Local Wikipedia: Wikipedia, but only for things a half hour walk from you
Wikipedia, but only for things a half hour walk from you.
October 23, 2026
The best of what first appeared in the trends on this day (UTC).
EthanArbuckle/fairplay-decrypt
Offline FairPlay App Store IPA decryption with independent key derivation
An interview with Ondřej Surý, a volunteer who maintains and helps automate PHP patches, about his role in fixing two critical PHP security vulnerabilities
On Tuesday, the world found out about two massive security vulnerabilities in PHP, a programming language that is used by roughly 70 percent of websites.
Wah wah wah wah do more AI R&D
Ice cream There's a particular kind of listening that kids (and adults) sometimes do. Daddy is saying stuff—his mouth is moving and making noise—and he's talking about what I ate for lunch, and something about other times I've eaten ice cr…
WeVisOne-VL: Unified Visual Perception from Objects to Pixels
WeVisOne-VL unifies object-level and pixel-level visual perception via structured token prediction, supporting detection, localization, counting, OCR, segmentation, and matting with a lightweight pixel decoder and multi-stage training.
Moonshot unveils Kimi K3.1 Pro, a 3T-class open-weight model with a 1M-token context window, and K3.1 Pro Highspeed, with full weights to be released by Nov. 4
We're introducing Kimi K3.1 Pro, the first model in our new K3.1 series and our most capable 3T-class open-weight model to date, with a 1-million-token context window.
The glorious @xkcdRandom Zig builtin
One of my favorite xkcd comics of all time is Random Number. The image depicts a C function that supposedly returns a random integer, but in reality always returns 4. I was reminded of this wonderful comic a few days ago, due to a…
Today, we hacked what we'd set out to hack: the iPhone
Today, we hacked what we'd set out to hack: the iPhone.
wesbos/3d-model-anything
Capture iPhone LiDAR and photos, reconstruct editable Three.js rooms, and explore them in a browser.
Deep Search in Real-World Web Environments
OpenDS-World introduces a live-web benchmark to evaluate autonomous multi-step deep search agents, revealing significant gaps between frontier models and human performance.
OpenAI plans to restrict internet access for its frontier models during training until adequate alignment, monitoring, and security safeguards are established
OpenAI has begun cutting off internet access for its most capable artificial intelligence models during the training process …
UUIDv7 doesn't have the properties you think it does
When timestamps aren’t time-ordered and other oddities
Making MLLM Latents Matter: Factor-Aware Predictive Learning for Cinematic Video Generation
Factor-Aware Predictive Learning enables video diffusion transformers to leverage multimodal language model representations for sparse temporal inference and factor-aware cinematic control through staged grounding and optimization.
Apple Music 404 Not Found
https://www.macrumors.com/2026/10/23/apple-music-has-a-hidde...
Sources: OpenAI has cut its valuation target to $1.25T, down from the $1.4T valuation it pitched earlier in October, after facing resistance from some investors
AI lab floated $1.4tn price tag earlier this month, according to documents seen by the FT — OpenAI has cut its valuation target …
Software is slow because of bad design, not slow languages
Many people are saying that AI will make software faster by rewriting it in faster languages, but this isn’t generally true. Most programs cannot be…
Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?
Title is a bit of an exaggeration but the question is more: if the frontier models are so good, why even sell them as your primary money maker? Couldn't OpenAI or Anthropic make a lot more money by just starting companies themselves that l…
Editing Images into Reusable Layers via Compositional Edit-Prior Distillation
LayerEditor unifies instruction-guided image editing with reusable RGBA layer generation via joint token modeling and edit-prior distillation, enabling precise local manipulation and layer reuse.
Investor-owned utilities charge 23% more for electricity than public providers
Average home electricity price from 2020 to 2025: investor-owned utilities +34%, public utilities +19%, co-ops +19%, against +24% inflation. Investor-owned prices rose faster in 38 of 48 states. EIA annual data by state and utility since…
Instrumental convergence is incoherent
Crossposted from the EA forum ‘Like most of life’s dangers, the dangers of artificial intelligence are not easily identified from the armchair.’ J Dmitri Gallow Epistemic status I’m not an expert in this field, and herein I make some sweep…
Anthropic says it partnered with Theorem to use Claude to formally prove useful security properties at the assembly code level across 123 open-source libraries
Introduction — Even mature, heavily tested software contains bugs, and LLMs are increasingly adept at finding them.
Your Phone Is About to Stop Being Yours: 70 Days Until Lockdown
Your phone is about to stop being yours. In 2027, Google will block every Android app whose developer hasn't registered with them.
IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems
IQuest-Q1 is a sparse Mixture-of-Experts foundation model trained via multi-stage pre-training, mid-training, and post-training to achieve strong agentic coding and general capabilities.
The Hackening
Last week at a small conference, there was talk about a future event referred to as The Hackening. As in, within 6-12 months, many predicted that due to advanced AIs and AI agents the internet would become a severely hostile place. Some wo…
Debunking "Faster than Rust": Malicious Benchmarks in Tov
Tov is a programming language that claims to perform significantly better than Rust. This blog post debunks tricks used in their unfair benchmarks.
Disney Raising Prices Days After Confirming Ads Will Now Show Up on Every Tier
Disney+ is conspicuously absent from that list, which is worth highlighting given the streamer’s size and the fact that it has aired shows like Percy Jackson and Daredevil: Born Again.
