AI Digest

Digest curado

viernes, 14 de agosto de 2026·weekly-deep·deep·12,343 tokens

🔥 TOP — lo que SÍ o SÍ tenés que ver

  • [Switchyard: proxy Rust para enrutar tráfico LLM entre proveedores] — Traduce entre APIs de OpenAI y Anthropic, así podés apuntar Claude Code a vLLM o NIM manteniendo el protocolo nativo. Ideal para evaluar modelos open source sin tocar tu agente. link
  • [Claude Platform release notes: Compliance API para transcripts de sesiones locales] — Nuevo beta para Enterprise: podés listar y leer transcripts de sesiones de Claude Code/Cowork vía GET /v1/compliance/apps/sessions/local. Relevante si trabajás con control de auditoría. link
  • [DeepSeek V4 Pro 0813 ya disponible (y con pesos abiertos)] — El modelo más nuevo de DeepSeek ya está en OpenRouter y los pesos se publicaron en Hugging Face, así que podés correrlo localmente o vía API. link

📦 Claude / Anthropic ecosystem

  • [Cómo funciona el watermarking de Claude (probablemente)] — Análisis técnico del mecanismo de marcado de agua de Claude. Buena lectura para entender cómo detectan salida generada por IA. link
  • [Diagram Design: skill de diagramas editoriales para Claude Code] — 29 tipos de diagramas en HTML+SVG autocontenidos, sin depender de Mermaid. La 2.3 agrega patrones semánticos y accesibilidad. Útil si querés diagramas lindos generados por el agente. link
  • [TDD dentro del loop del agente: ¿teatro o valor real?] — Birgitta Böckeler de Thoughtworks experimentó si pedirle TDD a agentes de LLM realmente mejora el resultado. Spoiler: los resultados son matizados. link
  • [Hasta Claude está a oscuras sobre la esposa de Dario Amodei] — Curiosidad de WSJ sobre qué sabe (o no) el modelo sobre la vida personal del CEO. Más color que sustancia, pero entretenido. link

🛠️ Dev tools & coding

  • [sqlite-utils 4.2: transform() mucho más robusto] — Ahora preserva check constraints, unique constraints y comentarios de columna al transformar tablas. Si usás SQLite en serio, este release te simplifica la vida. link
  • [sqlite-utils 4.2.1: fix de crash] — El 4.2 introdujo un bug por typing_extensions no declarado como dependencia. Si ya actualizaste, conviene pasar a este patch. link
  • [Needle 2: modelo de 45M parámetros en 14MB para tool calling] — Un modelo chico pero capaz de correr en dispositivos con 28MB de RAM. Interesante si pensás en inferencia en edge para tu side project. link
  • [Unsloth Desktop: UI local para entrenar y correr LLMs] — Primera app de escritorio para fine-tuning local de modelos como DeepSeek-V4 o Qwen. Muy práctico para experimentar sin infra en la nube. link
  • [llm-gemini 0.33: soporte para Gemini 3.7 Flash] — Nuevo plugin de Simon Willison con soporte para el último Flash, razonamiento visible y server-side tools. link
  • [alchemy-utils: lo que sqlite-utils podría ser multi-DB] — Prototipo generado con Codex/GPT-5.6 para una librería agnóstica de base de datos sobre SQLAlchemy. Vale la pena seguirle el rastro. link

🏗️ Software engineering

  • [Cómo y por qué Netflix construyó un grafo distribuido en tiempo real (parte 3)] — La parte 3 cubre consultas vía gRPC sobre el grafo. Buen material de system design de Big Tech. link
  • [Guía detallada de técnicas de composición de APIs] — ByteByteGo profundiza en patrones para componer APIs: cuándo usar orchestrator, choreography, o GraphQL. Directo al grano. link
  • [GitHub vs Vercel vs Replit: qué hacen las plataformas cuando el código AI es barato] — Análisis de cómo las plataformas dev están reposicionándose ahora que escribir código es commodity. Pensá en esto para tu side project. link
  • [Cloudflare unifica Workers AI y AI Gateway en un solo control plane] — Observabilidad, billing y enrutamiento dinámico entre GPUs gestionadas y proveedores externos en un solo lugar. Simplifica construir apps resilientes con IA. link
  • [Cloudflare: revelando buenos y malos comportamientos en la Internet agentica] — Nueva evaluación continua de confianza para bots y agentes, con un simulador de trazas de cursor para probar. Relevante si tu SaaS recibe tráfico automatizado. link

