Valentin Rodionov @arbitraryeffect.bsky.social · 27/08/2026I see the context of the post as, "replicability crisis is an overhyped right wing talking point". Which bad actors used to destroy whole fields. I agree that this is what happened (i.e. field destruction). But to claim that all is hunky dory on the science side is incorrect. 000
Valentin Rodionov @arbitraryeffect.bsky.social · 27/08/2026And here are the global publication numbers for all sciences, aggregated from OpenAlex. ~1.7M papers in 2000, 8.4M in 2025. There are not 5x more scientists now compared to 2000. Much of that output is not credible. It is published to exist. Science is in crisis. And bad people will use it. 200
Valentin Rodionov @arbitraryeffect.bsky.social · 27/08/2026Wow - Macchiarini has a Substack! 🤡 Also, these two names are not, in fact, inside the same argument. 010
Valentin Rodionov @arbitraryeffect.bsky.social · 18/07/2026The only thing he is "on to" here is offloading critical reading of cited papers on to an LLM. Checking links is not the same as proper, critical citation. Also, an LLM is not a search engine (even if it is integrated with one). 000
Valentin Rodionov @arbitraryeffect.bsky.social · 18/07/2026The problem is the phrase "checked the links." Checking a link tells you a URL resolves. It does not tell you the paper behind it says what you claim. He trusted the model on the content, did not read the sources himself, and announced it in public. 142
Valentin Rodionov @arbitraryeffect.bsky.social · 18/07/2026Screenshotted for the record, in case it vanishes. The Editor-in-Chief of Science, explaining how the citations in his latest piece were checked. Short thread. 🧵 1104
Valentin Rodionov @arbitraryeffect.bsky.social · 05/07/2026Sidney is wishing you a happy caturday! 081
Valentin Rodionov @arbitraryeffect.bsky.social · 26/05/2026Case in point: γ-graphyne, an acetylenic 2D carbon allotrope. Xiaoli Cui's group at Fudan claims to make it by ball-milling hexabromobenzene (or just benzene!) with calcium carbide. Their 2018 Carbon paper alone has 319 citations. The claimed chemistry is ridiculous and impossible. 110
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026Here is what the Russians actually published in 2013 (doi: 10.30906/0023-1134-2013-47-8-41-44). DBU is specifically used in the VI --> VII ring closing step. The advantage of using DBU there is that VII is clean enough to crystallize. No annoying column chromatography! 110
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026@theins.press article specifically quotes their source as claiming DBU being the "most important ... reliable marker" for the synthesis of epibatidine. That is simply incorrect, and the quote from their source (attached here) is something of a word salad. I doubt they were a real chemist. 140
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026DBU is easily accessible (standard industrial synthesis involves caprolactam and acrylonitrile). It is one of the most standard bases used in organic synthesis, and is also used as a curing agent for some epoxy resins. It is very inexpensive and universally available. Russia likely makes its own. 120
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026This brings me back to @theins.press reporting. Quoting the article: "...the key component for creating the bicyclic compound: 1,8-diazabicyclo[5.4.0]undec-7-ene..." The name sounds complicated, so that stuff must be rare and unique. Except it is not. It is just DBU - one of the commonest reagents. 130
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026Now, 6-chloropyridine-3-carboxaldehyde is not particularly hard to make, and not especially expensive. Current Thermo Fisher price is $127 per gram. Quirky building blocks like this would rarely if ever show up in importer databases. It is a small-volume, obscure research chemical. 110
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026The second thing that is apparent is that all the compounds on the list, with one critical exception, are generic, and not related specifically to the structure or synthesis of epibatidine. The one exception? 6-chloropyridine-3-carbaldehyde. If you think it looks like a part of epibatidine - yes! 110
Valentin Rodionov @arbitraryeffect.bsky.social · 31/03/2026@theins.press article lists compounds involved in the published synthesis of epibatidine, and the number of RU importers. So far, so good... or not. The first thing to notice is that 3-buten-2-one is methyl vinyl ketone (MVK), which Russia certainly makes on multi-ton scale, and likely imports too. 110
