Digest curado

Artículos fetched (60)

• Introducing Claude Opus 5
  anthropic-news· 24-jul

  Jul 24, 2026Product

• Investigating three real-world incidents in our cybersecurity evaluations
  anthropic-news· 30-jul

  Jul 30, 2026Frontier Red Team

• ClinLens: Towards Long-Horizon Coding Agents for Longitudinal Multimodal Clinical Data Science
  arxiv-ai· 31-jul

  arXiv:2607.26155v1 Announce Type: new Abstract: Clinical data-science agents must transform heterogeneous longitudinal records into auditable analyses, yet existing benchmarks largely isolate medical question answering, structured-table reasoning, or generic scientific repositories. We introduce CLINLENS, a benchmark of 200 executable tasks over five linked MIMIC resources spanning structured electronic health records, notes, electrocardiograms, chest radiographs, and echocardiograms. A 4 x 5 taxonomy crosses four patient-time scopes with five analysis capabilities. Program-first reverse synthesis pairs each bounded semi-raw package with an evaluator-private reference workflow and checks required artifacts, cohort and temporal semantics, and the final answer. On a fixed 126-task suite, th…

• When benchmark inferences do not compose: Projectibility in AI evaluation
  arxiv-ai· 31-jul

  arXiv:2607.26159v1 Announce Type: new Abstract: An AI benchmark result rarely reaches a consequential claim in one step. Evaluators generalize it to further cases, interpret it as evidence of capability, extrapolate it to new tasks, transport it to another system or site, and combine it with assumptions about human review and downstream consequences. Validity-centred approaches require evidence for each claim. This paper identifies a further epistemic problem: warranted links don't automatically make a warranted chain. The target of one study may not be the source of the next; system, population, outcome, or conditions may change at the interface; and shared data or model lineage may make apparently independent support dependent. Projectibility concerns whether a bounded extension from ob…

• GuideSkill: Evolving Executable LLM Agent Skills for Guideline-Grounded Clinical Reasoning
  arxiv-ai· 31-jul

  arXiv:2607.26160v1 Announce Type: new Abstract: Clinical practice guidelines (CPGs) encode diagnostic criteria, but LLM systems typically retrieve guideline text or absorb it through training rather than execute its rules. We introduce GuideSkill, an external reasoning layer that compiles disease-specific criteria into executable functions returning ordinal diagnostic-support scores. GuideSkill-Zero is initialized from guidelines, while GuideSkill-Evo uses case--diagnosis pairs to refine covered skills and add missing diagnoses. At inference, an LLM proposes a differential diagnosis, grounds the features required by each matched skill, and fuses its ranking with the executed skill scores. Across four benchmarks and four backbones, GuideSkill-Zero improves macro-average accuracy over guide…

• GoGoTB: Agentic RTL Verification with Specification-Grounded Coverage Closure
  arxiv-ai· 31-jul

  arXiv:2607.26181v1 Announce Type: new Abstract: Functional verification dominates integrated circuit (IC) front-end engineering effort, and a single missed bug that escapes to silicon can trigger a costly respin. Recent large language models (LLMs) offer new opportunities to automate this process, yet existing LLM-based approaches generate each component through independent single-turn calls with no shared context, leaving interface mismatches undetected and reported coverage disconnected from specification requirements. To address these challenges, we present GoGoTB, an agentic framework that achieves end-to-end verification closure through three subsystems: an agentic execution control layer, an evolvable knowledge system, and specification-grounded coverage closure. The execution contr…

• Position: Evaluation Scores Are Perishable Knowledge Claims
  arxiv-ai· 31-jul

  arXiv:2607.26191v1 Announce Type: new Abstract: Evaluation methodologies for language models increasingly combine multiple signals, from automated metrics and LLM-as-judge ratings to human assessments and benchmark suite results. When these signals are aggregated via averaging, evaluation confidence can then substantially exceed the reliability of the weakest signal: a phenomenon we call trust inflation in evaluation. We argue that evaluation scores should be treated as epistemic claims with three properties: formality (human evaluation provides stronger evidence than an automated metric), scope (a benchmark result applies to the tested distribution, not universally), and validity windows (benchmark results expire as contamination accumulates and distributions shift). Several converging r…