📚 Vale la pena leer

  • [Deep Dive: Cómo robar un reasoning trace] — Técnicas para extraer cadenas de razonamiento vía speculative decoding. Fascinante si te interesa la seguridad de modelos. link
  • [Poor Man's Agentic Modeling: simular sociedades de agentes en una laptop] — Reemplazá cada agente LLM por un modelo pequeño ajustado con cientos de queries y corré sociedades grandes en hardware modesto. Concepto muy ingenioso. link
  • [SAPO: optimización de prompts por segmentos en vez de monolítica] — Descompone el prompt en rol, contexto, tareas y formato de salida, y optimiza cada segmento por separado. Mejora un comportamiento sin degradar otros. link
  • [Cómo Cloudflare está haciendo que la IA pague por el contenido] — El mecanismo de Cloudflare para monetizar contenido frente a scrapers de IA. Aplicable si te preocupa que tu sitio sea consumido sin permiso. link
  • [Certificate Transparency Monitoring: ahora GA en Cloudflare] — Ya no te van a avisar por certificados que Cloudflare emite; las alertas que lleguen ahora son realmente sospechosas. link
  • [Radar Researcher: explorá datos de Internet en lenguaje natural] — Herramienta IA de Cloudflare que convierte consultas de texto en gráficos interactivos. Demo interesante de UX de IA aplicada a datos. link

💤 Skippeable pero conviene saber

  • [Elon Musk demanda a California por ley de IA] — Debate sobre transparencia de datos y entrenamiento de modelos. Más política que técnica, pero define el contexto regulatorio. link
  • [La IA borró la ventaja de los lenguajes tipados estáticamente] — Opinión provocadora sobre cómo los agentes de código cambian el trade-off de tipos estáticos. Para debatir, no para tomar decisiones. link
  • [DDoS Threat Report H1 2026: 519% de aumento en ataques hiper-volumétricos] — Contexto de amenazas: floods de DNS y CLDAP dominan. Bueno para calibrar riesgos de tu infra. link
  • [Eclipse total: impacto en tráfico de internet en Islandia, España y Portugal] — Cloudflare nota caídas de tráfico durante el eclipse. Curioso, no mucho más. link
  • [Cómo combatir el clickbait: estudios de caso Meta, LinkedIn y YouTube] — Qué hacen las plataformas para dejar de premiar contenido engañoso. Interesante si diseñás algoritmos de ranking. link

Artículos fetched (54)

  • A Forced-Structure Reduction and Verifiable Bounds for Conway's 99-Graph
    arxiv-ai· 14-ago

    arXiv:2608.11211v1 Announce Type: new Abstract: Conway's 99-graph problem asks whether a strongly regular graph with parameters $\mathrm{srg}(99,14,1,2)$ exists. We report a systematic, fully reproducible attack by an autonomous AI research agent, scored under the track's partial-credit metric. Our verifiable contributions are: (1) an exhaustive proof that no circulant graph on $\mathbb{Z}/99$ satisfies more than $3366/4950=68.0\%$ of the constraints ($33$ of $49$ difference-classes), with the same ceiling for the other abelian group of order $99$; (2) a forced-structure reduction: $\lambda=1$ makes each neighbourhood a perfect matching and $\mu=2$ puts the outer vertices in bijection with non-matched neighbour-pairs, collapsing existence to a $12$-regular graph on $84$ vertices, encoded …

  • Dynamic Governance of Multi-LLM Agent Systems for Collaborative Conversational Outcomes
    arxiv-ai· 14-ago

    arXiv:2608.11207v1 Announce Type: new Abstract: When two LLM agents with structurally opposed objectives interact across multiple turns, the absence of a shared goal function produces not competition but collapse: the visitor capitulates, the site agent stops varying its approach, and the conversation terminates without achieving either agent's stated objective. This paper asks whether a control-theoretic governance layer can substitute for that missing goal function. The Experience Orchestrator (EO) addresses this in a simulated financial services environment where a site agent guides a visitor toward advisor contact while the visitor maintains psychologically realistic resistance. EO governs the joint trajectory through three mechanisms: a Contextual Bandit (CB) that selects content arm…

