Top

Show HN: Scry, programmable internet search w/ congestion pricingscry.io

Scry is a read‑only, SQL‑like search service that lets LLM agents query billions of public‑web records (e.g. Reddit, Hacker News, academic papers) via a single MCP or HTTP endpoint. It enforces per‑query limits (deadline, memory, row cap), offers schema discovery, explain, rerank, and compile APIs, and reports freshness per source (e.g. Hacker News <15 min). In a DeepSearchQA benchmark Scry‑enabl…

Cache-to-Cache: Direct Semantic Communication Between Large Language Modelsarxiv.org

Grand MS-DOS Gaming General MIDI Showdownjohnnovak.net

The author records 46 classic DOS game soundtracks on seven MIDI modules (Roland SC‑55, SC‑VA, Yamaha MU80, S‑YXG50, etc.), shares 322 lossless recordings plus REAPER project files, and documents a repeatable workflow for extracting MIDI from DOSBox, cleaning it, and rendering audio on hardware and software synths.

Reviving the language that brought us the Jak & Daxter Seriesopengoal.dev

Self-Evolving Search Indexarxiv.org

The paper introduces SELF-INDEX, a framework that lets a search index automatically diagnose retrieval failures, revise its keys, and validate changes, using a query simulator to anticipate future queries. Experiments show consistent gains across corpora and downstream LLM agents.

Tell the speakers that you liked their talksohhelloana.blog

The author shares personal experiences encouraging conference attendees to approach speakers and give feedback, and urges speakers and organizers to facilitate those conversations. The post is a reminder that a simple compliment or chat can help speakers gauge impact and improve community connections.

Migrating the GitHub Copilot runtime to Rust, using Copilotgithub.blog

The Copilot agent runtime was rewritten from a 130‑k‑line TypeScript/Node.js codebase into a native Rust library (~830 k lines of Rust) to reduce startup latency, memory use, and improve reliability. The migration was done incrementally (in‑place) across 128 PRs, with AI‑generated code handling most of the work. The new runtime exposes a C ABI for in‑process embedding by all six Copilot SDK langu…

OpenAI's Misalignment Framework: A Tactical Bid to Preempt Global AI Governanceasiaai.fyi

Why MCP Was Always a Bad Ideamaharship.com

OmniEdu: Open Foundation Models for Learning and Teachingarxiv.org

OmniEdu is an open family of LLMs (4B‑27B) fine‑tuned on a curated, capability‑balanced educational corpus (≈70k examples, 16M tokens) covering subject competence, curriculum grounding, diagnostic reasoning, and pedagogical scaffolding. Across model scales it improves on K‑12 benchmarks (K12‑Bench EM 63.12 %/F1 76.69 %, MathFish 85.89 %, EDUMATH 86.95 %, MathTutorBench Scaffold 78.74 %) and achie…

CCC invites all model citizens to 40C3ccc.de

The Chaos Computer Club is announcing its 40th Congress (40C3) from Dec 27‑30 2026 in Hamburg, inviting speakers, artists, musicians, and punk performers via multiple calls for participation.

German Rheinmetall open-sources its Battlesuite connected weapon system protcolgithub.io

Rheinmetall has open‑sourced onboardapi 9.10.0, a C++ middleware that uses DDS (XTypes, XCDR2) for low‑latency, backward‑compatible data exchange between sensors and software. The library includes Java, C#/.NET, and Python wrappers and comprehensive docs.

When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Modelsarxiv.org

When2Think introduces a post‑training framework that lets a large reasoning model decide per‑instance how much reasoning depth to allocate, using difficulty‑aware reward shaping (IDAC) and verifier rewards. It cuts token usage by ~28% while boosting Pass@3 by 10% on AIME24 and reaches 40% Pass@3 on AIME25, outperforming compression and routing baselines.

Show HN: How Stale Is Your AI? Release age and training cutoff for 20 modelsjock.pl

The author curates release dates and training‑cutoff dates for 20 recent LLMs, exposing how many months each model lags behind its launch. He also provides JSON/txt files and instructions for agents to fetch this data to avoid stale assumptions.

