Entropy☃️Chase

Finished an article summarizing some tips I learned to write faster pandas code! Notebook included! [buymeacoffee.com/entropychase/f…] For me, this led to big speed improvements, esp for tasks that require row by row iteration!

@VBruttel @gadboit2 I think I was thinking about this exchange:

The pattern is one thing, the chance of that being natural is somewhere between 1 and 0.1%. The key finding is that these sites contain way more mutations than the rest of the genome (odds of this being natural somewhere below 1:10k and 1:1 mio.). @gadboit2 re-analysis confirmed that this is extremely unusual.

Since the COVID origins debate picked up steam recently, I decided to write a shorter summary of the basic facts we have available. I noticed that not everyone has been caught up to date with these details. Hopefully this improves the discourse! @entropychase/6-facts-for-a-more-informed-debate-on-covids-origins-1c8c1b01ca6a" target="_blank" rel="nofollow noopener">medium.com/@entropychase/…

@MeganSuspended appreciate the kind words!

@EntropyChase Excellent article. I’ve had to speed read thru the end as I’m sitting in the car in between jobs as taxi mum - but just wanted to say thank you, for what seems like a measured and logical summary of the ‘unknown’ origin. Great job.

@GaryMarcus how does Pangram do?

The book of Genesis, 84% created by AI! 🤣🤣🤣 Happy Easter/Happy Passover!

This song grooves. Great drumwork too imo. @KrisAbdelmessih

Went driving yesterday when it was really sunny, listening to this song, and all was healed. open.spotify.com/track/1da6XFL8…

@Roper_Lab @EMM_386 @DavidKlemitz @dkupiecki @PaulRic56123559 @Rebecca21951651 @angie_rasmussen @gdemaneuf @grok @MichaelWorobey @jepekar It seems like you're leaving out some important detail about what was found at the market.

@EMM_386 @DavidKlemitz @dkupiecki @PaulRic56123559 @Rebecca21951651 @angie_rasmussen @gdemaneuf @grok @MichaelWorobey @jepekar Analysis on sequences showed that the SARSCoV2/COVID19 RNA was found WITH samples that also had racoon dog DNA/RNA. It's additional circumstantial evidence, may identify the intermediate animal from bats. lots of these viruses around Asia #covidwasplanned nature.com/articles/s4158…

Rasmussen, the usual attack dog of the Morens/Daszak little cabal, doing her best to bail out the sinking zoonotic certitudes. She surely deserves a reward for her dedication.

@Roper_Lab @DavidKlemitz @dkupiecki @EMM_386 @PaulRic56123559 @Rebecca21951651 @angie_rasmussen @gdemaneuf @grok @MichaelWorobey @jepekar The problem is that the evidence we have today isn't supportive of a natural spillover. we still lack direct evidence for either theory. @entropychase/6-facts-for-a-more-informed-debate-on-covids-origins-1c8c1b01ca6a" target="_blank" rel="nofollow noopener">medium.com/@entropychase/…

@DavidKlemitz @dkupiecki @EMM_386 @PaulRic56123559 @Rebecca21951651 @angie_rasmussen @gdemaneuf @grok @MichaelWorobey @jepekar These coronaviruses circulate in the wild all over Asia. They recombine and jump species ALL THE TIME. That’s why scientists study them- to make drugs and vaccines to protect us #SARSCoV2 #covidwasplanned #virus

@JeromeAdamsMD That said, I think the denial of the harms of COVID in recent times is causing a lot of harm and impeding preventative efforts that could save many lives. I hope that changes. I compiled some research about this here. [reduce-airborne-disease.vercel.app/research-repor…]

@JeromeAdamsMD I feel differently. Even if folks believe COVID is nbd now (a point I personally disagree with), it still killed millions of people even early on. That's a big tragedy, and understanding the origin of a tragedy is not pointless or a "small fish." Important for prevention.

