Stavros Zenios

@lugaricano @joshgans 4.5 You do the revision.

@joshgans I have found this does not work. The referee pushes it in multiple different directions, some of them expansive, some narrowing, some sensible, some crazy. Then the revision tries to achieve all of those, and you end up with a complete mess.

Trying out a new adversarial LLM workflow to improve papers. Ask 1. Claude 4.7 to produce a grumpy report on a paper 2. GPT 5.4 Pro to evaluate the report, but do not be timid and push back. 3. GPT 5.4 Pro for revision suggestions 4. Send revision back to Claude 4.7 Repeat

Read your books faster than ever, from turning the pages, reading, thinking, and metabolizing knowledge, all in one platform.

Turn your ideas into published books faster than ever. 🎨 Design, research, and upload – all in one platform. From coloring books to storybooks, Book Bolt helps you publish faster and smarter.

@ahall_research @jasoncrawford @elmcaleavy @MikeRiggs @ben_golub @YanagizawaD @JohnHCochrane @xuyiqing @Yascha_Mounk Any links on your work using ai for geopolitical risk??

@jasoncrawford @elmcaleavy @MikeRiggs And featuring the research of @ben_golub , @YanagizawaD , @JohnHCochrane , @xuyiqing , @Yascha_Mounk and others

AI is already 10x-ing academic research in the social sciences. In a guest post for @rootsofprogress, I explore how we can get to 100x. Some of my ideas: build more prototypes, define open problems with objective benchmarks to compete on, and keep pressing on dynamic, replicable, agentic research. Check out the post here: newsletter.rootsofprogress.org/p/ai-is-alread…

@Bellmanequation What’s the model?

New setup

@packlesshepherd @CoarseDotInk Turning over to my own subscription does not violate copyright of submitted paper? Isn’t it uploaded?

Big update on @CoarseDotInk: you can now run reviews at ~zero marginal cost through your AI subscriptions! Just drop your paper into coarse.ink, click "Review with my subscription" and follow the instructions See coarse.ink/setup --> "Use my subscription"

The battle of the AI Referees…. something good comes out of this: high quality, low-price refereeing for the benefit of science.

Terribly pleased that my work influenced the integrative taxonomy of mosquitoes. You think this can be invoked next time I go camping?

«Me getting back at you for Claude!”

@Ryderfangfang @KaihatsuYT @Aelthemplaer No we don’t. Trump and Netanyahu did.

@KaihatsuYT @Aelthemplaer You fucking idiots want the war?? You ppl are spiteful nuts.

New favourite clip from Japanese news just dropped

I don’t really understand the maths it takes to send humans behind the Moon and bring them back safely. And the more I sit with that, the more it genuinely messes with my head even tho my love for physics and my knowledge of physics is astounding to a point Somebody had to work out a path where the Moon’s gravity is pulling you in, the Earth is pulling you back, and you’re moving just fast enough and not slow enough not to get trapped by either. They had to figure out the exact angle to come back into Earth’s atmosphere too. Too steep, you burn up. Too shallow, you bounce off and drift into space. And they had to get all of that right at the same time, for real people sitting in a small metal capsule about 400k kilometres away from home. Nothing in that system is standing still. The Moon is moving. The Earth is moving. Even the Sun is pulling on everything. And still, some people looked at all of that motion, all of that chaos, and turned it into numbers you can follow. Go here. Adjust here. Come back here. And unlike nepa light, it infact works. There’s also that moment in the journey where the crew passes behind the Moon. No contact with Earth. No signal. Just silence, with a massive rock blocking everything they’ve ever known. The only reason they can stay calm in that moment is because someone, somewhere, did the maths and proved they’ll come out the other side. I don’t know what it feels like to trust something that much. To put your life in an equation when you’re that far away from everything. But I do know this for sure, whatever that level of thinking is, whatever it takes to reach it, it might be one of the most extraordinary things human beings have ever done...

@ahall_research @karpathy So the model discovered martingales?

