@khinsen.net
Researcher at CNRS (France). Computational science, in particular computational biophysics. Metascience, in particular the evolution of science in the digital era. More active in the Fediverse: https://scholar.social/@khinsen
New blog posts: "Preparing for scientific deepfakes"
blog.khinsen.net/posts/2026/0...
"What I haven't seen yet, but expect to see soon, is the scientific equivalent of deepfakes: made-up results that come with made-up code that reproduces them."
π§ͺ #metascience
@khinsen.scholar.social.ap.brid.gy Talking to myself across the bridge...
A week ago, we had an open conversation amongst senior academics in leadership roles and their use of LLMs (in both their leadership capacities and individually). Small group, not representative, but it already taught me a great deal about the diversity of perspectives (& their consequences). 1/
An interesting experiment in reproducibility: the NanoBubbles project is looking for researchers/labs in nanotechnology who can help with understanding the discrepancy between an widely cited work and its mostly failed replication
I am looking for papers that report on quantum chemistry computations performed on quantum computers.
NOT theory
NOT running on simulators
NOT benchmarks
(all those are easy find)
but computations of real chemical interest on real quantum devices.
No, DeepMind has not solved the protein folding problem.
#Alphafold predictions are valuable hypotheses and accelerate but do not replace experimental structure determination.
RE: https://spore.social/@yoginho/115918117109281530
Recommended!
We've got ISSUES. Literally.
We scraped >100k special issues & over 1 million articles to bring you a PISS-poor paper. We quantify just how many excess papers are published by guest editors abusing special issues to boost their CVs. How bad is it & what can we do?
arxiv.org/abs/2601.07563
A 𧡠1/n
My wish for 2026 for academia specifically, is to push back against fascist framings, reasoning, and industry capture in our universities. My suggestions are below. But please do anything you can if you're an academic. πΏ Even if it's tiny. π Every pebble helps build a damn against fascist forces. πͺ¨
RE: https://mastodon.social/@Viss/115841064419194751
Hello Readers on #Bluesky: there really is no excuse anymore, *not* to migrate to the #Fediverse and escape your corporate overlords.
Empire or rebellion? This is the question.
Which one do you support?
Your actions shout louder than your [β¦]
2020s: AI "System Prompts" are lengthy, carefully constructed sets of expert rules about a particular domain, created by "prompt engineers".
1980s: AI "Expert Systems" were lengthy, carefully constructed sets of expert rules about a particular domain, created by "knowledge engineers".
Many toots, but nothing long-form I can think of. Open Source is ultimately a means that can be applied towards many ends. There's full corporate Open Source (e.g. Chromium and Firefox). There's also corporate takeover of Open Source with academic origins (e.g. Scientific Python).
Rate your score on Factor Fexcectorn.
Well done, Scientific Reports. pubpeer.com/publications...
Proteins are dynamic structures, but structural biology often shows them as static snapshots. Inspired by long-exposure photography and generative art, I built ProteinCHAOS, an artistic tool inspired by molecular dynamics to capture protein flexibility over time, much like long-exposure images.
Call for contributions/demos for programming 2026 in Munich, March. Ooh I'll have some serious shit cooking by then.
2026.programming-conference.org/home/substra...
Very happy to share our collaborative project on FAM118 proteins - noncanonical sirtuins that form filaments and process NAD in human and other vertebrate cells.
However, for the blog, I'll keep the static site as well. It's archivable and usable without JavaScript. Plus the posts get DOIs.
We have a long way to go before we can put together all the desirable features for scientific publishing!
Turning my blog into a HyperDoc is one of my projects!
As for network/server dependency, I am betting on having Common Lisp on WASM in the near future, and then I can run everything in the browser, while retaining the server option for heavier computations or larger datasets.
Thanks for reporting! I had forgotten install the SBCL source package on the server. That should be fixed now.
New blog post: "Explorable explorable explanations"
Building on Bret Victor's concept of "explorable explanations" for scientific publishing.
blog.khinsen.net/posts/2025/1...
π§ͺ #MetaSci
AI hype letter remains open for more signatures tinyurl.com/Sign-Letter-...
Scientists and scholars in AI and its social impacts call on von der Leyen to retract #AIHype statement.
@olivia.science
@abeba.bsky.social
@irisvanrooij.bsky.social
@alexhanna.bsky.social
@rocher.lc
@danmcquillan.bsky.social
@robin.berjon.com
& many others have signed
www.iccl.ie/press-releas...
Played with @khinsen.net's hypedoc today. Wow, really good. @tomasp.net u would like. ALT + CLICK on most elements to reveal the source code, all the source code elements are hyperlinked. So deep for exploration.
hyperdoc.khinsen.net/94FE4-microg...
Revised preprint: "Establishing trust in automated reasoning"
osf.io/preprints/me...
Based on two very helpful reviews from @metaror.bsky.social:
metaror.org/kotahi/artic...
More references, better discussion of unfamiliar concepts (e.g. conviviality), and more.
π§ͺ #metasci #compsci
Had a fun meeting with @ronentk.me and we set a simple landing page for @atproto.science π§ͺ
Take a look at atproto.science
Then please formulate it that way! The current wording suggests that floats may not be handled correctly by some implementations.
Yes, the current float support is useful, e.g. for small observational datasets. But only if support is reliable.
That makes perfect sense, but it also means that DASL is no good for scientific data. So: no need to worry about BigNums.
However, I don't see the point of having some float support but recommending not to use it for the sake of interoperability. It sends a message of confusion.
"most of the IEEE 754 "special" floating points (infinity, negative infinity, minimal NaN, and NaN with payloads) MUST NOT be encoded."
Doesn't sound good for storing data that fully exploits the IEEE formats (which isn't that common, but it happens).
"Completely avoiding floating-point numbers is RECOMMENDED to minimize interoperability and tooling issues."
Doesn't sound promising for long-term storage and interoperability.
