iiinova — where your digital self keeps evolving

Reposted by Ragnar {Groot Koerkamp}

What I like about not using LLM-based tools is writing code is so tedious you are looking for solutions to write less code or no code at all and you usually find them. Which improves your project dramatically in a longer run.

616811

Reposted by Ragnar {Groot Koerkamp}

Robin Hawkes @robhawkes.bsky.social · 11h

What if offshore wind farms were star constellations? 🌌 I've had this idea in my head for years, ever since I first saw satellite radar imagery of wind farms. They often look surprisingly like a telescope image of a star field. Over the weekend I finally made it a reality…

1124

Ragnar {Groot Koerkamp} @curiouscoding.nl · 29/09/2026

Randomly browsing crates.io, when suddenly... Can LLM-generated testimonials be part of academic CVs already?

This module is a derivative work; the algorithm — the cache-line-sized
implicit S(+)-tree, the branchless SIMD node compare, and the software-pipelined/prefetched batch traversal that hides memory latency — is entirely Ragnar's. **Thank you, Ragnar Groot Koerkamp**, for the design and the beautifully written explanation; every fast path here traces back to your work, including the B-factor and batching trade-offs we re-measured for i64.

1130

Reposted by Ragnar {Groot Koerkamp}

Rasmus Pagh @rasmuspagh.net · 29/09/2026

Very interesting report about AI and theoretical computer science, based on a recent meeting at @simonsinstitute.bsky.social simons.berkeley.edu/ai-tcs-worki... These are times of change, but I am optimistic that there will always be a valuable place for humans who seek to understand.

simons.berkeley.edu

AI + TCS Working Group

0123

Reposted by Ragnar {Groot Koerkamp}

Grace @gracekind.net · 30/11/2025

I don’t think this is a narcissistic injury, but it is an egoistic injury. And it hurts a lot! If a computer can do everything I can do, what makes me valuable? This is a difficult question to answer

4035016

Ragnar {Groot Koerkamp} @curiouscoding.nl · 28/09/2026

Google scholar sent me a notification of a new preprint with a highly relevant but obivously LLM-generated title. On 10s of scanning, many headers in the text are also LLM-generated. It's just so repulsive I cannot put myself to looking at the thing more than 20 seconds total...

110

Reposted by Ragnar {Groot Koerkamp}

Resist 🕎🍉 @antizionistjew.bsky.social · 28/09/2026

Google Maps Updated Their Maps - The Atrocities Of Gaza Are Now EXPOSED To The World. Heartbreaking 💔

13527

Ragnar {Groot Koerkamp} @curiouscoding.nl · 27/09/2026

We have AI, but gmail still doesn't have a feature where it suggests to snooze a mail to the one date that's mentioned in it.

020

Reposted by Ragnar {Groot Koerkamp}

Rob Patro @robp.bsky.social · 27/09/2026

Tangentially, @jermp.bsky.social also points out @recombconf.bsky.social is listed as B tier in CORE portal.core.edu.au/conf-ranks/?.... This is something that we should figure out how to address. RECOMB is a premiere conference in comp bio, & it is highly selective. It is `A` tier without question.

portal.core.edu.au

043

Reposted by Ragnar {Groot Koerkamp}

Rob Patro @robp.bsky.social · 25/09/2026

Ok comp. bio / bioinformatics / genomics friends. Let's talk about @recombconf.bsky.social and, specifically, the way in which the "proceedings" have evolved. This seems like something that we, as a community, should address. RECOMB now has no "published" proceedings in the classic sense.

195

Reposted by Ragnar {Groot Koerkamp}

Greg Egan @gregegansf.bsky.social · 27/09/2026

I was aware of Google announcing an implementation of Shor’s algorithm that needed substantially fewer qubits than previous methods, but I missed the bit where they released a Zero Knowledge Proof of their claim ... www.scientificamerican.com/article/why-...

scientificamerican.com

Why mathematicians are suddenly hiding research results

Google made a major breakthrough in quantum computing. To prevent the results from being misused, the company kept them secret. That backfired

1194

Reposted by Ragnar {Groot Koerkamp}

jazz mickle @queenjazz.gay · 27/09/2026

conventional wisdom for the entire time i've been programming has always been "less code is better". the idea that we must write more code faster is ridiculous. the problem has never been rate of output

435962

Reposted by Ragnar {Groot Koerkamp}

amos @fasterthanli.me · 26/09/2026

Q: what is it called when you start letting AI write some of your docs, then some more, then... A: "slippery slop"

1653

Ragnar {Groot Koerkamp} @curiouscoding.nl · 26/09/2026

Woke up at 2500m this morning 🤩

Picture of a rugged mountain ridge with a perfectly clear blue lake in front. The ridge is reflected in the lake. The foreground shows rocks at the bottom of the (shallow) lake, while the shoreline on the back shows a red upside down tent drying in the sun.