Valentin Rodionov @arbitraryeffect.bsky.social · 28/03/2026This wonderful letter is also very likely generated by an LLM. Even without running it through the detector, the slop is easy to see. So classy. I feel bad for those on the receiving end. 011
Valentin Rodionov @arbitraryeffect.bsky.social · 04/03/2026This you? For the interested readers: first image failing vertical line test is a crop from doi.org/10.1016/j.im... Second is from a PubPeer post related to doi.org/10.1038/s414... The 2024 "NanoImpact" paper already cited ~24 times, Nat Comm ~255 times. Good times. 111
Valentin Rodionov @arbitraryeffect.bsky.social · 26/02/2026So many things wrong with this post. 1. In the 90s RU, Communists were the *conservatives*. There were ~14 parties in the '93 Congress, with a slight conservative majority 2. The crisis was not "oligarchs vs people". It was infighting in the power structure, with Yeltsin arguably the "better" guy. 110
Valentin Rodionov @arbitraryeffect.bsky.social · 06/02/20261. You used "Democrat" precisely because it is relevant. 2. We have a 2-party system. The right-wing party devolved into Nazis. Your (in)actions helped Nazis. The dismantling of USAID and ongoing genocides in Gaza and Ukraine are on you. 3. Yes, Greens support genocide (see attached). Any comments? 1600
Valentin Rodionov @arbitraryeffect.bsky.social · 30/01/2026It's a shame he is a Putin supporter. ocdn.eu/sport-images... 120
Valentin Rodionov @arbitraryeffect.bsky.social · 15/12/2025Around 2016–17, creeping Apple hardware enshittification pushed me to Windows. By then Windows had WSL2 providing a proper bash shell, and, crucially, Photoshop, Illustrator, ChemDraw, and MS Office, all well integrated. 130
Valentin Rodionov @arbitraryeffect.bsky.social · 15/12/2025Perspective matters: I’m far from a Linux/Unix noob. I started on IRIX 6.5 on an SGI Octane, an Red Hat 7.x on my trusty PII-450 around 2000. SGI's Indigo Magic was polished and Just Worked™; early Linux… less so. 🧓 120
Valentin Rodionov @arbitraryeffect.bsky.social · 02/12/2025You’re in a desert walking along when all of the sudden you look down, and you see a tortoise. You reach down, you flip the tortoise over on its back. The tortoise lays on its back, beating its legs trying to turn itself over, but it can’t, not without your help. But you’re not helping. Why is that? 120
Valentin Rodionov @arbitraryeffect.bsky.social · 08/10/2025Here are a few more doi's from the same journal, some already flagged on PubPeer, some not, all 💯 fabricated with hand-drawn or nonsensical spectra and PXRDs, and citation padding. 10.1002/app.57648 10.1002/app.56053 10.1002/app.56837 (the "IR spectrum" is Fig. S2, 10.1002/app.57648). 130
Valentin Rodionov @arbitraryeffect.bsky.social · 08/10/2025Or how do you make peace with the main scheme of the paper? While the "initiator" is KMnO4/oxalic acid, PAN chains magically grow directly from chitosan. It must be magic, because this cannot be accomplished with chemistry. This passed 🍐review, and has been on display in this journal for 15 years. 120
Valentin Rodionov @arbitraryeffect.bsky.social · 08/10/2025"We have asked the authors to provide a detailed explanation and supporting data to address the concerns. The authors were able to do so to our satisfaction, hence we consider the matter resolved. " How do you "resolve" pixel-identical stretches of background in two NMR spectra? 120
Valentin Rodionov @arbitraryeffect.bsky.social · 08/10/2025Re-posting Dr. Wittmers' thread, with a some supporting info. Fabian flagged a paper in J. Appl. Polym. Sci for region duplication in PXRD (see pubpeer.com/publications... ), and reported it to Wiley. I have found additional duplications and crazy chemistry. Wiley's response? All 💯 👍 #ChemSky 161
Valentin Rodionov @arbitraryeffect.bsky.social · 24/09/2025De-extincted species be like (image attribution: halloweencostumes.com) 120
Valentin Rodionov @arbitraryeffect.bsky.social · 22/06/2025Sydney the woodworker cat. Happy #Caturday ! In case you were wondering: the plane is a Type 4 No. 5 Stanley (1874-1884). Probably the oldest I have ever restored. 040
Valentin Rodionov @arbitraryeffect.bsky.social · 11/06/2025And here are some cool pictures! The periodicity of the "bumps" on the lateral force image corresponds to the crystallographic planes bisecting aromatic rings. And the triple bonds are very visible in Raman, even in films ~20 atoms thick (or thin?). 001