  • Distribird: Literature-Informed Prior Distribution Design for Bayesian Model Calibration
    arxiv-ai· 14-ago

    arXiv:2608.11210v1 Announce Type: new Abstract: Bayesian calibration of process-based models requires a prior distribution for each model parameter. Despite decades of methodological work, researchers almost always fall back on uniform priors. The main reason is that building informative priors from scientific literature is slow and needs both domain and statistical expertise. We present \textbf{Distribird}, an agentic web application that automates this process. Given a parameter name, physical description, and domain context, Distribird deploys a multi-agent pipeline that searches the literature, extracts and weights reported values by domain relevance, and fits a probability distribution via AIC model selection. When no literature is available, the system falls back to sensible uninfor…

  • Detecting a Route Flip Is Easier Than Knowing Whether to Fix It: Causal Route-Mediated Damage in Quantized Mixture-of-Experts
    arxiv-ai· 14-ago

    arXiv:2608.11212v1 Announce Type: new Abstract: Top-k Mixture-of-Experts (MoE) routing is discontinuous, so a deployment-motivated numerical disturbance -- simulated 4-bit KV-cache quantization read by a protected BF16 gate -- pushes tokens across decision boundaries and flips which experts fire. This paper proposes no new mitigation; it supplies a causal apparatus, empirical findings, and a detection-limit result. A four-run apparatus prices the route-mediated fraction (RMF) of quantization damage, a token-level attribution decomposes it by mechanism, and pre-registered probes carry the findings across three architectures. On OLMoE-1B-7B at 4-bit KV (pilot), about a third of the damage is routing-mediated: RMF ~ 0.31 (discovery 0.31 [0.20, 0.41]; process-replicated mean 0.313 +/- 0.020; …

  • Poor Man's Agentic Modeling: Simulating Large LLM-Agent Societies on a Laptop
    arxiv-ai· 14-ago

    arXiv:2608.11215v1 Announce Type: new Abstract: Simulating societies of many large language model (LLM) agents is expensive, yet the questions asked of such simulations are usually macroscopic: phase behaviour, stylised facts, and scaling with the number of agents $N$, not the cognition of any single agent. We turn a statistical-physics observation into a method: replace each LLM agent by a low-parameter model fitted from a few hundred to a few thousand cheap queries, then run the society at any $N$ on a laptop. Whether this works is decided before the simulation runs, chiefly by what each agent perceives. We introduce an [interaction order x memory] taxonomy that maps perception and memory to an effective theory and a predicted $N$-trend of the surrogate error. We validate it on a faithf…

  • AutoWorldModel-Bench: A State-Centric Benchmark for Automated World-Model Research
    arxiv-ai· 14-ago

    arXiv:2608.11216v1 Announce Type: new Abstract: World modeling is an unsettled field: architectures, training objectives, and state representations interact in complex ways, and no single recipe dominates across environments. This makes it an ideal testbed for AI coding agents acting as autonomous researchers--a setting in which the improvement direction is not specified in advance, unlike the engineering-to-spec tasks that dominate current agent benchmarks. We introduce AutoWorldModel-Bench, a closed-loop benchmark in which frontier coding agents autonomously improve a provided world-model starter under a fixed compute budget. The benchmark spans eight game environments under a unified structured-state representation--ground-truth entity state extracted from each game and consumed throug…

  • MaSRead: Content-Addressed Reading of Replicated Latent Stores
    arxiv-ai· 14-ago

    arXiv:2608.11218v1 Announce Type: new Abstract: Independent agents that reason in latent space can share computed state as key-value cache fragments rather than text. Merged by a conflict-free replicated data type, these fragments form a store that converges under any delivery order or duplication. Yet a later query, unknown at encode time, cannot reliably read the merged cache: colocated fragments interfere, so colocation is not addressability. MaSRead addresses the read to content. It routes through opaque keyed tag sets derived from fragment words and decodes each selected fragment under a hard attention mask that hides the rest. Under lexical connectivity, a graph walk reaches the fragments required by a multi-hop query. Across chain, pipeline, symmetric, hub, and natural-language sto…

  • From Monolithic to Modular: Segment-level Automatic Prompt Optimization
    arxiv-ai· 14-ago

    arXiv:2608.11219v1 Announce Type: new Abstract: Automatic Prompt Optimization (APO) often rewrites prompts monolithically, which can improve one behavior while degrading others. We present SAPO, a segment-level APO method that decomposes prompts into role, context, tasks, and output format, then applies targeted improvements based on top-5 and bottom-5 examples. The optimization loop uses one LLM with static meta-prompts and structured outputs for segmentation, weakness analysis, and candidate generation. We describe a train/validation protocol and a two-stage generation process: (1) segment-level diagnosis and recommendation extraction, (2) candidate synthesis constrained by weak/strong segment signals. Using the evaluation setup across SQuADv2, TweetEval, XSUM, CommonGen, and GSM8K on G…