0140

Reposted by Ragnar {Groot Koerkamp}

Gro-Tsen @gro-tsen.bsky.social · 26/09/2026

The two answers to the MO question mathoverflow.net/q/515514/17064 exemplify these two mindsets. The main damage done by AI here is that the second motivation means you need to keep your questions absolutely secret and refuse to answer “what are you currently working on?”.

mathoverflow.net

Soft question: an era in my research terminated by AI (which is, in a way, not so bad)

I feel like writing a post about it, though I am not sure what kind of answers I will get. Here is my story. Ever since I have heard Michael Atiyah talk about his problem on configurations of points (

191

Reposted by Ragnar {Groot Koerkamp}

Bede Constantinides @bede.im · 24/09/2026

"Claude Opus 5.5 refused to answer on biosecurity grounds, its refusal costing half that of DeepSeek’s complete answer"

061

Reposted by Ragnar {Groot Koerkamp}

Antoine Leblanc @nicuveo.gay · 24/09/2026

AI is the triumph of "code is nothing but a means to an end", which is why we in the "software engineering is a craft" camp tend to hate it so much.

0278

Reposted by Ragnar {Groot Koerkamp}

Matt Henderson @matthen.com · 24/09/2026

every (convex) quadrilateral is a square- if you view it from the right perspective

1231432

Reposted by Ragnar {Groot Koerkamp}

Ryan Moulton @moultano.bsky.social · 23/09/2026

Over and over I find that the key to being a good experimentalist is to trust absolutely nothing.

51338

Ragnar {Groot Koerkamp} @curiouscoding.nl · 24/09/2026

I'm honestly surprised that at this point people are still surprised Claude can autonomously optimize code. It's just measure, identify hotspots, optimize, measure again, and repeat. Honestly all quite mechanical steps. The optimizations themselves are also mostly pattern matching I suspect.

130

Reposted by Ragnar {Groot Koerkamp}

Jarno N. Alanko @jnalanko.bsky.social · 23/09/2026

www.anthropic.com/news/claude-... 950 Claude agents spent 21 hours looking at sequences encoding reverse transcriptases and noticed nearby spaced repeats like in CRISPR. It's cool but.... how come the string matching community had not found this before? Embarrassing.

142

Reposted by Ragnar {Groot Koerkamp}

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 23/09/2026

If you view AI as filling in the convex hull of knowledge, this is great in the short term but bad in the long term if there's no one left to expand the hull after

1015715

Reposted by Ragnar {Groot Koerkamp}

Dirk Lorenz @dirque.bsky.social · 23/09/2026

Signs that academia may be for you: * eduroam works on the first try * you get your travel reimbursement without having to email anyone twice * you understand your institution’s course management system

3111

Reposted by Ragnar {Groot Koerkamp}

Jarno N. Alanko @jnalanko.bsky.social · 23/09/2026

I'm starting to believe in the labor theory of value for TCS/math papers. That is, given equally skilled workers, with equal tools, and equally important problems, the value of a paper is roughly proportional to the amount of (human) labor that has gone into it.

121

Reposted by Ragnar {Groot Koerkamp}

Camille Fournier @skamille.themanagerswrath.com · 22/09/2026

New critical skill: refusing to let people give you work by throwing AI crap at you. If it isn't immediately obvious what the value is from the doc it's going right back to sender

3315

Reposted by Ragnar {Groot Koerkamp}

Phil Eaton @eatonphil.bsky.social · 22/09/2026

Man there's no way to crash your reputation so quickly as an engineering org as publishing obviously AI-written articles.

0305

Reposted by Ragnar {Groot Koerkamp}

Yuri Sulyma @yurisulyma.bsky.social · 21/09/2026

A lot of AI/math discourse has a kind of sitcom-y view of mathematics, as if every open problem is a profound breakthrough Many open problems are there bc the author was exhausted and wanted to get the dang paper on arxiv already; or to leave low-hanging fruit for young people

0121

Reposted by Ragnar {Groot Koerkamp}

Zach Weinersmith @zachweinersmith.bsky.social · 22/09/2026

Artists: but I don't want to lose my job. I love it. Mathematicians: but I don't want to lose my job. I love it. Economists: LONG HAVE I AWAITED THIS HOUR! DESTROY ME, EFFICIENCY, AND LET UTILIZATION-ADJUSTED TOTAL FACTOR PRODUCTIVITY RISE ON THE WIND OF MY ASHES!

1869881

Ragnar {Groot Koerkamp} @curiouscoding.nl · 22/09/2026

I'll also present the SimdQuickHeap at P99, in case you missed it at the ESA best paper session ;) Basically, it's a pretty simple data structure inspired by quickselect with some SIMD sprinkled on top that's up to 2x faster than all other priority queues and up to 8x faster than binary heaps!

