October 16, 2026
The best of what first appeared in the trends on this day (UTC).
Zfinix/eleven
Agent-first reverse engineering for macOS, Windows and Linux: any ACP agent reads, renames and patches binaries, Android apps and GPU kernels next to you.
Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation
Intern Lumina U2 is a diffusion language model that uses multi-codebook discrete visual tokens to unify image, video, and 3D generation with text in a single backbone, achieving high performance and faster inference.
Netflix Chief Content Officer Bela Bajaria is leaving the company and will be replaced by former Disney and Fox executive Peter Rice, who starts October 26
In a major executive shakeup, Bela Bajaria is out at Netflix after 10 years at the streamer, most recently serving as its chief content officer.
Staying sane while trying to save the world: Highlights from Spencer Greenberg and Luisa Rodriguez
Many people have found value in this 80,000 Hours podcast episode from July 2026: Spencer Greenberg on staying sane while trying to save the world. The highlights below are Julia's picks of the best bits, lightly reformatted, for those who…
Apple routed a graphite pipe through the iPhone Duo's hinge to dissipate heat
In an exclusive interview with PCMag, Apple’s silicon and thermal teams explain the engineering that lets the A20 Pro chip run hotter and faster without cooking your phone.
0xProbabillity/house-control
Your home as a live 3D model in the Mac menu bar and on iPhone — lights and ACs via Smart Life, Home Assistant, Philips Hue and Apple Home.
JevAgentHarness: Benchmarking, Deploying and Training Decision Models That Share Control with LLMs
A harness that delegates agent tool choices to small decision models via fitted confidence thresholds reduces cost and latency while preserving task success, supported by a new benchmark and a 4B model trained on agent error states.
Two US officials and a Western official say the order to limit Ukraine's access to SpaceX's Starshield came from inside the White House
The order to limit a satellite system used by Ukraine for battlefield communications and drone operations came from inside the White House …
SWE-at-Scale: Scalable Issue Synthesis from Historical Bug Fixes
SWE-at-Scale synthesizes realistic software engineering tasks by inverting real-world fixes to generate diverse, executable bugs for evaluating coding agents.
Mental Health Navigator site for EA and AI safety
The recently updated EA Mental Health Navigator site provides resources on mental health in effective altruism and AI safety: * Mental health professionals (such as therapists and psychiatrists) recommended by people in these spaces (thera…
MyLuxy/overdrive
Overdrive, a PS5 homebrew game store with 1,800+ games.
FreeSolo: Linear-time Sparse Attention with Zero-Shot Length Extrapolation
FreeSolo introduces a native sparse attention mechanism with short-range and long-range branches that achieves linear-time complexity and improves zero-shot length extrapolation without costly long-context fine-tuning.
OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding
OmniCompass is a holistic benchmark for omni-modal video grounding and understanding that evaluates vision-language and audio-visual reasoning through temporal grounding, spatial grounding, and question answering tasks.
The White House's AI task force begins its 120-day window to evaluate AI, including producing a risk report, implementing the White House Accord on SI, and more
The White House's AI task force has formally adopted its charter, kicking off the short 120-day window it's been given to answer …
i12bp8/decomp.wtf
What The Function: your AI agent decompiles games you own into C that compiles to the exact same bytes. Runs locally.
Exploring Intrapersonal Game Theory
Work done as part of the MATS 9.1 extension program, mentored by Richard Ngo. 0: Introduction A previous post of mine presents a toy model of self-sabotage in intrapersonal games. This post applies the idea of cross-temporal incoherence to…
Speculative Sampling in Interactive Video Generation
Speculative Sampling enables interactive video generation with large bidirectional windows by caching predictions and delivering them incrementally, reducing latency without sacrificing quality.
Speculative Forcing for Efficient Autoregressive Video Generation
Speculative Forcing accelerates few-step autoregressive video generation by accepting early token drafts and skipping unnecessary denoising steps while preserving output quality.
BeBox is a Haiku application that emulates Be Inc.'s BeBox
Haiku native PowerPC and Hobbit BeBox Emulator. Contribute to johnrash/bebox development by creating an account on GitHub.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction
CalOn is a compact unified encoder that jointly performs entity, relation, and event extraction across diverse schemas by encoding text with type definitions and using shared probabilistic reasoning over combined type evidence.
datalab-to/indra
Fast CPU OCR: line detection, CTC recognition, reading order, markdown
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference
PN-Mix improves low-bit KV cache quantization by identifying salient RoPE channel pairs via query-key norms, achieving strong accuracy at lower bit rates.
Opsin, which helps companies show where AI agents are active, what data they access, and flag risky behavior, raised $21M led by SYN Ventures and Race Capital
AI agents are starting to outnumber humans in many high-tech workplaces, and they can be far trickier to monitor.
OpenAI continues to work to grow its ad business in ChatGPT, cutting ad rates, chasing small and midsize businesses, and hiring some new talent from Meta
As OpenAI has jockeyed with Anthropic to win over developers with new models and tools this year, its advertising business at times has seemed like an afterthought.
Three perspectives on corrigibility - shards, simulators, singular learning
This post was first drafted as part of the ML4Good bootcamp. Would it be possible to implement a corrigible agent? Is that likely to happen given current trajectories? Historically, the answer was a resounding "no", but with more recent th…
CoTranslator: Reverse to Forward Chain-of-Thought Translator for Reasoning Distillation
CoTranslator translates reverse reasoning traces to match verified forward reasoning distributions, enabling efficient and transferable reasoning distillation for student models.
GetLostIn/CTR-NitroFueled-PC
Crash Team Racing: Nitro-Fueled (PS4) running natively on Windows 10/11. Windows fork of SacredFire, built on bbport and shadPS4. No game files included.
MFVSR: A Multi-Style and Fine-Grained Benchmark for Video Subtitle Removal
MFVSR introduces a large-scale, multi-style benchmark for video subtitle removal that evaluates restoration fidelity, preservation, temporal consistency, and reliability through fine-grained, spatially decoupled protocols.