I can't square the cognitive dissonance of those who insist we wildly overreacted to COVID... yet also obsess over “lab leak” + “gain-of-function.” If a lab accidentally released a virus no worse than a bad cold, then don’t we have much bigger fish to fry? 🤷🏽‍♂️

I think @therobotjames will appreciate this one

this song is perfect for spring: open.spotify.com/track/2SDtvOPE…

@j_dekoninck @GregHBurnham @krishnanrohit Wonderful summary. Thanks for taking the time to make it. Excited by your work!

@EntropyChase @GregHBurnham @krishnanrohit But overall, I would say even without it, its clear that models are making progress: On our private IMProofBench (improofbench.math.ethz.ch/questions/lead…) consisting of research-level problems, models are also improving quickly. -- end of rant --

Question: what's a list of capabilities that AI has not meaningfully progressed on from 2023 till today? Writing well is my example, esp fiction, where it's gotten subject cohesion but the slope of the line is much flatter than eg coding. What else?

@hexane360 @gdemaneuf "virology of the virus" is an odd phrase, never seen that in the literature... regardless, the problem is that the evidence we have today isn't supportive of a natural spillover. we still lack direct evidence for either theory. @entropychase/6-facts-for-a-more-informed-debate-on-covids-origins-1c8c1b01ca6a" target="_blank" rel="nofollow noopener">medium.com/@entropychase/…

@gdemaneuf Believing the lab leak theory requires systematically throwing out all the data we have on the virology of the virus and where we found it. Then lab leak wins by default, hooray!

I am afraid Miller's argument shows that he has no real idea of what he is talking about. He makes a word salad out of 10% at most of the documentary material, and displays a profound misunderstanding of the ways the cases we sought for or reported across times, how they were classified, filtered, accepted or rejected. Sometime a bit of knowledge is the worst thing. In the case of Peter Miller, it resulted in him building up an argument on sand.

@GregHBurnham @krishnanrohit @j_dekoninck Big fan of Epoch's work with FrontierMath btw! I find it interesting scores on FrontierMath are so much lower than on 2026 MO problems. Do you think it's due to contamination, less novelty, or something else?

@EntropyChase @krishnanrohit @j_dekoninck That’s not to say there isn’t a lot of relevant background in the literature! There absolutely is. But “solving math problems no human presently has ever solved” is a separate question. Less progress there, but not zero eg epoch.ai/frontiermath/o…

@GregHBurnham @krishnanrohit @j_dekoninck do you think those other tests do a good job at mitigating contamination risk? I've noticed that it's an under discussed risk and results tend to get inflated, so it's just something i've learned to be cautious about.

@EntropyChase @krishnanrohit @j_dekoninck That, and there are many other examples pointing in the same direction, across other competitions and benchmarks

@VBruttel understandable. I considered discussing restriction site pattern you mentioned, but @gadboit2 noted that the chance that could arise via luck was higher than I initially thought. I might be way off there tho. I was unsure & tried avoiding anything potentially gray area

@EntropyChase Good article. I just found this sentence misleading👇 It is fine to focus on the non-molecular evidence, but one should then not conclude that there is only a low amount of evidence supporting a lab origin. IMO, that evidence is overwhelming.

@GregHBurnham @krishnanrohit @j_dekoninck yeah, this does seem like a solid improvement. I wonder if there is any reason to be concerned about data contamination (and how that risk compares to last year). hopefully unlikely if test was 6hrs after problems came out

@EntropyChase @krishnanrohit Yeah, not sure. @j_dekoninck could maybe shed some light re: odds something snuck through. These geometry solutions are just awful. But tbc even if you threw them out entirely there’s still a clear jump vs. last year.

@GregHBurnham @krishnanrohit this is promising! that said, the "human validation" methodology seems very strange. perhaps im missing something.

@EntropyChase @krishnanrohit They did much better on the 2026 version of the same competition. Reasoning on newly-devised math problems is advancing very quickly on multiple measures. matharena.ai/usamo/

well, this was awesome. RIP Steve Albini bobbyowsinskiblog.com/steve-albini-d…

cc @tryna_do_rite ever hear this one?

0:58 and beyond in this CSNY song is absolutely beautiful imo. Sounds like another world, makes me think of a floating castle in a far away land. open.spotify.com/track/3O4VF8UY…