Valentin Rodionov @arbitraryeffect.bsky.social · 11/06/2025We synthesized few-layer films of γ-graphyne (the ≡ bond carbon allotrope) at a liquid/liquid interface. It's a much cleaner synthesis than our original batch method - and the stacking of the layers is orderly! Coming soon to a 🍐-reviewed journal. Preprint 🠟🠟🠟 #ChemSky doi.org/10.26434/che... 131
Valentin Rodionov @arbitraryeffect.bsky.social · 04/05/2025It's pretty clear the model plagiarized this old Soviet anti-alcohol poster. It's kind of appropriately hilarious given who is depicted. 010
Valentin Rodionov @arbitraryeffect.bsky.social · 02/05/2025So I just opened my fridge, and discovered a very interesting product: poppi prebiotic soda! Pre/probiotic mix-up is one of my pet peeves. This one gives some definite Engineer vibes. The soda actually tastes pretty good. #ChemSky 🧪 1101
Valentin Rodionov @arbitraryeffect.bsky.social · 18/04/2025Found it! It is hiding in the "Save as Image". What threw me off was the .pdf option in "Save As". The resulting .SVG plays better with InkScape than any of my previous attempts. Progress! 010
Valentin Rodionov @arbitraryeffect.bsky.social · 18/04/2025Marvin cannot export SVG (or the option is cleverly hidden?). Also, structures copied as .PDF and saved as .PDF open differently by InkScape. Copying completely obliterates any formatting. 100
Valentin Rodionov @arbitraryeffect.bsky.social · 18/04/2025I am sure it is possible to get things done with IS, sort of. So far I am extremely unimpressed. It does not import well from most chemical drawing software. The UI is a disaster. The inspectors do not follow selection type, there is no "select same" for fonts, etc. But I can rotate artboards! 000
Valentin Rodionov @arbitraryeffect.bsky.social · 06/04/2025Also daffodils and hyacinths (though these are not as bad as lilies). In daffodils, the offending alkaloid is lycorine. www.aspca.org/pet-care/ani... www.aspca.org/pet-care/ani... 000
Valentin Rodionov @arbitraryeffect.bsky.social · 02/04/2025Here is the syllabus (on a side note, I was surprised you cannot post a .PDF to BlueSky). 110
Valentin Rodionov @arbitraryeffect.bsky.social · 02/04/2025Here is DeepSeek "thinking". It is very sure of itself. By the way, the results are generally better than ChatGPT or Gemini - but it's still not anywhere close to genuine reasoning. 010
Valentin Rodionov @arbitraryeffect.bsky.social · 23/03/2025On-the-spot guidance. en.m.wikipedia.org/wiki/On-the-... 070
Valentin Rodionov @arbitraryeffect.bsky.social · 17/03/2025And all to power the bullshit machines. Even if these constructs could be made intelligent, burning down fossil fuels to power them would be unconscionable. Since these are just LLMs, this is also incredibly stupid. thebullshitmachines.com 010
Valentin Rodionov @arbitraryeffect.bsky.social · 19/02/2025Jut looked it up. Yes, it is a Gemini LLM with bolted-on bits for web search and AlphaGo-like self-play (also done with LLMs). Taken in isolation, their block diagram could be used as an example of how *not* to do science. And LLMs fundamentally cannot reason. I just wish the LLM bubble pops soon. 210
Valentin Rodionov @arbitraryeffect.bsky.social · 18/02/2025I want to tap into the collective intelligence of ChemSky and learn a bit about EXAFS. I just came across the EXAFS spectra and fitting parameters attached here. Do these look broadly reasonable? If not, why? The source paper will for now remain nameless, but it has fake Raman data. #ChemSky 230
Valentin Rodionov @arbitraryeffect.bsky.social · 13/02/2025This is probably the funniest/fakest PXRD pattern I have ever seen (and I go through a lot of papermill products). Found in US patent application US 2023/0088084 A1. Brought to you by the same fine people who "discovered" "holey graphyne". www.chemistryworld.com/opinion/hole... #ChemSky 2244
Valentin Rodionov @arbitraryeffect.bsky.social · 27/01/2025Deepseek is better than ChatGPT 4o on my "brick/feathers" test. It gave a correct answer 4 out of 5 times. ChatGPT answered wrong 4 out of 5 times, and Gemini failed hilariously and spectacularly 100% of the time. 030
Valentin Rodionov @arbitraryeffect.bsky.social · 27/01/2025The current incarnation of imitative "AI" is as useful as a calculator that is wrong 35% of the time. I just asked Gemini and ChatGPT what is heavier, 1 kg of feathers or 2 lb of bricks. Both models said the bricks are heavier 😂 000
Valentin Rodionov @arbitraryeffect.bsky.social · 23/01/2025It's not the only "dark matter alkyne" paper from the same source. More are already on PubPeer, with the authors trying to defend the indefensible. And yes, the attached "reaction scheme" was published in an Elsevier journal in 2022. It's still there. pubpeer.com/publications... 120