A joint statement from the Lean and Mathlib communities says Levent Alpöge's 17-million-line Lean proof of the Riemann Hypothesis, released on November 20, 2026, compiles and checks out after about a day of community verification. →
The MIT-licensed LookAtMyQuant compiler squeezes thinkingmachines/Inkling-2-Mini from 18.89 GB into four MLX builds, the smallest 5.38 GB. The results are the authors' own. →
GnuPG released GPGME 2.3.1 to fix a critical double free that a crafted OpenPGP key with a very long user id can trigger. The advisory says remote code execution can't be ruled out and urges an immediate update. →
After security researchers reported that some inference providers had data compromised or were not upholding their data policies, OpenRouter off-boarded three providers and removed 125 endpoints from Zero Data Retention eligibility. →
pwn.ai details Upload2Shell (CVE-2026-125396), a use-after-free in Ghostscript's Type 1 font loader that lets a crafted document break out of -dSAFER and run commands. Ghostscript 10.08.0 and several distribution and application fixes are out. →
In a follow-up to his 2022 post, Raph Levien says AI is speeding up interest in Rust GUI, and he predicts a burst of experiments. He also covers GPUI, mobile, text layout, compositors and immediate mode. →
Leonida is an early open-source project that aims to run the Xbox Series version of GTA 6 natively on PC by recompiling it from the user's own copy. Its README says plainly that the game does not boot yet. →
A blogger on a 2.5 gig line found his Mac downloading at 600 Mbps through a Studio Display XDR. The cause was how many packets his USB-C Ethernet adapter sends per burst, and a 5G adapter fixed it. →
Nearly three years after Apple began allowing alternative browser engines on iOS in the EU, Mozilla says not one developer has shipped a non-WebKit browser, because Apple's conditions remain unworkable. →
Zero C#, a fork of bflat, compiles C# into small native programs with no .NET runtime and no garbage collector. The same source targets Linux, macOS, Windows, WebAssembly and DOS on a 386. →
A blog post says gpt-6.1-sol, given only xxd hex dumps, identified programs, patched binaries, wrote working ELF64 files and analyzed a malware sample, with no compiler or disassembler. The results are self-reported and small-scale. →
A new paper presents HiUni, which extends a pretrained vision language model to both understand and generate images while reusing the same ViT and dense Transformer backbone for both. →
Henrijs Dandzbergs, 16, is in Y Combinator's fall 2026 cohort with Polemos Industries, which aims to build cheap, jam-resistant, AI-guided missiles. He hopes to show a prototype by December. →
The Decoder·🔥59Glonce rating59Impact40Novelty45Relevance100Surprise40Players60Credibility70
An Nvidia research team says processing entire novels at once is often overkill and expensive, and that in many tests periodic summaries and targeted lookups worked about as well, The Decoder reports. →
PcapAnalyzer, an open-source beta under Apache 2.0, reads .pcap and .pcapng captures on your own machine and runs 27 problem detectors. An optional AI investigation uses your own Anthropic API key. →
gopgext compiles ordinary Go code with TinyGo into PostgreSQL extensions, generating the C glue, install script and control file. Its page says ports of contrib modules pass upstream's regression tests. →
The Decoder·🔥57Glonce rating57Impact45Novelty40Relevance95Surprise25Players90Credibility70
OpenAI released copy-ready prompt templates for GPT-6.2 Astra and Sol covering writing style, initiative and coding, and told developers to review their system prompts. The Responses API now discards earlier reasoning in some cases, and temperature and top_p are gone. →
The Verge·🔥56Glonce rating56Impact55Novelty40Relevance70Surprise35Players90Credibility75
In a Verge Decoder interview, senior AI reporter Hayden Field argues that Anthropic's Nasdaq listing, with roughly $900 billion in compute bills ahead, will test whether its safety branding survives pressure to grow revenue. →
OpenAI has released a guide, dated November 19, 2026, that sets out a reference architecture for running supervised security investigations with Daybreak Red in an isolated environment. →
The Verge·🔥52Glonce rating52Impact55Novelty30Relevance75Surprise30Players45Credibility85
After a largely fruitless week in Geneva, countries agreed to 'start discussing options' for rules on autonomous weapons. The Red Cross had wanted binding rules drafted within a year; the compromise gives two years to propose regulations. →
ZeroVLN is a slow-fast vision-language navigation framework that skips navigation-domain supervised fine-tuning of its VLM brain. The authors say it beats SFT-based hierarchical methods on VLN-CE benchmarks while barely changing the brain's weights. →