We're building a version of @karpathy 's autoresearch to help us predict prediction market prices and, ultimately, forecast the world more accurately. Good news: it has improved its accuracy over time! Bad news: it's done that by giving up, removing predictive features, and just using the previous market price as the predictor. We've got some more work to do! This is only week 2, so we've got plenty of room to run.

@LoizosAntoniou @DemPapadakis @makdrou Το ΑΛΜΑ κ ο Παπαδάκης, αν αποδεχθεί την απόφαση, βγαίνουν δικαιωμένοι.

@StavrosZenios @DemPapadakis @makdrou Το Volt σαφέστατα και είναι εκτεθειμένο ανεπανόρθωτα. Την ίδια όμως ώρα θα πρέπει να αναγνωρίζεται η θέση του θύματος και του θύτη.ΚΙ εδώ ο Παπαδάκης είναι το θύμα. Το ΆΛμα (του ΆΛματος) τον σταύρωσε γιατί χαλά το αφήγημα του χάους.

Νέα, και σωστά, ήθη στην πολιτική σκηνή. Ο @DemPapadakis καλό να αποσυρόταν ο ίδιος: θα τιμούσε τα ήθη του κόμματος του και θα βγεί πολύ ενδυναμωμένος (ότ)αν αθςωθεί. Θα τ μείνει το ήθος και τ @makdrou η τάτσα. (Αν δεν αθωωθεί, το αντίθετο βέβαια). cyprustimes.com/politiki/ektos…

@LoizosAntoniou @DemPapadakis @makdrou Άρα δεν είναι ανθρωποφαγία αλλά έπρεπε το ίδιο ήθος να εφαρμοστεί κ σ Δρουσώτη? Mε βρίσκεις απόλυτα σύμφωνο. Εγώ έκρινα την απόφαση του Άλμα και τι νομίζω πρέπει να κάνει ο Παπαδάκης. Αν ο Δρουσώτης αντιμετωπίζει σοβαρές κατηγορίες και Volt τις αγνοεί, τότε έχει πρόβλημα το Volt.

@StavrosZenios @DemPapadakis @makdrou Ο Παπαδάκης αναφέρεται σε κάποια μηνύματα χωρίς καμία κατηγορία. Ο Μακάριος κατηγορείται για βαριά ποινικά. Ποιος λες ότι έπρεπε να διαγραφεί από υποψήφιος πρώτος;

@NickSpisak_ Why don't just get Karpathy's gist and load to Claude and let it guide me? Not that I did not enjoy reading what you have here, but what is the real valuew added of following your steps then letting Claude do it for me, under supervision? Os it that some readers want hand-holding

@LoizosAntoniou @DemPapadakis @makdrou Είστε εκτός εποχής: "It is clear that my remaining in office has now become a distraction from this government's important work. For that reason, I offer my resignation." Μπορεί ο @makdrou να είναι καννίβαλος αλλά @DemPapadakis κ ΑΛΜΑ να έχουν το ήθος? reuters.com/world/uk/uk-mi…

@StavrosZenios @DemPapadakis @makdrou Πράγματι πρόκειται για νέα ήθη στις δυτικές κοινωνίες. Μέχρι πρόσφατα μόνο στους καννίβαλους τα συναντούσαμε

On this I disagree with @karpathy: Writing *IS* thinking! I am adopting his main idea to organize and engage with my research corpus, but not to skip the reading, metabolizing, and writing. Skipping these three, I will atrophy and become obsolete.

@lugaricano Luis, your post prompted me to check with @pangram a piece I did last night. The 82% did not surprise me since I interacted extensively with the AI to produce the piece. But the 18% it marked as AI-generated is the part I totally wrote myself. Admittedly, just one test.

Supposedly smart people try to smear all AI detectors by association with a terrible one. People who want to argue AI detection fails need to use Pangram Labs.

The fine line: "You can't outsource reading and its metabolisation." Use LLM to your heart's content but don't cross the line.