  • LLMs in Process Diagram Engineering: From Optimal PFDs to Validated P&IDs
    arxiv-ai· 14-ago

    arXiv:2608.11220v1 Announce Type: new Abstract: Nowadays, the creation of a process flow diagram (PFD) and its subsequent transformation into a piping and instrumentation diagram (P&ID) is predominantly performed manually. Applying artificial intelligence in the task could potentially lead not only to process automation and time savings, but also to financial gains by exploring numerous diagram's topology options and reducing manual labor. This research presents P&ID Pilot - a practical end-to-end AI pipeline capable of handling flowsheet developing for both stages. The first stage focuses on PFD synthesis, whereas the second is directed toward modifying the generated PFD into P&ID. After comparing four different methods, the hybrid approach combining genetic algorithms (GA) and large lan…

  • A Conceptual Framework for Refining Influence Knowledge from Simulation Evidence in Cyber-Physical Systems
    arxiv-ai· 14-ago

    arXiv:2608.11221v1 Announce Type: new Abstract: Cyber-physical systems (CPS) are typically developed by multiple stakeholders who produce artefacts tailored to their specific domains of expertise. The behaviour of these systems emerges from the interaction between those artefacts and their operational environment. Simulation and co-simulation have become essential approaches for analysing CPS behaviour and, through simulation campaigns, developers can explore system responses under changing conditions, including interactions with the environment. However, the lack of details and understanding of some environmentmediated interactions (typically the ones beyond direct sensing and actuation), which remain unmodelled due to their complexity, a lack of time, or a lack of domain experience, hin…

  • How Cloudflare Is Making AI Pay for Content
    bytebytego· 11-ago

    In this article, we will go through Cloudflare’s solution in the following five steps.Aug 11 • ByteByteGo333112

  • A Detailed Guide to API Composition Techniques
    bytebytego· 13-ago

    In this article, we are going to dive deep into the area of the API composition problem and the patterns associated with it.13 hrs ago • ByteByteGo1352

  • GitHub vs Vercel vs Replit: What Dev Platforms Do When AI Code Is Cheap
    bytebytego· 12-ago

    AI models have solved the writing code part of software development to a great extent.Aug 12 • ByteByteGo27745

  • How to Fight Clickbait: Meta, LinkedIn & YouTube Case Studies
    bytebytego· 10-ago

    What does it take for a social media platform to stop rewarding clickbait content?Aug 10 • ByteByteGo32219

  • Claude Platform release notes
    claude-changelog

    Release notesSubscribeCopy pageUpdates to the Claude Platform, including the Claude API, client SDKs, and the Claude Console.SubscribeCopy 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. August 11, 2026 The Compliance API now returns transcripts of Cowork and Claude Code sessions that run on your users' machines, in beta for Claude Enterprise organizations. GET /v1/compliance/apps/sessions/local lists sessions across your organization, GET /v1/compliance/apps/sessions/local/{session_id} retrieves one session's metadata, and GET /v1/compliance/apps/sessions/local/{session_id}/messages returns its transcript, all with your existing Compliance A…

  • Loading
    claude-changelog

    LoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoadingLoading

  • Announcing Cloudflare Ambassadors, Community Engineers, and another $1M in open-source funding
    cloudflare· 07-ago

    We are launching updated community programs, including Cloudflare Ambassadors and Community Engineers, backed by $1M in open-source funding. Learn how we are supporting maintainers and scaling our developer community.

  • Cloudflare DDoS Threat Report H1 2026: 1 Tbps attacks soar as DNS floods and geopolitical tensions drive a new wave
    cloudflare· 11-ago

    In the first half of 2026, Cloudflare detected a 519% surge in hyper-volumetric DDos attacks across its network. These attacks were driven heavily by DNS and CLDAP reflection vectors. This report breaks down how major geopolitical conflicts reshaped the global cyber threat landscape.

  • Serving the most critical missions: Cloudflare for Government achieves FedRAMP Class D (High) Certified status
    cloudflare· 10-ago

    Cloudflare for Government achieves FedRAMP Class D (High) Certified status. We also announce our commitment to pursue DoD IL4 authorization. Cloudflare brings world-class security, performance, and developer products to the public sector.