171

Reposted by Ragnar {Groot Koerkamp}

Jarno N. Alanko @jnalanko.bsky.social · 22/09/2026

It is really starting to feel like at least a decade worth of pure math/TCS results is going to be happening in Q4 2026. Human understanding, digestation and dissemination is going to take a while. But wow.

031

Reposted by Ragnar {Groot Koerkamp}

Zalea 💜 @zalea.bsky.social · 22/09/2026

this comparison to alchemy is super interesting. LLMs feel similar to mercury in that they could be a red herring and exposing yourself to them too much messes up your brain

The alchemists also had clues that led them astray. For one thing, the were obsessed with mercury. Mercury is not very interesting chemically, but it is the only metal that is a liquid at room temperature.

This seemed very significant to the alchemists, and caused them to place mercury at the heart of their alchemical system, and their search for the Philosopher's Stone, a way to turn base metals into gold.

It didn't help that mercury was neurotoxic, so if you spent too much time playing with it, you started to think weird thoughts. In that way, it resembles our current thought experiments with superintelligence.

37215

Reposted by Ragnar {Groot Koerkamp}

Cooper Lund @cooperlund.online · 22/09/2026

Damn, he had Sam Altman dead to rights in 2016

3744169

Reposted by Ragnar {Groot Koerkamp}

Alexander Kasprzyk @amkasprzyk.bsky.social · 22/09/2026

Public perception about the future of mathematics is an enormous challenge facing our community. We are currently not even remotely in control of the narrative. For example, the photo shows what’s on display in every newsagents and supermarket in the UK. Will future students be scared away? #MathSky

The latest issue of New Scientist magazine, with the front cover “How AI blew up maths”.

5288

Ragnar {Groot Koerkamp} @curiouscoding.nl · 22/09/2026

At least so far they're only proving open problems, i.e. known-unknowns. Here's to hoping there'll still be work for us in defining what we don't know... (not sure I'm optimistic.) Also, given that they `grep conjecture` I wonder if they deemed my minimizer conjectures worthy. 😆

110

Reposted by Ragnar {Groot Koerkamp}

Huck Bennett @huckbennett.bsky.social · 22/09/2026

Wow does the post title bury the lede! I particularly like the pictured comment.

35715

Reposted by Ragnar {Groot Koerkamp}

Jarno N. Alanko @jnalanko.bsky.social · 21/09/2026

Exciting news regarding the SBWT rust API! We now have an implementation of the subset rank method in this preprint, written Anastasia Diseth. It takes the space down from 5 bits/kmer to 2.5 bits/kmer (double that if both strands are indexed). About 4-5 times slower. www.biorxiv.org/content/10.6...

biorxiv.org

New Space-Time Tradeoffs for Subset Rank and k-mer Lookup

Given a sequence S of subsets of symbols drawn from an alphabet of size σ , a subset rank query srank( i, c ) asks for the number of subsets before the i th subset that contain the symbol c . It was r...

182

Ragnar {Groot Koerkamp} @curiouscoding.nl · 21/09/2026

The math of MSCA applications: - Suppose writing an application takes 2 months. - Total cost of hiring a PhD is maybe €5k/month (??) - Applying costs €10k. - Total cost for 21k proposals: €210M - Total funding: €400M So half the budget is wasted on just applications.

472

Ragnar {Groot Koerkamp} @curiouscoding.nl · 20/09/2026

This stuff is all wrong. Turns out that a B200 GPU costs ~40k and draws 1kW. At 5ct/kW prices, that means you have to run it for 91 years to pay as much in energy as for the hardware... Assuming they depreciate in 5 years, only 5% of their cost is the energy; or maybe 10% with cooling overhead.

120

Reposted by Ragnar {Groot Koerkamp}

Ted Underwood @tedunderwood.com · 19/09/2026

Embarrassed by the baseless speculation I was spreading nine days ago. Should have waited for the base.

engadget.com

Anthropic has set up a bio research lab for physical experiments - Engadget

Anthropic now has its own bio research lab in San Francisco.

1526528

Reposted by Ragnar {Groot Koerkamp}

Daniel Lemire @lemire.bsky.social · 19/09/2026

How did Apple Silicon get 50% faster in three years? lemire.me/blog/2026/09/19/how-did-a…

052

Reposted by Ragnar {Groot Koerkamp}

Lea Verou, PhD @lea.verou.me · 19/09/2026

“These five paragraphs of Claude’s PR description could have been a sentence” is a the new “this meeting could have been an email”

511515

Reposted by Ragnar {Groot Koerkamp}

conputer dipshit @davidcrespo.bsky.social · 19/09/2026

what Sanderson is calling "motivated explanation" means roughly what he does with his videos, namely making the case for why we should care about this question just as much as clarifying the question

In a proof, definitions sit at the start. It is common and expected to begin with a new construction and proceed by analyzing its properties.

