https://geneguessr.brinedew.bio/
Inspired by Geoguessr and Wordle, this is a free web game where you get shown a random human protein each day, and you have to triangulate its gene name using similarity clues.
My background is in wet lab biology and I intend this to be comprehensible 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 from 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. If you've been asking yourself "Now... (read more)
Actually, I just tried Claude 4.1 Opus with extended thinking on a fresh account, and I think it gets the closest to what I myself would say.
... (read more)Scientific Validity (Score: 7/10)
The document demonstrates solid grounding in real biological phenomena.
[...]
However, the document selectively presents evidence supporting its thesis while largely ignoring counter-examples and alternative explanations. For instance, naked mole rats have exceptional cancer resistance AND exceptional longevity, contradicting the strict trade-off narrative. The document also oversimplifies complex mechanisms - aging is multifactorial, involving mitochondrial dysfunction, protein aggregation, and metabolic changes that aren't all directly related to cancer suppression.
Scientific Novelty (Score: 2/10)
The cancer-aging trade-off has been extensively studied in evolutionary biology and gerontology for decades
[...]
Classification:
ChatGPT 5 Thinking is a bit more critical and thought for far longer (5 minutes and multiple online searches)
Caveat: this temporary chat window still has access to my account's saved memory, but it didn't seem to explicitly come up in a thinking trace:
... (read 402 more words →)Bottom line (validity).
With careful corrections and caveats, the essay’s backbone—multilevel anti-cancer governance and its costs—is scientifically defensible as a major contributor to aging, though likely incomplete as a unified theory. I would rate overall validity as moderate-to-high contingent on clarifying the overreaches above.[...]
Scientific novelty
What seems novel: the piece integrates disparate literatures (Peto’s paradox, tissue governance, immunosurveillance, senescence) into a single “managed fragility” lens and draws operational predictions (prioritize genome stability
As a test, I tried using your step 1 prompt to see what LLMs think about one of my crank-flavored essay drafts lying around, where I claim that aging extends lifespan.
(I added a request to use online search explicitly).
These are excerpts from Gemini 2.5 Pro (fresh account):
... (read 556 more words →)This is an excellent and insightful piece of scientific writing. It synthesizes complex ideas from evolutionary biology, oncology, and cell biology into a coherent and compelling narrative.
[...]
Scientific ValidityTo a very large extent, this project is scientifically valid. The author demonstrates a strong command of the subject matter and accurately represents established scientific concepts.
[...]
The project is a successful synthesis of existing knowledge, not a presentation of new
To give an example of how disastrously incompetence can interact with the lack of personal accountability in medicine, a recent horrifying case I found was this one:
Doctor indicted without being charged for professional negligence resulting in injury
According to the hospital, Matsui has been involved in a number of medical accidents during surgeries he performed over a period of around six months since joining the hospital in 2019, resulting in the deaths of two patients and leaving six others with disabilities.
Matsui was subsequently banned from performing surgery by the hospital and resigned in 2021.
This youtube video goes over the case. An excerpt:
... (read more)January 22nd: Dr. Chiba's heart sinks when he learns that Matsui has
Thanks for checking it out.
It took me many iterations to settle down on the exact logic. At first, I started with just the HiG2Vec GO similarity embedding. It did alright, but I didn't like how the same protein family gets wildly different scores based on just pathway participation or tissue expression. I added ESM2 sequence-based embedding to tame this inconsistency. It also resulted in the "your guess is top-9 similar" hint to be arranged in the order of increasing sequence similarity, which is a nice bonus for late-game triangulation.
I tried making a shared embedding out of two separate ones, but ran into statistical issues with how differently I needed to normalize them.... (read more)