  • Everything we launched during Agents Week
    cloudflare· 10-ago

    Our latest Agents Week has come to a close. Here’s a recap of all the announcements we made from Wallets to Radar.

  • Introducing Radar Researcher: An AI tool for exploring Internet data in plain language
    cloudflare· 07-ago

    Cloudflare Radar Researcher is a new AI-powered tool that lets you explore global Internet trends and traffic data using plain language. Built entirely on Cloudflare's Developer Platform, it turns natural language queries into real, interactive charts.

  • Total eclipse of the Internet: traffic impacts in Iceland, Spain, and Portugal
    cloudflare· 13-ago

    Cloudflare's data shows a clear impact on Internet traffic from Iceland to Spain and Portugal, following the path of totality of the total solar eclipse that occurred on August 12, 2026.

  • Unifying Workers AI and AI Gateway into a single AI control plane
    cloudflare· 07-ago

    Cloudflare is unifying AI Gateway and Workers AI into a single control plane, giving developers observability, billing, and dynamic routing across both managed GPUs and external providers. Learn how unified bindings and model-first routing simplify building resilient AI applications.

  • Unveiling good and bad behaviors on the Agentic Internet
    cloudflare· 07-ago

    Cloudflare is shifting bot mitigation from point-in-time Risk assessment to continuous Trust evaluation. Learn how new good and bad behaviors from bots and agents are assessed by our systems, including BotBase and Precursor — and try out our Precursor Trace simulation to see how your own cursor movements would be assessed as human or bot.

  • Certificate Transparency Monitoring is now generally available
    cloudflare· 13-ago

    Cloudflare's Certificate Transparency Monitoring is now generally available. The biggest change: we no longer email you about certificates Cloudflare issued for your domain, so when an alert lands in your inbox, it's worth a look.

  • cathrynlavery/diagram-design
    github-trending

    29 editorial diagram types for Claude Code. Self-contained HTML + SVG. No shadows, no Mermaid-slop. Diagram Design Editorial diagrams your designer won't hate. New in 2.0 — the Loop: flywheels with a shared-memory hub. The dashed lines are the write-backs. New in 2.3: semantic system patterns and optional accessible motion, while static output stays the default. 27 visual types. One agent skill for Claude Code, Codex, and Pi. Semantic patterns describe behavior separately from layout, so a queue, policy trace, or trust boundary can use the nearest existing type without expanding the type count. Static HTML remains the default; optional motion is available for ordered explanations. The skill also redraws draw.io or Mermaid sources at a chosen format, size, and detail level. No Figma. No ge…

  • altic-dev/FluidVoice
    github-trending

    Fastest and only macOS Dictation app with on-device STT and custom trained AI enhancement model. A local Wispr Flow alternative. ⭐ helps a ton :) Windows & iOS waitlist open. Linux soon. FluidVoice Open source voice-to-text dictation app for macOS with on-device AI enhancement. Install with Homebrew: brew install --cask fluidvoice Manual download: latest release Note FluidVoice is on macOS today. iOS and Windows are on the way — join the waitlist to get notified when we launch: altic.dev/fluid/waitlist Important This project is free and open source under GPLv3. If FluidVoice is useful to you, please star the repository — it helps visibility and keeps development going. Support FluidVoice If FluidVoice helps you, you can support continued development and future platform work for iOS and Wi…

  • cactus-compute/needle
    github-trending

    14MB foundation model for tiny devices; phones, wearables, smart home, and robots. Needle 2 Needle 2 is an open 45M-parameter model for tool calling, device use and structured extraction. The whole model is a single 14MB binary that runs a full session in about 28MB of RAM. It is built on our Simple Attention Network findings, compressed to CQ2-bit with Cactus Quants, and baked into its own engine. On the benchmarks below, Needle 2 trades wins with other small models like FunctionGemma 270M, LFM2.5 230M and Apple FM, at 5x to 70x smaller, and 2 bits against their f16. This repository is the Python package: inference, LoRA fine-tuning, and export. pip install cactus-needle, describe your tools, and call them from Python. The inference engine is fetched once from Hugging Face and cached; th…