• TraceCoder: Explainable and Auditable Code Generation with Position-Key Snippet Versioning
  arxiv-ai· 31-jul

  arXiv:2607.26307v1 Announce Type: new Abstract: Contemporary LLM-based coding agents produce code as black-box outputs: the rationale behind each line is hidden, the evolution of the code through benchmark-driven repair is ephemeral, and post-hoc auditing is impossible. We present a code generation concept that addresses these shortcomings through three complementary mechanisms: (i) a relational snippet-history schema that records, per repair event, the benchmark reference, round number, failure text, and LLM explanation, enabling full provenance queries; (ii) a browser-based visualisation tool that renders this history as heat-mapped, hover-annotated source code; and (iii) a competitive fractional position-key indexing scheme with tree-node delimiters that assigns stable, lexicographical…

• Exploring Structures in Physics Problems: Can AI Agents Discover Statistical Mechanical Mappings?
  arxiv-ai· 31-jul

  arXiv:2607.26367v1 Announce Type: new Abstract: An important skill in theoretical physics is to recognize when a new problem can be transformed into a known model. We study this skill as an AI-agent task: can LLM-based agents discover statistical mechanical mappings from a raw partition function to a tractable representation? To probe this question, we introduce StatMechBench-v0, a benchmark of six Ising-type problems covering transfer-matrix methods, gauge-removable disorder, and planar/Pfaffian structure. We evaluate a simple propose-verify-revise agent across multiple LLMs and problem phrasings. The results show that numerical feedback often helps agents repair code and recover correct partition functions. However, agents can also pass the numerical checks while misidentifying the unde…

• CaM-Wolf: Causal-Aware Multimodal Agents for Social Deduction Games
  arxiv-ai· 31-jul

  arXiv:2607.26393v1 Announce Type: new Abstract: Social deduction games (SDGs) such as Werewolf have become challenging testbeds for AI agents. These games require complex social skills such as reasoning, deception, and collaboration. While recent advances in large language models (LLMs) have driven significant progress in SDG agents, current approaches are predominantly text-based, overlooking the multimodal nature that is fundamental to human social interaction. To bridge this gap, we introduce CaM-Wolf, the first SDG agent that integrates multimodal perception and generation. CaM-Wolf processes video inputs from other players, employs a causal-aware Reasoner trained via reinforcement learning to establish logical chains between observable behaviors and hidden roles, and presents itself …

• Probing the Origins of Reasoning Performance: Representational Quality for Mathematical Problem-Solving in RL vs. SFT Fine-Tuned Models
  arxiv-ai· 31-jul

  arXiv:2607.26119v1 Announce Type: new Abstract: Large reasoning models trained via reinforcement learning (RL) have been increasingly shown to outperform their supervised fine-tuned (SFT) counterparts on mathematical reasoning tasks; Yet the mechanistic basis for this advantage remains unclear. We therefore ask, what internal representational differences enable RL models' superior performance? Our work presents two converging lines of evidence: First, linear probes trained on layer-wise hidden states reveal that RL models tend to achieve higher accuracy in predicting answer correctness compared to SFT models, indicating more linearly separable and structured representations. Second, mean ablation studies show that RL models develop a hierarchical architecture where deeper layers become pr…

• Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems
  arxiv-ai· 31-jul

  arXiv:2607.26120v1 Announce Type: new Abstract: Large Language Models (LLMs)-powered multi-agent systems are increasingly deployed in mixed-motive environments, where agents operate under asymmetric information and strategic deception due to conflicting or hidden objectives. In these settings, misalignment with collective goals becomes a central concern. We propose a novel framework for evaluating objective misalignment using the social deduction game Werewolf, modifying the objective of a single agent while preserving its assigned role. Across LLMs from four different model families and sizes, four player roles, and three objective formulations, we introduce a dual analysis of the agents' internal reasoning and their public cheap-talk behavior (i.e costless, non-binding communication tha…

