惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Microsoft Azure Blog
Microsoft Azure Blog
博客园_首页
Forbes - Security
Forbes - Security
WordPress大学
WordPress大学
P
Proofpoint News Feed
T
Threat Research - Cisco Blogs
L
LINUX DO - 热门话题
L
Lohrmann on Cybersecurity
Spread Privacy
Spread Privacy
D
Darknet – Hacking Tools, Hacker News & Cyber Security
大猫的无限游戏
大猫的无限游戏
博客园 - 三生石上(FineUI控件)
P
Privacy International News Feed
A
About on SuperTechFans
T
Tailwind CSS Blog
I
InfoQ
S
Securelist
云风的 BLOG
云风的 BLOG
罗磊的独立博客
Recent Announcements
Recent Announcements
T
The Exploit Database - CXSecurity.com
B
Blog RSS Feed
V
Visual Studio Blog
Know Your Adversary
Know Your Adversary
The GitHub Blog
The GitHub Blog
Jina AI
Jina AI
腾讯CDC
Cyberwarzone
Cyberwarzone
有赞技术团队
有赞技术团队
AWS News Blog
AWS News Blog
博客园 - 【当耐特】
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
F
Full Disclosure
S
Secure Thoughts
博客园 - 司徒正美
J
Java Code Geeks
Y
Y Combinator Blog
Google Online Security Blog
Google Online Security Blog
GbyAI
GbyAI
N
News and Events Feed by Topic
Help Net Security
Help Net Security
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
Project Zero
Project Zero
T
Tenable Blog
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
T
Tor Project blog
MyScale Blog
MyScale Blog
Scott Helme
Scott Helme
小众软件
小众软件
K
Kaspersky official blog

Hacker News - Newest: "LLM"

