iiinova — where your digital self keeps evolving

Nicholas Mancuso @nmancuso.bsky.social · 7h

Builders/creatives clasping hands meme

Nicholas Mancuso @nmancuso.bsky.social · 11h

"In any case, what really matters is that the true inner sanctum of human creativity hasn’t been breached and probably never will be, and also, that Sam Altman and Dario Amodei are contemptible little nerds." scottaaronson.blog?p=10169

scottaaronson.blog

The Mathocalypse

Last night my 9-year-old son was taunting my wife, complexity theorist Dana Moshkovitz, as follows: “mommy, I heard you got cooked! I heard that a robot solved the math problem you worked on …

010

Reposted by Nicholas Mancuso

Will Ratcliff @wcratcliff.bsky.social · 14h

Out today in @nature.com: our review on how single cells evolve into multicellular organisms. With Ozan Bozdag, @kaitong25.bsky.social, Peter Yunker, and @matthewherron.bsky.social. rdcu.be/9ZuiKow9twKa For an overview, check out the video below.

620390

Reposted by Nicholas Mancuso

Quanta Magazine @quantamagazine.org · 20h

Last month, Dor Minzer and his graduate students rushed to post a milestone result in theoretical computer science, hoping to get ahead of a rumored AI proof. They were just in time. OpenAI put out a proof of a related conjecture yesterday. www.quantamagazine.org/as-ai-closed...

quantamagazine.org

As AI Closed In on ‘Unique Games’ Proof, Researchers Raced to Beat the Machines | Quanta Magazine

In the shadow of a rumored AI proof of one of the biggest problems in their field, three computer scientists rushed to publish their own milestone result.

0166

Reposted by Nicholas Mancuso

GOLIKEHELLMACHINE @golikehellmachine.com · 19h

all the promo stuff for this show is so much fun (for a change) — i will have more official flyers with details later, but, lol

1012016

Reposted by Nicholas Mancuso

bioRxiv Genetics @biorxiv-genetic.bsky.social · 21h

Placental cis-eQTL analysis identifies preeclampsia-dependent genetic regulation at promoters www.biorxiv.org/content/10.64898/20…

031

Reposted by Nicholas Mancuso

Ben Recht @beenwrekt.bsky.social · 21h

I have no idea if this one is true, but all this result proves is the absurdity of big-O notation.

3445

Nicholas Mancuso @nmancuso.bsky.social · 22h

Ive been out of the CS game for far too long to try and judge the veracity of the proofs. Given UGC longstanding importance to cs theory it'll be valuable if it holds up, but I'll wait until (if at all) it's verified by experts.

000

Reposted by Nicholas Mancuso

Hélianthe @helianthe.net · 07/10/2026

i think it's fundamentally because closed source software is sad in a mirror universe where openai released the weights, training data, and logs reproducibly, then we could all do science together on what's actually a miraculous machine instead they want to be secret wizards, so it's sad

2254

Reposted by Nicholas Mancuso

Carl T. Bergstrom @carlbergstrom.com · 07/10/2026

Terry Tao @teorth.bsky.social: Indiscriminate brute-force application of AI to mathematics is depleting the field of good questions without renewing this resource or providing commensurate value. OpenAI: We hear you. Here's a .zip file with 377 hasty AI solutions to important open problems.

nytimes.com

OpenAI Releases Findings on 377 Math Problems, Further Roiling Field (Gift Article)

After a previous solution angered mathematicians, the company characterized the new release as being more responsive to concerns about A.I. disrupting research.

19409119

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Heaven forbid this was pointed out independent of any value of that statement bc lemme tell you

020

Reposted by Nicholas Mancuso

Steve Huntsman @stevehuntsman.bsky.social · 07/10/2026

I think they fucked up again Not impressed by the spamming Plenty of answers look interesting but now I fear much of the earth has been salted. Who is gonna want to do the mop up work? What are we gonna learn about constructions and techniques that isn’t shallow?

2214

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Ok there is a LOT of variability in quality re the writeups. Some aren't bad?

010

Reposted by Nicholas Mancuso

Quantian @quantian.bsky.social · 07/10/2026

Are you fucking kidding me with this one

1936239

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

I mean, I don't think there is anything wrong with using it, provided its made clear that results/math/etc were obtained with AI's help. I use it frequently, but in bio we have enough runway due to needing experiments to keep AI at bay for a bit

010

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Yeah the math/cs fields are getting the brunt of this rn and like any paradigm shift (*shudders*) its going to be a while until researchers figure out how to adapt and integrate it [if at all].

110

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Yeah this sucks. I mean a ton of cool results, but its just garbage writeups, no human involvement to curate and improve arguments. Just a data dump.

1100

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Paper is the most AI written monstrosity ever. Just straight into the proof, no motivation, no connection to other areas, no implications for what this means and why its important.

250

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

god the write ups are... well... technically write ups

0270

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Okay. I'm much less excited about -how- these results have been released, but between yesterday's 3SUM/APSP and now UGC? I feel conflicted bc I want to be very excited about these results while not wanting to give free advertising to OpenAI that just said 'let er rip' w/o human involvement

290

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Precisely because here the authors spent time to produce *understanding*. I would be shocked if additional work isn't built upon these results both by AI and researchers alike. It's not my field and I could easily be wrong however.