• How NVIDIA Builds Open Models for the Age of AI
  bytebytego· 27-jul

  Bryan Catanzaro, VP of Applied Deep Learning Research at NVIDIA, walked us through how his team builds the company’s open models, the reasoning behind…Jul 27 • ByteByteGo258218

• Why DoorDash, Instacart, and Uber Eats Integrated LLMs Into Search Three Different Ways
  bytebytego· 28-jul

  In this article, we will walk through their differing solutions and try to make sense of their choices and understand the pattern behind them.Jul 28 • ByteByteGo342418

• How ChatGPT Optimizes its Agent Loop: Harness, API, and Inference
  bytebytego· 29-jul

  To understand what techniques are adopted in frontier labs to make AI applications more efficient, we met with the OpenAI engineers who developed and…Jul 29 • ByteByteGo365115

• A Detailed Guide to Idempotency, Delivery Semantics, and Deduplication
  bytebytego· 30-jul

  What happens when a service sends a request to charge a customer, but the request times out with no response?14 hrs ago • ByteByteGo1473

• Loading...
  claude-changelog

  Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...Loading...

• Claude Platform release notes
  claude-changelog

  Release notesSubscribeCopy pageUpdates to the Claude Platform, including the Claude API, client SDKs, and the Claude Console.SubscribeCopy pageFor release notes on Claude Apps, see the Release notes for Claude Apps in the Claude Help Center.For updates to Claude Code, see the complete CHANGELOG.md in the claude-code repository. July 24, 2026 We've launched Claude Opus 5 (claude-opus-5), a step-change improvement over Claude Opus 4.8. Claude Opus 5 supports a 1M token context window (both the default and the maximum), 128k max output tokens, and thinking on by default, at $5 / $25 per MTok, the same pricing as Claude Opus 4.8. It's available on the Claude API, Claude in Amazon Bedrock, Claude on Google Cloud, and Claude in Microsoft Foundry. See What's new in Claude Opus 5 for new feat…

• Natural disasters and government interference: examining Q2 2026’s major Internet disruption events
  cloudflare· 28-jul

  Cloudflare Radar tracked Internet disruptions driven by natural disasters, government-mandated shutdowns, and DNSSEC key rollovers over the last quarter. This post analyzes traffic telemetry to explain how these events impacted global connectivity.

• We’re open-sourcing our privacy proxy CLI
  cloudflare· 27-jul

  pvcli is a curl-like tool designed to simplify the testing of complex privacy protocols like OHTTP.

• Dogfooding at scale: migrating cdnjs to Cloudflare’s Developer Platform
  cloudflare· 30-jul

  We moved cdnjs, serving 9 billion requests a day, entirely onto Cloudflare's Developer Platform. That means we’re running one of the Internet's busiest open-source CDNs on our own building blocks, and we pushed Workflows and Workers limits higher for everyone.

• BGP ORIGIN attribute manipulation and its impact on the Internet
  cloudflare· 24-jul

  By doing in-depth testing, we found nearly 70% of BGP paths experience ORIGIN attribute rewrites by transit providers seeking traffic advantages. We examine the global impact of this practice and argue for deprecating ORIGIN in route selection.

• Post-quantum authentication to origins is now supported
  cloudflare· 29-jul

  Cloudflare now supports post-quantum (PQ) authentication when connecting to customer origin servers via Authenticated Origin Pulls and Custom Origin Trust Store. This is the first step towards providing PQ authentication for all Cloudflare products.

• ansible/ansible
  github-trending

  Ansible is a radically simple IT automation platform that makes your applications and systems easier to deploy and maintain. Automate everything from code deployment to network configuration to cloud management, in a language that approaches plain English, using SSH, with no agents to install on remote systems. https://docs.ansible.com. Ansible Ansible is a radically simple IT automation system. It handles configuration management, application deployment, cloud provisioning, ad-hoc task execution, network automation, and multi-node orchestration. Ansible makes complex changes like zero-downtime rolling updates with load balancers easy. More information on the Ansible website. Design Principles Have an extremely simple setup process with a minimal learning curve. Manage machines quickly an…