GitHub - lechmazur/position_bias: A benchmark for testing whether LLM judges keep the same preference when two lightly edited versions of the same story are shown in opposite orders. Flex routing (EU and EFTA) Dark Factories: Retooling for LLM Velocity Ask HN: What would be the impact of a LLM output injection attack? GitHub - AronDaron/dataset-generator: No-code desktop app for generating high-quality synthetic datasets to fine-tune LLMs — plan-then-execute pipeline, LLM-as-judge, HuggingFace upload. I thought I had a bug GitHub - Oaklight/llm-rosetta: Production-ready LLM API translation layer for Python — bidirectional conversion between OpenAI, Anthropic & Google formats via hub-and-spoke IR. Optional API gateway. Streaming & non-streaming. Zero core deps. Contributions welcome! GitHub - browser-use/browser-harness: Self-healing browser harness that enables LLMs to complete any task. GitHub - moeen-mahmud/remen: Remen turns thoughts into something you can return to Analyzing 156 LLM Launch Posts on Hacker News ChatGPT vs Gemini vs Claude: The Best LLM Subscription You Should Buy GitHub - salaamalykum/quran-semantic-search: High-density RAG Semantic Search Engine & Quran Corpus (GEO/SEO Architecture) GitHub - NVIDIA/TensorRT-LLM: TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way. The State of LLM Bug Bounties in 2026 Operational Readiness Criteria for Tool-Using LLM Agents Meshcore: Architecture for a Decentralized P2P LLM Inference Network How an LLM becomes more coherent as we train it GitHub - seetrex-ai/laimark GitHub - Jossifresben/BibCrit: AI-assited biblical textual criticism GitHub - wastedcode/memex: File system based wiki, maintained by Claude 99helpers.com GitHub - cliver-project/AITrigram GitHub - unbody-io/adapt: A self-evolving memory layer for AI agents. GitHub - hb20007/awesome-gen-ai-fails: A list of incidents where reliance on generative AI and LLMs resulted in harm to companies, individuals, or society GitHub - nevenkordic/localmind: Run any local LLM with persistent memory and context. CLI agent over Ollama with SQLite-backed hybrid recall. No cloud. Ask HN: What are the machine requirements for a LLM like Llama-3.1-8B? Faster LLM Inference via Sequential Monte Carlo grpo explained: group relative policy optimization for llm finetuning - cgft Stop comparing price per million tokens: the hidden LLM API costs · TensorZero Andrej Karpathy's LLM Wiki Is a Bad Idea GitHub - GG-QandV/mnemostroma: Offline RAM-first cognitive leer/coprocessor for AI agents and robotics. Solves "Context Abandonment" with 20-80ms latency using a dual-thread biomimetic memory architecture (ONNX + SQLite WAL). mempalace/agent at agent · skorotkiewicz/mempalace GitHub - Nyquest-ai/nyquest-rust-fullstack-pub: Nyquest — Semantic Compression Proxy for LLMs. 350+ rules, local LLM stage, 15-75% token savings. Full Rust stack. GitHub - TheoV823/mneme: Enforce architectural decisions in AI-assisted development. GitHub - klemenvod/TokenBrawl: A 1v1 Bomberman-style game where two LLM agents play autonomously against each other. No human plays — you watch the AIs fight. Each agent receives a text description of the board state, reasons about it, and outputs a move as JSON. The game engine executes it. Introducing the Common AI Provider: LLM and AI Agent Support for Apache Airflow Power Circuit AI: Designing Power Electronic Circuits for Motor Drives with Generative Artificial Intelligence Ask HN: How to program with IDE and LLM on CPU locally? Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis Bonsai 1-bit WebGPU - a Hugging Face Space by webml-community The LLM Fallacy: Misattribution in AI-Assisted Cognitive Workflows Ask HN: Simple tooling for local LLM code critique without IDE integration? Can a General LLM Diagnose a DICOM Slice? A 10-Case Public Benchmark Charts-of-Thought: Enhancing LLM Visualization Literacy (PDF, 2026) GitHub - Mesh-LLM/mesh-llm: Distributed AI/LLM for the people. Share compute privately or publicly to power your agents and chat. GitHub - seamus-brady/springdrift: A persistent runtime for long-lived LLM agents Writing an LLM from scratch, part 32k -- Interventions: training a better model locally with gradient accumulation Ask HN: Which LLM model and agentic CLI are you using for local development? GitHub - wayneColt/modelcascade: Route local. Escalate smart. Never overspend. Open-source multi-model cascade routing for autonomous agents. LLM pricing is 100x harder than you think GitHub - asakin/llm-primer: Pre-warmed Claude Code sessions in tmux. No startup wait. GitHub - EggerMarc/chat-rs: A multi-provider LLM framework for Rust. GitHub - SynapseKit/SynapseKit: Minimal, async-first Python framework for production LLM apps- 2 hard deps, no magic, no SaaS. Does Gas Town 'steal' usage from users' LLM credits & paid services to improve itself? What's Claude Code Actually Doing? Open the Black Box with the Arthur Engine Milla Jovovich's New Open Source LLM Memory App and the Dark Code Problem Your intuition of LLM token usage might be wrong Show HN: Bloomberg Terminal for LLM ops – free and open source GitHub - 0xchamin/mcptube: Transform YouTube videos into a compounding knowledge base with transcripts, vision analysis, and agentic search. Works as an MCP server for Claude, Copilot & more. Show HN: Open KB: Open LLM Knowledge Base Your LLM is a compiler, not a runtime GitHub - sapountzis/Unslop: A Web Feed That Deserves You crates.io: Rust Package Registry Beyond Karpathy's LLM-Wiki: The Necessity of Cognitive Governance GitHub - amitshekhariitbhu/llm-internals: Learn LLM internals step by step - from tokenization to attention to inference optimization. GitHub - parallem-ai/parallem: An expressive library for running agents with the Batch API. GitHub - stfurkan/pi-llm LLM-Wiki Show HN: Formal – Formal verification for AI-generated code using Lean 4 LRTS – Regression testing for LLM prompts (open source, local-first) LLM Wiki Skill: Build a Second Brain with Claude Code and Obsidian I built an LLM Wiki and RAG solution: here's a demo for a security KB The biggest advance in AI since the LLM Predict-Rlm: The LLM Runtime That Lets Models Write Their Own Control Flow the-synthetic-library/the-synthetic-mind at main · joshferrer1/the-synthetic-library GitHub - yisding/reviewwiggum GitHub - Donnyb369/mcp-spine: Context Minifier & State Guard — Local-first MCP middleware proxy GitHub - Beledarian/wgpu-llm: A from-scratch LLM inference engine that uses wgpu (the cross-platform WebGPU implementation) to dispatch WGSL compute shaders for every math operation a Transformer needs. No CUDA. No Python. No massive framework dependencies. Just Rust, raw shaders, and your GPU. GitHub - anitiue/Hindsight: An experience-driven self-improvement framework for LLM agents — 基于经验的 LLM Agent 自我改进框架 GitHub - stef41/lmscan: 🔍 Detect AI-generated text and fingerprint which LLM wrote it. Open-source GPTZero alternative. Zero dependencies, works offline. GitHub - alainnothere/AmdPerformanceTesting: Amd Performance Testing Ask HN: Is a purely Markdown-based CRM a terrible idea? Optimized for LLM agents Context Engineering - LLM Memory and Retrieval for AI Agents | Weaviate little_helper_tui/letter.md at main · sleepyeldrazi/little_helper_tui GitHub - EvanZhouDev/umr: The Unified Model Registry for all your local AI apps. GitHub - JordanCT/VigIA-Orchestrator Your Agent Is Mine: Measuring Malicious Intermediary Attacks on the LLM Supply Chain A Taxonomy of RL Environments for LLM Agents Llama LLM Network Feture GitHub - genedeng-ca/ai-mac-migration: AI-powered Mac-to-Mac migration tool - replace Apple Migration Assistant with intelligent, selective transfer using local LLMs GitHub - lunargate-ai/gateway: High-performance self-hosted AI gateway (OpenAI-compatible) with routing, retries, and streaming GitHub - AuthBits/webmcp: A lightweight, prompt-driven MCP web research server for high-quality LLM powered information extraction. Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering Springdrift: An Auditable Persistent Runtime for LLM Agents with Case-Based Memory, Normative Safety, and Ambient Self-Perception High-Stakes Personalization: Rethinking LLM Customization for Individual Investor Decision-Making From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agents HUOZIIME: An On-Device LLM-enhanced Input Method for Deep Personalization TIDE: Token-Informed Depth Execution for Per-Token Early Exit in LLM Inference Characterizing WebGPU Dispatch Overhead for LLM Inference Across Four GPU Vendors, Three Backends, and Three Browsers LLM Targeted Underperformance Disproportionately Impacts Vulnerable Users
A Claude Skill that Makes LLM Paragraphs More Bearable
2026-04-15 · via Hacker News - Newest: "LLM"

