP, where it could put each individual issue on a shared Kanban board. This, and Playwright MCP for testing on live site, were the two workhorses that got me through this. For bugs Claude couldn't one-shot, Linear was great for consolidating issue information so that I could dump it into ChatGPT Codex - it would usually think for like half an hour, output very confusing explanations, but the bug was gone.
Game is free, no log-in required, sorry if you run into any mobile bugs - didn't test it much there.
https://geneguessr.brinedew.bio/
▲boldlybold59 minutes ago
[-] Great work! I really like the interface here, I think you put a lot of taste into the way this all came together. It's quite hard (appropriately so!) I would love to see a experienced structural biologist go at this.
reply▲brinedew39 minutes ago
[-] Thank you. I started the development with the interface first, basically making a mockup of how the finished game should look like, and then prodding LLMs with a stick to make a backend that would support this interface without crashing.
Making a nice-looking web GUI without knowing relevant vocabulary was a very clunky process in comparison to code pipelines, basically just pasting screenshots into the chat window and asking LLMs to "line up stuff properly", which they still couldn't manage to do in places.
Huh. Is this kind of manifold something that a group of humans can now identify on sight and associate with a part of a DNA sequence? That's pretty spectacular. I have a friend who worked for years on AlphaFold, but wasn't aware that people had gotten to this level of confidence in visually identifying proteins.
reply▲My dream is forming a team of Tetlock-style superbiologists who can identify gene names on sight like Rainbolt, beat prediction markets on shorting biotech stocks, and smell out pre-cancerous cells like it's the final round of amongus.
reply▲Hahahaa ... I kinda love this, along with humans who can sit at a piano and play a song they've only heard once, or cook a complicated dish they only tasted once. That seems like the great Turing test... zero-shot humans.
But seriously, there probably are a few people who can see genes from proteins like that, faster than a whole datacenter of GPUs. Putting together such a brain trust could be invaluable.
wigglewoggle7 hours ago
[-] I went straight to cd-4 and was crushed to find out I didn't get it in one guess
reply▲RestartKernel1 hour ago
[-] This seems super well-done, albeit completely impossible for laymen. Great work, I'd love to see someone with the necessary knowledge play this.
reply▲Clicking the leftmost menu icon raise the following error:
Geneguessr encountered a problem:
runtime-error
SyntaxError: redeclaration of const NAVIGATION_START
That's the frontpage link, fixed it, thank you.
reply▲For a moment thought this was gonna be some ethnoguessr clone
reply▲lovegrenoble8 hours ago
[-] Dayly, that means one puzzle per day? And how many of them in total?
reply▲You can play more than once in the practice mode. It turns on after you complete the daily puzzle.
There's almost 20k entries in my human gene database. You can choose any of them as your guess, but the "gene of the day" is chosen from a subset that has decent 3D structure coverage (meaning, not falling back to alphafold2 for visualization).