Aller au contenu

Riemann Hypothesis proof claims update & Daraxonrasib pancreatic cancer pill results - Hacker News (Nov 22, 2026)

22 novembre 2026 · 7 min

Notes de l’épisode

Support The Automated Daily directly:
15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support

Please support this podcast by checking out our sponsors:
- SurveyMonkey, Using AI to surface insights faster and reduce manual analysis time - https://get.surveymonkey.com/tad
- Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad
- Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad


Today's topics: Riemann Hypothesis proof claims update - The Lean and Mathlib communities say Levent Alpöge's Lean proof of the Riemann Hypothesis is formally verified by multiple proof checkers. Meanwhile, Alessandro Maria Masullo's separate RH proof draft has errors found with help from AI models. Daraxonrasib pancreatic cancer pill results - Early trial data for daraxonrasib, a KRAS-targeting pill, showed tumor shrinkage in over 60% of patients with metastatic pancreatic cancer. One-year survival reached 76.6%, though researchers say it is not a cure. Army detonating unstable lead azide - The U.S. Army plans to destroy a 20-year-old stockpile of recrystallized, shock-sensitive lead azide near Burlington, Iowa. Monitors will track lead-contaminated dust and soil after the detonation. Toyota Nissan Honda parts standardization - Toyota, Nissan, and Honda plan to standardize steel specs, plastics and resin protocols, and wiring harness color codes. The goal is to cut costs and compete with Chinese automakers. Studio Display Ethernet speed bottleneck - A blogger traced slow 2.5 Gbps Ethernet through an Apple Studio Display to USB burst limits in a Realtek adapter. A newer RTL8157 5 Gbps adapter restored full network speed. LookAtMyQuant model quantization compiler - LookAtMyQuant is an open-source MIT quantization compiler that assigns bits based on each weight's sensitivity. It claims 12x lower KL divergence than Q4_K_M GGUF on Gemma 4 E2B at the same size. GPT-6.1 Sol reads raw binaries - Researchers report GPT-6.1 Sol can read hex dumps, patch binaries, and write working ELF executables without compilers or disassemblers. It also analyzed malware, identifying C2 servers and API hashing. Steering the body beyond willpower - An essay argues that cravings, pain, sleep, and stress often respond to mental tricks and framing rather than willpower. It cites smoking cessation, sugar cravings, and pain syndromes. Stullensimulator sandwich-building web app - Stullensimulator is a playful German JavaScript web app for building and browsing virtual sandwiches with spreads and toppings.

-Urgent Post Claims Proposed Proof of GRH, Then Reports Problems
-Daraxonrasib Trial Shows Major Promise Against Advanced Pancreatic Cancer
-Lean community verifies formal proof of the Riemann Hypothesis
-Army Plans Controlled Detonation of Degraded Lead Azide Stockpile
-Toyota, Nissan, and Honda Plan to Standardize Parts to Cut Costs
-Monitor Hub Bottleneck Slashed Ethernet Speeds Until a New Adapter Fixed It
-LookAtMyQuant: A Quantization Compiler for Smaller, Higher-Fidelity Models
-Stullensimulator Homepage
-Getting the Mind to Steer the Body
-LLM Successfully Reads and Edits Executables Byte by Byte


Episode Transcript

Riemann Hypothesis proof claims update
We start with the Riemann Hypothesis story we've been following, which now has two very different threads. First, the Lean and Mathlib communities, along with the Lean FRO and the Mathlib Initiative, issued a joint statement about Levent Alpöge's Lean library proof released on November 20th. After a day of checking, they say they fully certify both the statement and the proof. The code depends only on Mathlib, it compiles, independent tools confirmed the theorem matches the formal Riemann Hypothesis statement, and several separate proof checkers, including Lean's own kernel, agree. That last part matters because it guards against both logical gaps and sneaky code. If it holds, this would be a landmark result confirmed by machine rather than only by traditional peer review. The other thread is going the opposite direction. Alessandro Maria Masullo's separate proposed proof, which Brian Conrad had been reviewing, was rushed out as a draft and presented as entirely human work. By the end of the same day, an update said several people had already found problems, some with AI assistance, and that it should not be treated as accepted unless those issues are resolved. Two claims, two very different levels of confidence.

Daraxonrasib pancreatic cancer pill results
Now to medicine, where there's genuinely encouraging news. A pill called daraxonrasib is drawing attention after early trial results in advanced, metastatic pancreatic cancer. Among 61 patients, nearly everyone saw their tumors stop growing, more than 60 percent saw them shrink or vanish, and in roughly one in twelve the cancer became undetectable. One-year survival was about 77 percent, compared with around 40 to 45 percent typically seen with chemotherapy alone. The drug targets KRAS mutations, which drive most pancreatic cancers and have long been notoriously hard to hit. Researchers are careful to say it's not a cure, and they need longer follow-up to see how lasting these responses are. Still, for one of the deadliest cancers, these are unusually strong numbers, and the approach could eventually extend to other KRAS-driven tumors.