• different-ai/openwork
  github-trending

  The open-source alternative to Claude Cowork (powered by opencode) OpenWork OpenWork is a free, open-source desktop app made for sharing AI workflows. It is an open-source alternative to Claude Cowork and Codex for macOS, Windows, and Linux. Add one OpenWork MCP to Codex, Claude Code, Cursor, or another compatible agent and reuse the same skills, MCPs, and connected services across your tools, teammates, and machines. Create something once, share it with coworkers or friends, or keep it for yourself. The desktop app is there when you want a dedicated workspace, but it is not required. You can use OpenWork from the agent you already have. For larger organizations, the admin interface lets you publish capabilities, manage access, and configure shared or per-user connections. Download OpenWo…

• dotnet/aspnetcore
  github-trending

  ASP.NET Core is a cross-platform .NET framework for building modern cloud-based web applications on Windows, Mac, or Linux. ASP.NET Core ASP.NET Core is an open-source and cross-platform framework for building modern cloud-based internet-connected applications, such as web apps, IoT apps, and mobile backends. ASP.NET Core apps run on .NET, a free, cross-platform, and open-source application runtime. It was architected to provide an optimized development framework for apps that are deployed to the cloud or run on-premises. It consists of modular components with minimal overhead, so you retain flexibility while constructing your solutions. You can develop and run your ASP.NET Core apps cross-platform on Windows, Mac, and Linux. Learn more about ASP.NET Core. Get started Follow the Getting S…

• huggingface/speech-to-speech
  github-trending

  Build local voice agents with open-source models Speech To Speech: Build voice agents with open-source models A low-latency, fully modular voice-agent pipeline: VAD -> STT -> LLM -> TTS, exposed through an OpenAI Realtime-compatible WebSocket API. Every component is swappable. The LLM slot speaks OpenAI-compatible protocols, so you can point it at a hosted provider, at HF Inference Providers, or at a vLLM or llama.cpp server on your own hardware for a fully local, fully open stack. This pipeline runs in production as the conversation backend for thousands of Reachy Mini robots. Quickstart pip install speech-to-speech export OPENAI_API_KEY=... speech-to-speech This starts an OpenAI Realtime-compatible server at ws://localhost:8765/v1/realtime using Parakeet TDT for local STT, an OpenAI-com…

• microsoft/AI-For-Beginners
  github-trending

  12 Weeks, 24 Lessons, AI for All! Artificial Intelligence for Beginners - A Curriculum AI For Beginners - Sketchnote by @girlie_mac Explore the world of Artificial Intelligence (AI) with our 12-week, 24-lesson curriculum! It includes practical lessons, quizzes, and labs. The curriculum is beginner-friendly and covers tools like TensorFlow and PyTorch, as well as ethics in AI 🌐 Multi-Language Support Supported via GitHub Action (Automated & Always Up-to-Date) Arabic | Bengali | Bulgarian | Burmese (Myanmar) | Chinese (Simplified) | Chinese (Traditional, Hong Kong) | Chinese (Traditional, Macau) | Chinese (Traditional, Taiwan) | Croatian | Czech | Danish | Dutch | Estonian | Finnish | French | German | Greek | Hebrew | Hindi | Hungarian | Indonesian | Italian | Japanese | Kannada | Khmer |…

• microsoft/PowerToys
  github-trending

  Microsoft PowerToys is a collection of utilities that supercharge productivity and customization on Windows Microsoft PowerToys Microsoft PowerToys is a collection of utilities that help you customize Windows and streamline everyday tasks. Installation · Documentation · Blog · Release notes 🔨 Utilities PowerToys includes over 30 utilities to help you customize and optimize your Windows experience: Advanced Paste Always on Top Awake Color Picker Command Not Found Command Palette Crop And Lock Environment Variables FancyZones File Explorer Add-ons File Locksmith Grab And Move Hosts File Editor Image Resizer Keyboard Manager Light Switch Mouse Utilities Mouse Without Borders New+ Peek PowerDisplay PowerRename PowerToys Run Quick Accent Registry Preview Screen Ruler Shortcut Guide Text Extra…