A preprint measures how much different reviews of the same paper repeat each other's critiques. At ICLR, overlap rose from 12.0% in 2026 to 21.3% in 2027, and reviews flagged as LLM-written overlap more than unflagged ones. →
Mind2Act World is a 30-task simulated robot manipulation benchmark that separates reasoning, execution and their composition. The strongest agent succeeds in 52.7% of Mind and 39.0% of Act episodes, but in only 12.0% of Mind2Act episodes. →
Pandex serves legislation and highest-court decisions of the EU and seven European countries through a REST API and an MCP server. It claims 2.8M documents, with 33 of 41 institutions held complete. →
Microdrama startup Shortical premiered 'Inevitable', a wholly AI-generated series built on Israeli actor Aki Avni's licensed face. It also invites others to 'license your face', but says no contracts have been signed through the platform yet. →
The Decoder·🔥47Glonce rating47Impact35Novelty40Relevance60Surprise55Players25Credibility70
A New Yorker report, summarised by The Decoder, says Maple, a Vermont Buddhist monastery, recruits AI-industry workers. Former members accuse it of manipulation and surveillance; its leadership denies almost all of the claims. →
Neuroscientist Anil Seth, winner of the 2025 Berggruen Prize Essay Competition, argues in Noema that intelligence and consciousness are different things, and that human biases push us to treat them as one. →
Simon Willison·🔥46Glonce rating46Impact35Novelty45Relevance70Surprise25Players20Credibility75
mod_wsgi creator Graham Dumpleton is writing Muru, an AI-assisted Rust server for ASGI and WSGI Python apps, and publishing a detailed account of it as he goes. Simon Willison flags the series. →
Take-Two, parent of Rockstar Games, estimates Grand Theft Auto 6 reached $3.8bn (£2.9bn) in retail sales on its first full day, which the BBC calls the biggest entertainment launch in history. The figures are company estimates through end of day Thursday. →
A developer building Arcana, his own Magic: The Gathering rules engine for RL experiments, found random-play test games stalling. One emblem resolution with thousands of Zombies took 35 seconds on a laptop. →
A new statistical test, AltRA-Test, checks whether an LLM can stand in for human annotators even when different items were labelled by different people. Across 20 datasets, LLMs pass in about half of the setups. →
A new preprint proposes AFFECT, an auxiliary objective that teaches world models how small action changes alter what happens next. On DROID, correlation with real-robot success rises from 0.47 to 0.79. →
A preprint introduces VSpark, a speculative-decoding method for autoregressive vision-language-action models. The authors report 1.28x to 2.10x speedups across four VLAs and four manipulation benchmarks, with lossless decoding. →
A new diffusion distillation method, Posterior Matching Distillation (PMD), aims to keep sample diversity while compressing a teacher into few steps. The authors report the best FID among 4-step methods on SDXL and FLUX.2 [klein] 4B. →
A preprint proposes Skill-FT, which uses natural-language skills to steer a teacher model's rollouts before fine-tuning, and AutoMaster, an LLM agent that designs those skills. The authors report mean gains of 1.42-8.99 percentage points over vanilla SFT. →
Wired's Kernel Panic newsletter says last week's outages were not an AI hacking spree and have routine explanations. But it argues AI-driven bug discovery is pushing patching to the limit and starting to play a role in real-world failures. →
National cyber director Sean Cairncross is leaving the Trump administration after a week of outages hit companies and government agencies. Defense Department CIO Kirsten Davies is set to replace him in an acting role. →
Ars Technica·🔥39Glonce rating39Impact35Novelty40Relevance45Surprise20Players40Credibility90
The FCC has ordered California to stop applying its Carrier of Last Resort rules to AT&T, clearing the way for AT&T to end basic phone service for 184,000 residential and 15,000 business customers on June 1, 2027. California could still sue. →
The Verge·🔥39Glonce rating39Impact35Novelty45Relevance40Surprise20Players35Credibility85
Chess.com is adding 10 games, from mahjong to sudoku, in a Classic Games public beta on web, iOS and Android. Its chief growth officer says AI sped up the coding but not the art and design. →
An essay on OpenAI's site, adapted from a Harvard lecture, names an unspoken expectation that humans will hand control of world affairs to machines, and argues against it. The author says the views are personal, not OpenAI's. →
A new paper describes WeVisForensics, an agent framework that spots AI-generated video from visual and audio cues. The authors report 95.2% average accuracy, 9.0% above the strongest prior detector. →
A new paper introduces FlashMorph2, an on-policy method for choosing which layers of a pretrained Transformer keep full attention when it is converted to a hybrid. It reports gains over FlashMorph on RULER, LongBench and recall-intensive tasks. →