In a motivated explanation, definitions sit in the middle. New constructions are only allowed to enter the vocabulary if the problem they are addressing has been clearly established.

In a proof, all statements must be correct, each claim following as a necessary implication from what comes before.

In a motivated explanation, it is okay and often desirable to start with an idea that is not quite right and requires correction, but whose origins are relatable.

A genre of motivated explanation I’m fond of is “discovery fiction”, a term coined by Michael Nielsen. You develop an idea with a narrative that starts with a simple-but-wrong solution to a problem, see where it breaks down, fix that problem, discover a new problem, and so on.

The scope of a proof is to explain why a particular theorem is true.

The scope of a motivated explanation is not only to clarify why a theorem is true, but why the theorem is the right one to pose in the first place, and how it is used in the surrounding context.

1383

Reposted by Ragnar {Groot Koerkamp}

Jarno N. Alanko @jnalanko.bsky.social · 19/09/2026

This race to be the first to publish an AI-generated result is getting silly. If the answer is just a prompt away for anyone, why should *any* human get credit for it? The value is in the quality of the exposition. Racing to be the first to publish is detrimental to that.

051

Reposted by Ragnar {Groot Koerkamp}

Mara Bos @mara.bsky.social · 19/09/2026

I love this. More companies need this type of energy.

Screenshot of a Prusa assembly manual webpage. The breadcrumb navigation at the top reads: “Support > Prusa CORE One+ > Prusa CORE One+ (Gen 2) kit assembly (1.00) > 1. Introduction | Reward yourself”. The page heading is “12. Reward yourself”.

The visible instructions read:

“Building the CORE One+ is a rewarding experience. After finishing each chapter, reward yourself. There is a bag of Haribo bears in the box just for that!

The biggest issue from our experience (MK4S, MK4, MK3S+, MK3S, MK3, MK2S, ...) is inadequate bear consumption. Many of you didn't have enough gummy bears for all chapters, some even ate them all before they started!

After years of thorough scientific research, we came to a solution => At the end of each chapter, you will be told a specific amount of bears to consume.

Eating an incorrect amount than prescribed in the manual might lead to a sudden boost of energy. Please consult a professional in the closest candy store.

Hide the Haribo for now! From our experience, an unattended bag with sweets will suddenly disappear. Confirmed by multiple cases all around the World.”

On the right is a large photograph of a gold “Haribo Goldbären” Pocket Size bag containing multicolored gummy bears.

1594

Reposted by Ragnar {Groot Koerkamp}

Dominik 🔮 @tkdodo.eu · 19/09/2026

How it started ... how it’s going I just can’t do this anymore 😔

Opened 30 other pull requests in 4 repositories

4 open in TanStack/Query

fix(solid-query): follow queryClient changes in useQueries 🤖🤖🤖
fix(solid-query): follow queryClient changes in useMutationState 🤖🤖🤖
fix(solid-query): follow queryClient changes in useMutation 🤖🤖🤖fix(vue-query): accept optional infinite query initial data 🤖🤖🤖

6615

Reposted by Ragnar {Groot Koerkamp}

Lana Sinapayen @sinalana.eurosky.social · 19/09/2026

Can someone who does UX professionally explain to me why you went from "Enter ID and password, click button" to Enter ID click button Enter pw click button I know that login became hidden under huge CREATE ACCOUNT because greed, but why this?

4101

Reposted by Ragnar {Groot Koerkamp}

Grant Sanderson @3blue1brown.com · 18/09/2026

The last IMO problem AI could not solve. For when you want to sit down for a longer problem-solving saga, here's the full video: youtu.be/Nbwv5wHQoj0

514518

Reposted by Ragnar {Groot Koerkamp}

Tommy Löfstedt @tommylofstedt.bsky.social · 19/09/2026

So, things appear to have to become much worse than killing 123 school girls (!) before people understand that LLMs are toy computer programs and not to be relied on for anything of any importance.

163

Reposted by Ragnar {Groot Koerkamp}

Clément Canonne @ccanonne.github.io · 17/09/2026

The Call for Papers for #STOC2027 is up! Importantly, the PC "will place substantial weight on the quality of exposition, and clarity of technical arguments and proofs" Also: - Public posting requirement - Required video submission and more. Deadline: ⏰ Nov 2, AoE acm-stoc.org/stoc2027/sto...

Policy experiments for STOC 2027: In light of rapid advances in generative AI and their impact on research and scientific communication, STOC 2027 is experimenting with several new policies intended to encourage high-quality submissions and promote clear and effective communication of research. The policies below include mandatory public posting and mandatory video submission. Detailed instructions for these two requirements will be released closer to the paper submission deadline.

44014