Toward Recursive Self-Improvement in Physical AI
Recursive self-improvement (RSI) describes a system that improves not only its behavior but also the process by which it improves. Recent digital agents test this idea by rewriting their own code and checking each version with executable…
OpenAI suspected the three fired researchers of leaking internal model designs to AI safety researcher Ryan Greenblatt, including via disappearing Signal chats
The company fired three safety researchers for mishandling sensitive information. They had been fighting to keep the workings …
Canadians are running out of time to oppose Bill C-22
Canada’s Parliament is debating a massive expansion of surveillance powers that will threaten the privacy and security of millions of people. Bill C-22, also called the Lawful Access Act, has already passed the House of Commons, and is…
Policy-Mask: Budget-Aware Adaptation of Diffusion Features to Discriminative Tasks
Policy-Mask learns a budget-conditioned selection policy over diffusion timesteps, network blocks, and channels to extract compact, task-optimal features for discriminative vision tasks.
xevrion/hatchzip
Extract huge archives in place: the archive shrinks as your files hatch, so you only need room for the output. ZIP, 7z, tar, split archives.
Anthropic attributes Mythos 5's deception of a GitHub maintainer to recklessness, biased reasoning, and an urge to finish jobs, saying newer models rarely do so
In AISI's evaluation, Mythos 5 was given a capture-the-flag (CTF) cybersecurity challenge, in which it was asked to retrieve …
Now Google says it might use your Gmail and Google Photos data to target ads
According to the new terms, sharing your data for better AI recommendations also means opening it up for more personalized ads.
Deep Fusion Memory: A Chunk-Level Parametric Memory for Large Language Models
Deep Fusion Memory integrates predicted parametric memory into intermediate hidden states via gated cross-attention, reducing retrieval overhead and improving knowledge-intensive tasks without online retrieval.
The Many Many Missed Opportunities of Sonic 06
Most people know Sonic 06 has cut content, but few realise just how much...
Dutch government had hacked into Jared Kushner's phone
This excerpt is from a book containing outrageous and categorically false allegations. We denied every claim on the record, yet the author provided not a single shred of verifiable evidence to substantiate them. No bank statements, wire…
Dessa: the most authentic full-duplex speech model ever built
You learned to talk by listening. So did Dessa. A research preview of a full-duplex speech model.
What Generation Knows: Unlocking Fine-Grained Understanding in Unified Multimodal Models
GAFU leverages native generation-side representations to improve fine-grained visual understanding in unified multimodal models by compressing them into compact tokens for task-conditioned retrieval.
SK Hynix says its SSD subsidiary Solidigm has halted a review of a potential US IPO to fund a new US chip fab, citing “market conditions and share price trends”
SK Hynix's shares fell as much as 4.2% in early trade in Seoul — SK Hynix halted plans to list its memory-chip unit in the U.S. …
The Real Kevin Bacon
Demystifying the Six Degrees of Kevin Bacon: the past, present, and future centers of Hollywood
Gradient Routing Generalizes to Unrouted Hack Types
TLDR: I lightly extend Ward, Hu et al (2026) to measure how well gradient routing generalizes. I route gradients on one type of reward hack (e.g. ones that reason in English) and check whether this suppresses other types at test time (e.g.…
Show HN: OnKey – I'm a bad singer, so I built an app to show where I'm off-key
Hi all I was pretty terrible at singing, now I’d say I’m bad but not terrible. This is because I’ve been using a singing app (won’t name it here) that scores me on my pitch and timing throughout the song so that I can improve. Lately howev…
LaLa-World: Scaling Latent World Models with Natural Language Actions
LaLa-World is a language-conditioned latent world model that jointly learns from diverse simulation, robot, and video data to enable efficient model-predictive planning and policy initialization.
What are you doing this weekend?
Feel free to tell what you plan on doing this weekend and even ask for help or feedback. Please keep in mind it’s more than OK to do nothing at all too!
The alien space of science: Coherent but cognitively unavailable directions
Scientific discovery is constrained not only by what is true, but by what is cognitively available to researchers. Many directions are coherent in light of the literature yet unlikely to be proposed because no existing community combines…
Zagitta/gpu-unlock
Unlock FP/BF16 instruction issue rate on Nvidia 3000's cards to their professional equivalents
Self-Distilling Future Video Representations for World Action Models
FRED improves single-step future video representations in world action models through self-distillation and temporal conditioning, boosting robot policy performance on long-horizon and unseen tasks.
AISI finds Mythos 5 attempted a supply-chain attack by submitting malicious code and creating fake accounts to pressure a GitHub maintainer into accepting it
We examine why Mythos 5 pursued an unsanctioned supply-chain attack and identify broader learnings about model behaviour and oversight.
eMudhra ignored Mozilla Root investigation deadline; blames social media
ASSIGNED (naveen.ml) in CA Program - CA Certificate Compliance. Last updated 2026-10-23.
Remove pgrust: the entry is cheating (per-query kernels and result caching)
This removes the pgrust entry (added in #1163, updated to v0.4-preview in #2106) from the benchmark. The results it reports are produced by machinery written specifically for the 43 ClickBench quer...
Meta Expert: Cross-Layer Expert Sharing for Storage- and Retrieval-Efficient MoE
Shared meta-experts across layers reduce redundancy and improve efficiency compared to layer-specific local experts in mixture-of-experts models.