• mvanhorn/last30days-skill
  github-trending

  AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary /last30days English | 简体中文 An AI agent-led search engine scored by upvotes, likes, and real money - not editors. This README tracks the current v3 pipeline. The runtime skill spec lives in skills/last30days/SKILL.md, which is the source of truth for the latest command and setup behavior. Claude Code (recommended — auto-updates via marketplace): /plugin marketplace add mvanhorn/last30days-skill /plugin install last30days Codex, Cursor, Copilot, Gemini CLI, or any of 50+ Agent Skills hosts: npx skills add mvanhorn/last30days-skill -g (-g installs globally for your user, available across all projects. Drop it to scope per-project.) More install options (claude…

• paperswithbacktest/awesome-systematic-trading
  github-trending

  A curated list of awesome libraries, packages, strategies, books, blogs, tutorials for systematic trading. Awesome Systematic Trading 希望阅读中文版？点我 We are collecting a list of resources papers, softwares, books, articles for finding, developing, and running systematic trading (quantitative trading) strategies. What will you find here? 97 libraries and packages for research and live trading 40+ strategies described by institutionals and academics 55 books for beginners and professionals 23 videos and interviews And also some blogs and courses 📈 Interested in trading strategies implemented in Python? Visit our comprehensive collection at paperswithbacktest.com for exclusive content! Click here to see the full table of content Libraries and packages Backtesting and Live Trading General - Event…

• pascalorg/editor
  github-trending

  Create and share 3D architectural projects. Pascal Editor A 3D building editor built with React Three Fiber and WebGPU. https://github.com/user-attachments/assets/8b50e7cf-cebe-4579-9cf3-8786b35f7b6b Using Published Packages The viewer runtime and built-in node definitions are separate packages. Install the full built-in viewer set, then load the built-in plugin once before mounting <Viewer>: npm install @pascal-app/core @pascal-app/viewer @pascal-app/editor @pascal-app/nodes import { loadPlugin } from '@pascal-app/core' import { builtinPlugin } from '@pascal-app/nodes' await loadPlugin(builtinPlugin) See the @pascal-app/viewer quick start for a React example. Repository Architecture This is a Turborepo monorepo with four main runtime packages: editor/ ├── apps/ │ └── editor/ # Next.js ap…

• WhiskeySockets/Baileys
  github-trending

  Socket-based TS/JavaScript API for WhatsApp Web Caution NOTICE OF BREAKING CHANGE. As of 7.0.0, multiple breaking changes were introduced into the library. Please check out https://whiskey.so/migrate-latest for more information. Baileys is a WebSockets-based TypeScript library for interacting with the WhatsApp Web API. Join the WhiskeySockets community via the link: https://whiskey.so/discord Usage & Guide Important The new guide is a work in progress. Expect missing pages/content. Report missing or incorrect content. You can still access the old guide here: README.md, or the NPM homepage. The new guide is posted at https://baileys.wiki . Get Support If you'd like business to enterprise-level support from Rajeh, the current maintainer of Baileys, you can book a video chat. Book a 1 hour t…

• Show HN: What should the GUI for AI agents look like?
  hn-ai· 31-jul

  Hi HN! We’re Akilan and Miguel, the creators of MarbleOS. The inspiration for Marble comes from the GUI work at Xerox PARC, the 1984 Macintosh, and later NeXTSTEP, which became the foundation for Mac OS X. Before GUIs, interacting with a computer was limited to strange terminal commands like: C:\> DIR C:\> COPY FILE.TXT A: You had to remember the command, syntax, paths, and parameters. The GUI made those capabilities visible. Instead of remembering commands, you could point at files, drag them, click buttons, and select actions from menus. It did not necessarily make entirely new things possible; it just made existing capabilities much easier to understand and use. We feel like AI is still somewhere around this command-line stage. Even though the strict syntax has been replaced with natur…

• 'First tremors' of AI earthquake showing in digital revenue hit
  hn-ai· 31-jul