Army detonating unstable lead azide
Next, a hazardous waste problem with no easy answer. The U.S. Army is preparing to blow up a large stockpile of lead azide near Burlington, Iowa, after legal challenges delayed the plan. Lead azide is a very sensitive explosive used in detonators, and this batch has sat underground for about 20 years. Over that time it recrystallized and became even more shock-sensitive, so moving it or treating it chemically is considered too dangerous. Destroying it in place is the only option officials see. The site has been fitted with barriers, soil and water measures, and weather monitoring, but the core worry remains lead-laden dust drifting over a wide area. Air and soil monitors will be watching closely.

Toyota Nissan Honda parts standardization
Over in the auto industry, Toyota, Nissan, and Honda are reportedly teaming up to standardize some parts. The first targets are steel quality criteria, plastics and resin protocols, and a shared color-coding system for wiring harnesses. These are pieces drivers never see, so brand identity shouldn't suffer. Domestic Japanese suppliers come first, with international ones later. The motivation is clear: save money on the invisible stuff and spend it on innovation, as pressure from Chinese competitors keeps growing.

Studio Display Ethernet speed bottleneck
Here's a satisfying debugging tale. One writer had a 2.5 gigabit internet connection that dropped to around 600 megabits whenever their USB-C Ethernet adapter was plugged into an Apple Studio Display. Plugged straight into the Mac, it was fine. Drivers and router flow control helped partway, but the real culprit turned out to be how the adapter talked over USB: it sent data in very small bursts, and behind the monitor's hub that capped how many packets it could push before dropping them. Swapping to a newer Realtek adapter that handles larger bursts fixed everything. The lesson: a monitor acting as a hub can quietly throttle your network in ways that are very hard to spot.

LookAtMyQuant model quantization compiler
In AI tooling, an open-source project called LookAtMyQuant takes a smarter approach to shrinking language models. Instead of compressing every weight the same way, it measures how much each piece hurts the model when compressed, then spends its size budget where it matters most. The author claims dramatically better output fidelity than a popular standard format at the same file size on a Gemma model, plus case studies on several larger models. If those claims hold up, it's another step toward running capable models cheaply on laptops and other modest hardware.

GPT-6.1 Sol reads raw binaries
And speaking of AI, a striking report says GPT-6.1 Sol can work directly with raw executable bytes, with no compiler, assembler, or disassembler. Given hex dumps, it correctly described what programs did and even spotted a real bug in upstream code. It patched binaries to add new features and wrote working executables from scratch. On real malware, it identified the command-and-control server and the way the sample hid its system calls, with many claims later confirmed. The authors think this could reshape reverse engineering and malware analysis, though limited context still makes large binaries a hurdle.

Steering the body beyond willpower
On a more personal note, an essay making the rounds argues that willpower and logic often fail at changing behavior, and that what works is a kind of mental trick that gets the body on board. The author points to quitting smoking, beating sugar cravings, and some pain and stress symptoms easing once a person finds the right framing. Books, rituals, even timers or vivid mental images can do what sheer effort can't. It's a thought-provoking take on how much our bodies, not our reasoning, steer outcomes.

Stullensimulator sandwich-building web app
And finally, something lighter: the Stullensimulator, a playful German web app for building virtual sandwiches, picking spreads and toppings, and browsing everyone else's creations. No big lesson here, just a reminder that the web is still a fun place to tinker.



Subscribe to edition specific feeds:
- Space news
* Apple Podcast English
* Spotify English
* RSS English Spanish French
- Top news
* Apple Podcast English Spanish French
* Spotify English Spanish French
* RSS English Spanish French
- Tech news
* Apple Podcast English Spanish French
* Spotify English Spanish Spanish
* RSS English Spanish French
- Hacker news
* Apple Podcast English Spanish French
* Spotify English Spanish French
* RSS English Spanish French
- AI news
* Apple Podcast English Spanish French
* Spotify English Spanish French
* RSS English Spanish French

Visit our website at https://theautomateddaily.com/
Send feedback to [email protected]
Youtube
LinkedIn
X (Twitter)

Tous les épisodes de The Automated Daily – Hacker News Edition

Paramètres

Affichage

Langue

Détermine les stations affichées en premier sur la page d’accueil.

Historique

Les stations et épisodes écoutés récemment et tes dernières recherches sont seulement enregistrés dans ce navigateur.

Notifications

Un court message quand un podcast que tu suis publie un nouvel épisode. Désactive un podcast précis avec la cloche sur sa page.

Compte

Enregistrez vos favoris – dans l’app et sur le site.

Raccourcis clavier

/
Ouvrir la recherche
Espace
Lecture ou pause
← →
Dans le lecteur : 15 s en arrière, 30 s en avant (épisodes) · dans une rangée : tuile suivante
↑ ↓
Dans une rangée : cible dans la tuile
Début / Fin
Dans une rangée : première ou dernière tuile
Esc
Fermer la fenêtre ou le menu
?
Cet aperçu
En cours de lecture

Voir la station

À suivre

    Ajoutez des épisodes avec « À suivre » – ils seront lus après l'épisode en cours.