000

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Yeah I'm not 100% sold on this to be clear, and the marketing aspect is precisely why my enthusiasm was dampened. But as I mentioned downthread, if these companies/tools aren't going away this seems like a better approach compared with previous 'ai solves problem here is incomprehensible paper'

110

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

Theres litany of other issues that arise with using/releasing math from ai models that range from moral/ethical and I wont pretend to be an expert and claim this is how things -should- be done. But if these systems really aren't going away, this seems like a step in the + direction compared to prev

100

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

For all the recent AI-math results lately, I think if companies are going to be involved, and experts are willing, this not such a bad model imo. The authors clarified and expanded up on the core results, and released work that presents that understanding alongside the idea.

200

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

As far as COI goes, I'm not quite sure. Academics are paid consultants all the time, provided that is reported to their Universities. The article was quite upfront about it, so I can't read that as anything resembling misconduct.

120

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

I can't really answer on behalf of the authors and its pointless to speculate tbh. As I mentioned in another thread, my enthusiasm was somewhat diminished (minor!) when I realized this was the result of Anthropic reaching out to the authors, rather than the authors using it themselves.

110

Reposted by Nicholas Mancuso

bioRxiv Genetics @biorxiv-genetic.bsky.social · 06/10/2026

Linkage disequilibrium moderates how evolving genetic background shapes variant-level architecture www.biorxiv.org/content/10.64898/20…

092

Reposted by Nicholas Mancuso

Hanbin Lee @epigenci.bsky.social · 06/10/2026

We studied the evolution of genome-wide linkage disequilibrium and its influence on genetic architecture with genetic drift. We found that the Bulmer effect is population size and demography dependent. For nerds: genetic architecture follows from solving PDEs.

14821

Nicholas Mancuso @nmancuso.bsky.social · 07/10/2026

I'm reading that as Anthropic compensated the authors. Given the statements about how this work is independent of their work at University [maybe for ownership legality reasons?], it reads as if this was consulting work. That is my interpretation, and the authors would be the ones to contact tho

100

Reposted by Nicholas Mancuso

Daniel Swain @weatherwest.bsky.social · 06/10/2026

Just did some number crunching using latest (Oct 1) ECMWF seasonal data, and I must say that I've never seen such high predictive confidence in a very wet California winter. This is a truly extraordinary forecast amid a record-strength El Niño event. [Thread: please read to end.]

Model-relative sea surface temperature anomalies for November 2026–January 2027. The supplied grid is displayed without added smoothing or coastal filling. Beige marks land and gray marks missing ocean values. The main colour scale spans -2 to +10°F, with white at zero. Hot pink marks anomalies above +10°F. This map does not establish an operational ENSO category.

16534217

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

We're lucky enough to live within 3 blocks of our kids elementary school. But the sad reality is that I dont feel comfortable with them walking/biking themselves because they need to cross a 4-lane street where cars regularly speed ~ 50mph with no crossing guards! It's unreal. 😡

021

Reposted by Nicholas Mancuso

Mark Peifer (He, him) @peiferlabunc.bsky.social · 06/10/2026

Please find 5 minutes today to draft your response to this drastic change in NIH Peer review. The deadline is October 13. I pasted my own response below as an example--please share widely 🧪

I am an NIH-funded scientist who has run a lab for 34 years. I have served on literally dozens of NIH peer-review study sections, most recently this past June. I also have served in Council for both NIGMS and for CSR. I thus am very familiar with the current review process from both the prospective of an applicant and a reviewer. I strongly oppose the proposed changes. While NIH peer review is not perfect, it is the best system thus far devised, and is the envy of colleagues in other countries. Three experts in the field spend hours reviewing each proposal, using their knowledge of important issues to rank proposals by signficance, feasibility and impact. Then they present their opinions to a panel of 20 or more expert sin the broader field, and a spirited discussion ensues. Thus each proposal is put through very stringent process, and the resulting ranking has substantial value in times when funding is limited. Having review many times, I know that the proposed binning would provide equal "scores" to proposals that were greet with substanailly different degrees of enthusiasm. Further, grants binned just bowl and just above the 25% cutoff would be subsequently viewed as if the enthusiasm for them quite divergent, certainly not an accurate summary. The current fine-grained system avoids these issues. Removing it will make the task of NIH Program officers more difficult. Binning also allows non-scientific issues to take on disproportionate weight in funding, an issue that will be accentuated if political appointees are making the ultimate decisions. Please retain the current system.

23128

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

dont get mad in my mentions about *a joke that is in regard to the lack of a TOC*. chill

000

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Yup, much to my shock last night! Still blown away by this. bsky.app/profile/nman...

040

Reposted by Nicholas Mancuso

Damien C-C @dccc-phd.bsky.social · 06/10/2026

"You can teach a man what consent looks like but you can’t force him to care about it. It follows, I think, that we need radical solutions here: abolishing all-male institutions where such misogyny proliferates and festers and is fastidiously protected."

0295

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Glad to see the media finally being responsible for once

000

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

To be clear I'm not accusing the authors of any misconduct whatsoever. I saw the abstract and was floored initially, given the history of these problems and hypotheses. Making my way through the introduction I was further blown away by the fact that the spark to this work was due to AI.

010

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Damn! Hadn't gotten that far yet (still making through html version on phone) . I was really hoping this wouldn't be another PR case for them. Still excited by the result however.

020

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Now whether or not the lean proof encodes the problem at hand is another story, is that what you're getting at?

160

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Lean is the verifier here. It won't compile if the proof is wrong. It could be an awkward or uninformative proof, but it shouldn't be wrong (barring bugs in the compiler)

2110

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Im not sure i follow, the proof was written in lean. Are you saying that a separate LLM should write the formal proof? Sorry just trying to make sure I better understand

160

Nicholas Mancuso @nmancuso.bsky.social · 06/10/2026

Oh, not sure. It wrote it's proof for lean, couldn't have been done using agents? What's the argument for diff llm here?

1180