OpenAI and Anthropic oversold AI security breachesnypost.com

Markdown in /srchtmx.org

The essay argues that Markdown should be treated as source code and checked into a /src/md directory alongside the implementation it describes. Keeping intent in diffable, reviewable Markdown lets teams generate code and tests from it while preserving locality for both humans and AI agents.

Lucasart's Afterlifewordpress.com

This blog post revisits Lucasart's 1996 game "Afterlife," a unique SimCity-like game where players manage Heaven and Hell, offering strategies for new and returning players. It details specific game mechanics, zoning tactics, and common pitfalls to help optimize soul collection and progress.

Have it both ways: stay discoverable in search while disallowing AI trainingcloudflare.com

Cloudflare adds a Disallow AI Training setting that lets sites stay indexed for search while blocking crawlers from using their content for AI model training. The feature replaces the generic Block AI Bots option with granular Search, Training, and Agent controls.

This Digital Radio Gets Messages to the World’s Remotest Locationsieee.org

HERMES is an open‑source digital shortwave radio system that runs on 20 W HF transceivers (3–30 MHz) and can reliably send files over 400–600 km links, with optional encryption. A pilot with Bangladeshi fishers demonstrated SOS GPS transmission and rescue without any terrestrial network.

Internet Phone Bookinternetphonebook.net

The Internet Phone Book is a physical, art‑focused publication that curates personal websites of designers, developers, writers, and other creators. The page lists upcoming tour events, library locations, and where to buy the book, and invites submissions. No technical content, code, or engineering insights are present.

Scott Jenson: Are we really going to use the same Desktop UX forever?youtube.com

VideoGen-Agent: Reinforcing Video Generation Agentsarxiv.org

VideoGen-Agent is a multimodal RL‑trained agent that orchestrates augmentation, generation, and verification tools to improve text‑to‑video synthesis on a new 600‑prompt benchmark (VABench). It lifts a base generator’s score from 56.5 to 75.6 (‑19.1 pts) and to 86.1 when the generation tools are upgraded, with 84.3% human preference over the strongest baseline.

An actively maintained and updated Motif fork actually existsosnews.com

A community‑maintained fork of the Motif X11 UI toolkit (https://github.com/thentenaar/motif) has released five versions in its first year, incorporating long‑stale upstream fixes and Gentoo patches. The original Motif project on SourceForge has been inactive for over two years, but this fork aims to keep Motif and related X11 components alive, hoping to eventually support a niche desktop environ…

Brood War Benchswerdlow.dev

Show HN: Foremerge – Catch intent conflicts between parallel coding agentsgithub.com

Foremerge is an open‑source Rust CLI that lets multiple AI coding agents publish their intended semantic changes before they edit files, detecting intent conflicts that Git's line‑based merge cannot see. It stores intents in a local SQLite DB under .git and reports advisory warnings without locking files.

Spymarks, Not Watermarksbrand.io

Vinix – A modern operating system written in Vvinix-os.org

The first new cat species discovered in 100 yearsnationalgeographic.com

A new small wild cat species, *Leopardus tilcayo*, was identified in Bolivia after genomic analysis of a specimen first found in 2017. The study, published in *Current Biology*, shows the cat diverged from related tiger cats ~1.4 Myr ago and suggests many small‑cat taxa are cryptic species, with implications for conservation assessments.

Region-Level Policy Optimization for Fine-grained MLLM Perceptionarxiv.org

Vision‑RL2 trains a lightweight proposal network via region‑level reinforcement learning to select high‑resolution evidence for multimodal LLMs, allowing coarse‑resolution localization and fine‑resolution recognition. Across six fine‑grained vision benchmarks it reduces visual token count by ~4× while matching or surpassing full‑resolution accuracy.

Jev introduces a new shape of LLM - System One, aka Decision Modelssimonwillison.net

Jev is a new “System One” decision model that consumes text and returns numeric probabilities (yes/no, choice, score) instead of generated text, priced at $0.042 / M input tokens with free output. It’s fast, cheap, and suited for classification‑type tasks, but its opaque numeric output raises bias and explainability concerns.