Sabitlenmiş Tweet
@MorlockP @ark_press Post terraform Martian water depth makes quite a pleasant little inlet from the northern sea
English
graffle
773 posts
graffle
@graffle4more
Exoplanet mapping, energy systems modelling, astrobiology. Interested in automating energy/agri tech with GIS based digital twinning.
Katılım Kasım 2022
1.8K Takip Edilen216 Takipçiler
graffle retweetledi
Very cool! I've followed you and would be thrilled to see more of your work as it comes along. I spent some time last year adapting Casey's models to CUDA to run Mars terraform sims, and I'd really enjoy engaging more in this field. Perhaps collaborate on some papers at some point?
English
@graffle4more @Robotbeat My models. I'm working very slowly as its my 10th priority or so on actual hydrological sims
English
No, my own thoughts as a query. I'd value your feedback. I think I'm getting hung up on what we should allocate as the unit of information being valued. Should we place it at the level of "model being checked against data"? And then rate the information value of given models?
Also unsure I'm correctly treating the interaction between p <= 0.05 claims the a 50/50 replication rate properly.
English
“Jo Boaler is a professor of education at the Stanford Graduate School of Education, with an enormously influential body of work …. Her work got Algebra removed from middle schools across the Bay Area.
It is some of the most incompetently or dishonestly conducted research I have seen in a decade as a journalist.”
English
@grok think deeper: the unit of information in each paper is a model being advanced with the claims that if >=95 of 100 observations fit the model then the model should be considered a proven fact (unless p hacking).
And the distribution of those models within the field matters when evaluating the information content. How can we apply individual selection criteria based on metadata to identify the sections of the field that has information value?
English
Yes, it's a solid analogy. For a binary outcome like "replicates or not," Shannon entropy peaks exactly at 50/50 (1 bit max uncertainty). That models a field where published results are as informative as fair coin flips—pure noise on average. A randomly selected paper then carries near-zero expected information value about underlying reality.
The chart's social-science rates (~40-60%) fit this well, explaining the replication crisis. Not every paper is worthless individually, but the literature as a whole is.
English
Perhaps there’s somewhere in middle that represents a useful unit of information between "entire article" and "individual words randomly picked to state a fact". What if we treated each model as the unit of information we're measuring the value of? Where a model is proposed with the argument that >95 of 100 samples measured from reality fits to the model, the presented model aligns with the proposed facts stated in the hypothesis, therefore the fact should be considered true.
English
I would say a 50/50 replication rate is not Shannon noise, because you also have selection of facts under question. If you could make up an arbitrary fact (say by throwing dice for words in a sentence) and then have a 50/50 chance of getting that published, *then* you would have Shannon noise.
English
We should make more waterfront property
bone@boneGPT
no matter how good the future is, they aren't making any more waterfront property
English
Is saying "at a 50/50 replication rate, these fields have reached maximum Shannon's information entropy and therefore are purely noise and therefore a given paper randomly selected has no information value" a valid hypothesis? Think I'm missing something about the distribution of the replication.
English
I thought that outcropping looked familiar! Perhaps we could publish a follow on paper together exploring the shoreline ecosystem types at varying water levels? Much of the depth modelling is already done and sitting in my geodatabase from the CUDA adaption of your original flooding scripts.
English
Researchers just uncovered the strongest evidence yet for a vast ancient ocean on Mars — a massive coastal shelf hidden in the northern hemisphere.
The feature is a wide, flat band of terrain (hundreds of miles across) with gentle slopes and low curvature, wrapping around the northern plains. It closely matches Earth’s continental shelves, which form over long periods as rivers deposit sediment into standing oceans and waves shape the coastline.
Using computer simulations that “dried up” Earth’s oceans to identify the most durable topographic signatures, the team then scanned high resolution Mars orbiter data and found the matching shelf signatures there. Shorelines alone are subtle and sit at inconsistent elevations due to billions of years of sea level change, erosion, volcanism, and wind — but a coastal shelf is far more stable and would only form around a long lived body of water.
Author Abdallah Zaki explained: “It’s very fundamental. If there is an ocean, there must be a shelf. This is a more stable topographic signature.”
The discovery, published in Nature on April 15, 2026, strengthens the case for a stable ocean early in Mars’ history and points to promising new targets for rovers to hunt for ancient coastal sediments that could preserve signs of past habitability.
📸 A. Zaki
English
graffle retweetledi
Loved Project Hail Mary and want more hard sci-fi with momentum?
Start here: The Martian, Rendezvous with Rama, Tau Zero, Children of Time, and more.
Real scientific pressure. Low hidden-agenda drag.
clearshelf.xyz/shelf/hard-sci…
English
graffle retweetledi
@IterIntellectus At 50/50 replication, is it fair to say these fields are operating at maximum information uncertainty with peak Shannon's entropy? A given paper is a coinflip, meaning the information content is just noise and no signal?
English
no
the takeaway is that half of published studies across every discipline don't replicate
it's just a coin flip
Dr Kareem Carr@kareem_carr
This new study in Nature is kind of a bombshell. The takeaway seems to be that you should trust Psych research more than Economics!
English
At 50/50 replication, is it fair to say these fields are operating at maximum information uncertainty with peak Shannon's entropy? A given paper is a coinflip, meaning the information content is just noise and no signal?
Dr Kareem Carr@kareem_carr
This new study in Nature is kind of a bombshell. The takeaway seems to be that you should trust Psych research more than Economics!
English
.@uppityhobbit I was thinking about bringing something of yours next time I'm in town. Existing offerings were flowers, coins, letters, and a ring. Perhaps I could attach one of your designs to a small base token or coin; a smaller version of the ornament you make.
English
graffle retweetledi
アメリカでこの『ピザの上にピザをトッピングしたピザ』を見たとき「こんなヤツらに勝てるわけない…」と思った
日本語