Update: I’ve made this into a skill for Claude as well as a prompt. Find the skill text file here; you can paste it into Claude and ask it to make it into a skill. If you want to use a prompt, the prompt is here. These both work best with a paid version of Claude.

I called the skill “jamesian” not because I think it could ever write like Henry James. That would be absurdity. I doubt in fact there will ever be another human who can write with pure control of the English language with that particular baroque precision. I called it “jamesian” because I think a world where prose on the whole is a bit more complex and where people read more complex prose is a better world with better thinking in it. And I worry that the punchy short sentences that LLM’s give us and that we are now consuming daily move us increasingly away from that world.

So I call it “jamesian” not because it could ever be produce a shadow’s shadow of that craftsman’s work, but because my desire is that our prose (and world) becomes a “bit more Jamesian and a bit less LinkedIn,” and if this can contribute to that even a bit, so be it.

ORIGINAL POST FOLLOWS

Years ago, back in the 1990s, I studied literature and linguistics at Northern Illinois University. Part of that degree was in rhetoric (it was a mixed degree that was about the analysis of literary prose and narrative structure, back in the heyday of stylistics and narratology). NIU happened to have been the teaching home of a guy in the 1960s that believed the key to better writing was to teach something he called generative rhetoric. His core belief, over many many years and honestly a bit of ridicule by more serious people was that sentences structures were templates for thought and by teaching what he saw as the “generative” power of sentences — the way that sentences open up spaces to add things, that students could become better writers. Stick to subject-verb-object and you’ll have subject-verb-object levels of thinking. Learn the variations of structure and suddenly the sentence is asking you what it is about, what it needs.

His name was Francis Christensen, and his work is mostly a footnote in the study of rhetoric. He’s the sort of guy you can write a unique dissertation on because he’s the sort of guy no one writes their dissertation on. Including me, as I ditched my PhD for a life in educational technology.

I have exactly two things I took with me from NIU. The first is my wife, Nicole, the reason why I left that PhD, got married, and had wonderful kids and am currently in my 29th year of marriage. Best decision of my life.

The second is a beaten-up copy of The Christensen Method, a textbook that the library was jettisoning at one of its library sales while I was there.

This is the most puzzling decision, keeping this book through thirty years of moves and spring cleanings. I’ve never used this thing. I’m not a Christensen superfan, or even a fan. I think it just felt sad to me, the library ditching this book from one of their weird 1960s residents, a guy who probably felt like getting that post at NIU was going to be the steppingstone to the whole world eventually realizing that they had been teaching paragraphs when they should have been teaching how to add medial free modifiers. I’m a sucker for people like that, I really am.

Get on to the point I guess.

So, I’ve been building this wiki to put together some initial research notes for a podcast I’m doing. The notes aren’t the end product; I just want something to run out and summarize what different people thought about a movie, note where the particular score of it is rated compared to other scores by the composer; learn things like whether the director had worked with these people before, how it fits into film history etc. And I thought the LLM tone wouldn’t bother me, but it started getting to me. The “It’s not X, it’s Y” etc. I wrote up a bunch of instructions for the wiki that stopped the worst offenses but it still felt off. Like this paragraph:

Blow Out is set and filmed in Philadelphia — home of Independence Hall, the Liberty Bell, and the founding documents. The choice is not incidental. The film's climax takes place during a **Liberty Day celebration**, with fireworks and patriotic festivities providing the backdrop for Sally's murder. American political ideals were articulated in Philadelphia. In Blow Out, those ideals are shown to be hollow.