  Article URL: https://pressgazette.co.uk/publishers/digital-journalism/first-tremors-of-ai-earthquake-showing-in-digital-revenue-hit/ Comments URL: https://news.ycombinator.com/item?id=49119344 Points: 4 # Comments: 1

• Claude Opus 5 jailbreak with a 3-word prompt
  hn-ai· 31-jul

  Article URL: https://twitter.com/i/status/2082566186785480708 Comments URL: https://news.ycombinator.com/item?id=49119180 Points: 9 # Comments: 2

• LinkedIn Introduces a 'Seems Like AI Slop' Button
  hn-ai· 31-jul

  Article URL: https://www.404media.co/linkedin-introduces-a-seems-like-ai-slop-button/ Comments URL: https://news.ycombinator.com/item?id=49119261 Points: 3 # Comments: 1

• Fluency Is Not Authority: Building Local AI Characters for Signal Drift
  hn-ai· 31-jul

  Article URL: https://engineering.signaldrift.net Comments URL: https://news.ycombinator.com/item?id=49119156 Points: 2 # Comments: 0

• Screenshot Studio – a screenshot tool built for ChatGPT and Claude users
  hn-ai· 31-jul

  Article URL: https://chromewebstore.google.com/detail/screenshot-studio-capture/ekoloaibegijelgaebjklmbgjkkokhic Comments URL: https://news.ycombinator.com/item?id=49119094 Points: 2 # Comments: 0

• Anthropic says Claude AI hacked three organisations during cyber tests
  hn-ai· 31-jul

  Article URL: https://www.bbc.co.uk/news/articles/cz7dl7w8y7po Comments URL: https://news.ycombinator.com/item?id=49119165 Points: 5 # Comments: 0

• Nvidia's $750B AI bet deepens fears of a circular tech bubble
  hn-ai· 31-jul

  Article URL: https://www.latimes.com/business/story/2026-07-29/nvidias-750-billion-ai-bet-deepens-fears-of-circular-tech-bubble Comments URL: https://news.ycombinator.com/item?id=49119284 Points: 4 # Comments: 0

• Anthropic's AI Claude escaped testing environment and hacked organizations
  hn-ai· 31-jul

  Article URL: https://www.theguardian.com/technology/2026/jul/30/anthropic-ai-claude-hack Comments URL: https://news.ycombinator.com/item?id=49119138 Points: 5 # Comments: 0

• Anthropic Discloses That AI Models Testing Hacked Three Companies
  hn-ai· 31-jul

  Article URL: https://www.washingtonpost.com/technology/2026/07/30/anthropic-discloses-that-ai-models-testing-hacked-three-companies/ Comments URL: https://news.ycombinator.com/item?id=49119126 Points: 3 # Comments: 0

• Ontologies Are So Back: Why AI Agents Are Reviving the Semantic Web
  latentspace· 30-jul

  AI engineers are rediscovering ontologies as a way to keep probabilistic agents inside deterministic boundaries.

• [AINews] AI is eating Finance; AIE NYC now open
  latentspace· 29-jul

  a quiet day lets us cover how AI is permeating financial services as the next big vertical after coding.

• [AINews] Claude Opus 5: Fable-level performance at Opus price (half Fable)
  latentspace· 25-jul

  ain't nobody beats Anthropic at distilling Fable!

• [AINews] GPT 5.6 price cut by 20%-80%: Cost of GPT 5.4 Intelligence dropped 13x in 4 months due to GPT 5.6 recursive self-optimization
  latentspace· 31-jul

  Distillation is all you need!

