Louis Guthmann zk/acc

Here it goes: ZKP is to computation what zipping is to a file. Someone has done the computation and compresses it for you to check it. Sometime this computation hides things sometimes it does not… When hides, it aims at privacy. When it does not, it aims at scaling

Codex refreshing limits at random time is pretty awesome! Thanks @thsottiaux

Introducing the Monastery for AI-native founders. A single builder can now outperform a publicly traded company. $2 million. 12 weeks. Do the impossible.

@defidave Wait until you look at the number of chabad Houses. I quizzed Vitalik on it he could not have been more far off

Hasedics are the most successful network state by a mile

@cryptodavidw Actually disagree strongly. French politics is so stupid to me (other countries have more interesting politics) I just avoid the topic all together.

this video also highlights one cultural difference I've noticed between France and the US, when you hang out in France you discuss politics A LOT, while in the US it seems like everyone avoids it

@0xSero How do you get /goal with those harness? This feature alone dramatically improved my workflow.

I still stand by Droid being the best agent harness out there. I’ve tested everything under the sun. #1 Droid #2 Pi #3 Amp #4 OpenCode #5 Codex CLI I am still working on a few reviews but performance wise this has been my experience.

@DavidDeutschOxf @Dkarbowitz The French had the Dreyfus affairs years before and were about to elect a Jew as President du Conseil so not sure. I would be very curious of a vintage LLM trained in French, German and Polish of the time.

@GuthL @Dkarbowitz I would expect every other European language to have a less favourable opinion than that.

Yes, exactly. When Max Nordau told the First Zionist Congress in 1897 that the emancipation of Jews in Europe was fake, and prophesying disaster, he made an exception for Britain: there, he said, it was genuine. Anglosphere values are poison to the Pattern—but also vice-versa.

@DavidDeutschOxf @Dkarbowitz I don't know if you saw but talkie (the new LLM trained on pre 1930 English only data) had pretty much a similar stand which would appear to have been the mainstream view of the time

@Dkarbowitz Do you have a reference for that? Jabotinsky realised in 1920 that Britain could not be relied upon to protect the Jews of the Yishuv. But did he ever suspect that British Jews were not genuinely safe in Britain?

@NewSamawal @Shokooh1373 You can indeed see the leftovers all on the coast line near Nahariya

@Shokooh1373 It’s no longer there 💔

My great grandfather built this railway.

hi friends, i have some personal news to share with you. tomorrow is my last day at @StarkWareLtd. i vividly remember joining starkware with 0 knowledge (see what i did there) about crypto and i had no idea what i was walking into. close to 3 years later, i leave with so much more than knowledge.. memories, friendships, and a perspective i didnt have before. at its core, my job was always to make starknet feel a bit more human and accessible and i hope, in some way, i managed to do that for a few of you. thank you for everything and see you around 🧡

Knowing that the model has been trained on pre 1930 English corpus, it provides an insight in the mainstream position of the English world (UK and US) around the Jews pre WWII. Wonder how a model trained on other languages (German, French) would reveal about the period psyche

@GerardAraud J'imagine que vous ne payez pas. L'ai gratuite est inutile, l'IA payant est exceptionnelle

Étrange expérience. Je cherche une citation de Proust, dont je me souviens vaguement. ChatGPT me répond non seulement en l’inventant mais en me fournissant sa localisation inexistante dans l’œuvre. Heureusement que j’ai vérifié. Troublant.

@shabashewitz How did that happen? A grand father was in a young movement and then married a Jewish woman after the war?

@marmaduke091 What's the prompt?

🚨 First outputs from GPT 5.5 (Spud) Pro ! Three.js capability is next level. Blew my mind.

Guys look Claude helped me - a random guy in his basement - build a wetlab and do vibe genomics! I sequenced my whole genome despite zero lab experience, without my DNA leaving home! I put together my notes and a step by step guide here: vibe-genomics.replit.app It was a lot easier than I was expecting! Ultimately I hit ~16x coverage and compared my results against my 600k raw 23andme SNPs, and it held up!

@HeyGen Hyperframes

Prompt Guide: hyperframes.heygen.com/guides/prompti…

We built our launch video in Claude Code using HyperFrames. Now it's yours. Open source, agent-native framework. HTML to MP4. $ npx skills add heygen-com/hyperframes RT + Comment "HyperFrames" to get the full source code of this launch video (must follow)

@kobigurk Could this be used to prove signatures on large softwares/files instead of downloading locally and hash it?

It’d be interesting to see more efficient and post quantum variants of proximity signatures :) See the note here: eprint.iacr.org/2026/694 8/8

Happy to share Proximity signatures, a note with @angeris! What if we want to verify signatures on messages that are way too big? And in addition, get some guarantees on them? Verifiers are then in a problem! 1/8

@eniwhere_ @ceccon_me It is surpaijgnly better than Seedream by ByteDance who refuses even very uncontroversial dead figures

@ceccon_me the biggest argument against AGI is Google models lol they don't even know what they are

> open gemini > select pro > select "create image" > type detailed image prompt > "I cannot directly create or generate images, but I can certainly help you conceptualize" does google actually have AI? has anyone ever managed to use their products?

@fuckpoasting @kr0der Codex also has a extremely annoying tendency to use subtext to title as place to put comments which makes the UX weird af

@kr0der for landing pages, idt its ever worth using an ai design, doesnt matter which one it is

i have officially churned from claude code to codex, i grossly underestimated how good gpt-5.4 fast truly is

@sergey_nog It is 59 bit of security for a grover algorithm and 118 in the classical sense.

@apruden08 Calling Avihu a random dude is pretty insulting... He is StarkWare most formidable and was also behind ColliderVM and ColliderScript.

Yesterday: Quantum physicists: publish paper about breakthrough resource reduction for Shor's algorithms. Bitcoiners: It's total BS quantum is fake! Today: One random guy from Starkware: publishes paper about a possible way to do emergency PQ transactions that require a dedicated GPU and $200 (no formal security proof) Bitcoiners: Quantum is solved we're saved!