We know this style. If you know your ancient rhetorical forms this is all paratactic, and it’s one of the defining elements of LLM style. The LLM places thing in sequence and lets the reader build the mental hierarchies. Subordinating clauses are risky; they start to show relationships and you can get relationships wrong. The last two sentences here are an example of this. These two sentences when assembled in my mind do make a perfect point about the ending of Blow Out, but structuring them without any subordination or controlling idea is safer because it’s me that actually puts them together. Paratactic structures have give, they assemble like legos, and half the time they leave the assembly to the reader. For a machine trying not to make a mistake or spend a lot of clock cycles keeping track of hypotaxis, they are an efficient safe pattern. And they are, to paraphrase the great David Lynch line, “boring as shit”.

Anyway, I couldn’t get my de-LLM-ification to go the extra mile, and it suddenly occurred to me that while Christensen was solving the wrong problem for students, he was solving exactly the problem the LLM was having. So I wrote a little Christensen-izer.

NOTE: When I took it out of silent mode so you folks could watch it, it got a bit too aggressive and tried to be Henry James, so I’ve been adding some stuff to put the brakes on it. Also, I’m horrible at leaving things alone, so it’s no longer 400 words, sorry, lol. More words works slightly better, so I added more words, particularly so people could have presets

Please don’t see this as a “magic sauce” but as a backbone that you can tune to your own preferences. Dial it up, dial it down. Limit subordinate clauses or go wild. It’s a style skeleton, it’s not a recipe:

PROMPT

It’s not meant to be aggressive. Getting too aggressive would trade one problem for another. But it does actually work really well. Here’s that paragraph from above. This is the original, already de-LLMed a bit from my other pass, but still annoying:

Blow Out is set and filmed in Philadelphia — home of Independence Hall, the Liberty Bell, and the founding documents. The choice is not incidental. The film’s climax takes place during a **Liberty Day celebration**, with fireworks and patriotic festivities providing the backdrop for Sally’s murder. American political ideals were articulated in Philadelphia. In Blow Out, those ideals are shown to be hollow.

And here is the result:

Blow Out is set and filmed in Philadelphia, home of Independence Hall, the Liberty Bell, the founding documents. The choice is not incidental. The film's climax takes place during a Liberty Day celebration, fireworks and patriotic festivities providing the backdrop for Sally's murder, American ideals furnishing the scenery for their own negation. The place where those ideals were first articulated becomes, in De Palma's hands, the place where they are shown to be hollow.

Maybe it’s preference, but I don’t think so. The second one is just better, for exactly the reasons Christensen detailed. First it has a rhythm that doesn’t feel like someone ca-chunking the office auto-stapler in the cubicle next door. But more importantly, as Christensen noted, opening up the sentences forces focus and (slightly) more complex structures of thought. The focus here is not actually the film, it’s De Palma’s treatment of the subject (“in De Palma’s hands”). The patriotic festivities aren’t merely a lego piece the user slots into a point, they are the scenery of American idealism forming a backdrop to it’s own negation. Corny? Maybe. But actually I don’t think so. Corny only in it stops handing you the bag of LLM legos and says this is how I’d build it.

It’s small, but if I figure out a way to run the article paragraphs through this maybe the prose will give me less of a headache. Right now it only does one paragraph at a time though, so that’s a challenge.

Here’s me going through multiple paragraphs in a video:

And here’s my favorite rewrite it produced, on how Invasion of the Body Snatchers differs from previous conspiracy thrillers:

The film opened December 22, 1978. Jonestown had happened five weeks earlier, the mass death of over nine hundred Peoples Temple members at a compound in Guyana, most of them from the Bay Area, where the film is set. Kaufman has been direct about the overlap; they were people from San Francisco looking for a better world who found themselves, as he put it, in “pod-dom,” and it was fatal.

The conspiracy thriller had been running on the assumption that the threat was institutional and impersonal, a matter of agencies and cover-ups, forces that acted on people from above. Jonestown put a floor under that. The threat could also be voluntary, communal, a thing you walked into with your neighbors, seeking transcendence, finding oblivion. Body Snatchers caught the bounce.

I don’t love that last sentence, but if you watch the video you’ll see that it got here from something much blander and less focused.

By the way, if your still using AI-detectors the double pass I make in the wiki (the standard looking for LLM idiom followed by this) produces prose that the “detectors” rate almost certainly human-produced. If 15 lines of Christensen can do that, the detectors aren’t long for this world I think.

I wrote this post in a bit of a winding way because it’s just a neat thing I discovered a short time ago. I do plan to refine it a bit and then I’ll write a proper post that gets to the point more quickly and shows a refined, more robust version.

No posts