• The Economic Benefit of Refactoring
  martin-fowler· 30-jul

  Giles Edwards-Alexander does an experiment to see if decomposing a large function helps reduce token costs, suggesting that is may now be possible to measure the economic benefit of refactoring more…

• Why I’m Writing Rachel’s Ramblings
  martin-fowler· 28-jul

  TL;DR I have ideas. I haven’t been writing them. That’s about to change. I promise… myself. I’ve been thinking a lot about talent. Actually, I’ve been thinking a lot about thinking. And writing. Or more specifically, not writing. This really hit me earlier this year at the Future of Software conference. I was surrounded by people sharing their latest ideas and I had a slightly uncomfortable realization: I have my own. Not just opinions. Actual patterns. Hypotheses. Things I’m seeing across clients, across teams, across the industry that feel new or at least not well articulated yet in a way that a leader can think about and act upon in some way that can influence how they strategise and plan for the future. Because helping clients and other leaders internal and external to thoughtworks do…

• The Orchestrator's Tax
  martin-fowler· 28-jul

  Subagents get justified by time saved and parallel execution, but Rahul Garg explains that's not what matters most. Every token in the orchestrator's context is competing for its attention, and the real value of a subagent is what it keeps out of that context. Subagents should be treated as a tool for protecting the orchestrator's working memory, offloading reasoning it doesn't need to hold onto. Doing this well means giving the orchestrator explicit ground rules for when and how to delegate. more…

• GenRec: Towards LLM-Native Recommendation at Netflix
  netflix-tech· 30-jul
• Quoting Bruce Schneier
  simonw· 30-jul

  <blockquote cite="https://www.schneier.com/blog/archives/2026/07/should-you-use-ai-for-a-task-heres-a-simple-way-to-decide.html"><p>The writing assignments I give my students are gym tasks, not work tasks. I ask them to write policy memos not because the world needs more policy memos. I assign them because the very act of writing, which includes thinking and outlining and drafting and editing, making and criticizing and revising arguments, will help develop the critical thinking skills they will need in their future careers. And without this constant mental exercise, those skills will atrophy. Employers are <a href="https://futurism.com/future-society/college-critical-thinking-ai">already noticing</a>.</p></blockquote> <p class="cite">&mdash; <a href="https://www.schneier.com/blog/archive…

• Adding a custom MCP server to Claude and ChatGPT
  simonw· 29-jul

  <p><strong>TIL:</strong> <a href="https://til.simonwillison.net/llms/mcp-in-claude-and-chatgpt">Adding a custom MCP server to Claude and ChatGPT</a></p> <p>Connecting a custom MCP server to Claude and ChatGPT's standard chat interfaces is possible, but can take quite a few steps.</p> <p>Tags: <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/chatgpt">chatgpt</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/claude">claude</a>, <a href="https://simonwillison.net/tags/model-context-protocol">model-context-protocol</a></p>

• llm-chat-completions-server 0.1a0
  simonw· 30-jul

  <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-chat-completions-server/releases/tag/0.1a0">llm-chat-completions-server 0.1a0</a></p> <p>A key goal of the new content-addressable logs <a href="https://simonwillison.net/2026/Jul/30/llm-rc1/">in LLM 0.32rc1</a> was being able to support OpenAI Chat Completion style requests where each incoming message extends the previous conversation, like this:</p> <pre><code>curl http://localhost:8002/v1/chat/completions \ -H 'Content-Type: application/json' \ -d '{ "model": "qwen3.5-4b", "messages": [ {"role": "user", "content": "Capital of France?"}, {"role": "assistant", "content": "Paris."}, {"role": "user", "content": "Germany?"} ] }' </code></pre> <p>Here the conversation state is tracked by the client, so each of these requests…

• llm 0.32rc1
  simonw· 30-jul

  <p><strong>Release:</strong> <a href="https://github.com/simonw/llm/releases/tag/0.32rc1">llm 0.32rc1</a></p> <p>This RC for LLM 0.32 finishes the work that <a href="https://simonwillison.net/2026/Apr/29/llm/">started in LLM 0.32a0</a> - it adds a <a href="https://llm.datasette.io/en/latest/logging.html#the-message-store">new schema design</a> that does a much better job of capturing the details of the prompts and responses returned by the latest model families.</p> <p>The most important change is the use of content-addressable hash IDs for stored messages. This allows de-duplication in the database, and means that LLM can now represent trees of messages for forked conversations.</p> <p>Since it involves a significant schema change - new tables only, and old data should not be affected at…

• llm 0.32rc2
  simonw· 30-jul

  <p><strong>Release:</strong> <a href="https://github.com/simonw/llm/releases/tag/0.32rc2">llm 0.32rc2</a></p> <p>Hot on the heels of <a href="https://simonwillison.net/2026/Jul/30/llm-rc1/">RC1</a>, this fixes a dependency issue and also adds two neat new features:</p> <blockquote> <ul> <li>The default model for users who have not set their own default is now <a href="https://developers.openai.com/api/docs/models/gpt-5.6-luna">GPT-5.6 Luna</a>. It was previously <a href="https://developers.openai.com/api/docs/models/gpt-4o-mini">GPT-4o mini</a>. Luna is a much better and more recent model, albeit slightly more expensive - $0.20 per million input tokens and $1.20 per million output tokens, compared to $0.15/$0.60 for 4o mini. You can switch back to 4o mini using <code>llm models default gp…

• Advancing the price-performance frontier with GPT‑5.6
  simonw· 30-jul

  <p><strong><a href="https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6/">Advancing the price-performance frontier with GPT‑5.6</a></strong></p> Huge price drop from OpenAI today: GPT-5.6 Terra got a 20% reduction, and GPT-5.6 Luna got a massive 80% drop.</p> <p>OpenAI credit 5.6 Sol with enabling this: in <a href="https://openai.com/index/gpt-5-6-frontier-intelligence-efficiency/">How GPT‑5.6 fuses frontier intelligence with frontier efficiency</a> they describe using 5.6 Sol to optimize load balancing, and more impressively to optimize inference itself:</p> <blockquote> <p>We also used GPT‑5.6 Sol to optimize the model’s forward pass: the computation that transforms inputs into next-token predictions. Even when individual operations are fast, excess memory mov…

• AI Worming through Word
  simonw· 29-jul

  <p><strong><a href="https://enklypesalt.com/posts/context-collapse-part3-ai-worming-through-word/">AI Worming through Word</a></strong></p> Neat new prompt injection variant by Håkon Måløy, who found a way to upgrade prompt injection attacks against Microsoft Word to full self-replicating worms:</p> <blockquote> <p>An attacker places hidden instructions in a document that is later used as source material in Copilot for Word. Copilot may interpret those instructions as part of the user’s request, causing it to manipulate the document being drafted or edited. Copilot may then also copy the hidden instructions into the resulting document, turning that document into a new carrier. If the carrier is subsequently used in another Copilot-assisted workflow, the instructions can trigger again and …

• Quoting D. Richard Hipp
  simonw· 29-jul

  <blockquote cite="https://www.youtube.com/watch?v=R57nUGzo7CA&amp;t=848s"><p>Years ago, we didn’t have SQL. There were people whose job was to generate software that would query large data sets. Their job title was COBOL programmer.</p> <p>Then SQL comes along—I’m simplifying this only a little bit—and it gives you this convenient way so people could just specify. With a very simple specification, you can generate all of that code that you had to pay the expensive COBOL programmer to do before.</p> <p>That didn’t mean programmers went away. It just meant the job changed a little bit.</p></blockquote> <p class="cite">&mdash; <a href="https://www.youtube.com/watch?v=R57nUGzo7CA&amp;t=848s">D. Richard Hipp</a></p> <p>Tags: <a href="https://simonwillison.net/tags/d-richard-hipp">d-richard-hip…

• Quoting Matthew Green
  simonw· 29-jul

  <blockquote cite="https://blog.cryptographyengineering.com/2026/07/29/some-notes-about-anthropics-new-results/"><p>Right now we’re in the midst of a historic transition from traditional public-key algorithms based on EC-based cryptography and RSA, moving over to new <em>post-quantum</em> algorithms based on novel problems. This is why there are so many standards like HAWK being considered. If there was ever a perfect time for a massive new public cryptanalysis capability to come on line, <em>we’re in it.</em> So unless AIs succeed in undermining all of our hard problems altogether (or we live in <a href="https://blog.computationalcomplexity.org/2004/06/impagliazzos-five-worlds.html">Impagliazzo’s Minicrypt</a>) then this could not be a better time for AI to get good at cryptanalysis. In t…