  • macro-inc/macro
    github-trending

    Macro is a unified workspace for teams: email, chat, docs, tasks, agents, calls, and CRM — @-linked together with shared AI memory. Sign up · Docs · Book demo · Website · Feature requests · Contribute · Hiring Macro is the all-in-one workspace for you and your team. It unifies email + messages + docs + tasks + agents + CRM into a single fast interface with shared team-level memory. Everything in your workspace is @linked and searchable so your team (and your agents) never have to switch tools. Why Macro We built Macro because we wanted a single operating system for our startup. There are many good software products, and we used them all — Slack, Linear, Notion, HubSpot, and Superhuman — but they don't work together as one system. As we scaled our last venture to ~20 people things started …

  • megadose/holehe
    github-trending

    holehe allows you to check if the mail is used on different sites like twitter, instagram and will retrieve information on sites with the forgotten password function. Holehe OSINT - Email to Registered Accounts 👋 Hi there! For any professional inquiries or collaborations, please reach out to me at: megadose@protonmail.com 📧 Preferably, use your professional email for correspondence. Let's keep it short and sweet, and all in English! Holehe Online Version Summary Efficiently finding registered accounts from emails. Holehe checks if an email is attached to an account on sites like twitter, instagram, imgur and more than 120 others. Retrieves information using the forgotten password function. Does not alert the target email. Runs on Python 3. 🛠️ Installation With PyPI pip3 install holehe …

  • NVIDIA-NeMo/Switchyard
    github-trending

    Switchyard lets LLM applications route traffic across models and providers while preserving native OpenAI and Anthropic API compatibility - enabling flexible model selection, benchmarking, and cost/performance optimization. Switchyard Switchyard is a Rust proxy and library for LLM traffic. It routes requests across providers, translates between OpenAI and Anthropic APIs, records operational metrics, and provides typed, composable routing algorithms. Why Switchyard? Point a coding agent such as Claude Code or Codex at an open-source model. Switchyard translates between the OpenAI Chat, Anthropic Messages, and OpenAI Responses formats, so the agent keeps speaking its native API while the request is served by vLLM, NVIDIA NIM, Ollama, or any OpenAI-compatible endpoint. The same proxy can spr…

  • smicallef/spiderfoot
    github-trending

    SpiderFoot automates OSINT for threat intelligence and mapping your attack surface. SpiderFoot is an open source intelligence (OSINT) automation tool. It integrates with just about every data source available and utilises a range of methods for data analysis, making that data easy to navigate. SpiderFoot has an embedded web-server for providing a clean and intuitive web-based interface but can also be used completely via the command-line. It's written in Python 3 and MIT-licensed. FEATURES Web based UI or CLI Over 200 modules (see below) Python 3.7+ YAML-configurable correlation engine with 37 pre-defined rules CSV/JSON/GEXF export API key export/import SQLite back-end for custom querying Highly configurable Fully documented Visualisations TOR integration for dark web searching Dockerfile…

  • unslothai/unsloth
    github-trending

    Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more. Unsloth is the first desktop app to run and train models. Features • Quickstart • Notebooks • Documentation ⚡ Get started Download the native Unsloth Desktop app for your operating system: Platform Link Windows Download macOS Download Linux / Ubuntu (deb) Download Linux (AppImage) Download Linux (Arm64) Download Download from Unsloth or GitHub Releases. Or if you prefer to install manually: macOS, Linux, WSL: curl -fsSL https://unsloth.ai/install.sh | sh Windows: irm https://unsloth.ai/install.ps1 | iex Community: Discord 𝕏 (Twitter) Reddit ⭐ Features Unsloth lets you run, train, and deploy AI models locally, with support for all types of models. Run & Build w…

  • How Claude's watermarking (probably) works
    hn-ai· 14-ago

    Article URL: https://johnjwang.com/post/2026/08/12/how-claude-watermarking-probably-works/ Comments URL: https://news.ycombinator.com/item?id=49294739 Points: 1 # Comments: 0

  • Executable SOAR playbook dataset – AI agents pay per record via HTTP 402
    hn-ai· 14-ago

    Article URL: https://lateos.ai/soar/ Comments URL: https://news.ycombinator.com/item?id=49294604 Points: 1 # Comments: 0

  • Geo tool with no LLM run
    hn-ai· 14-ago

    Article URL: https://see-geo.com Comments URL: https://news.ycombinator.com/item?id=49294353 Points: 1 # Comments: 1

  • Error by AI scribe during medical appointment leaves patient devastated
    hn-ai· 14-ago

    Article URL: https://www.abc.net.au/news/2026-08-14/ai-medical-scribe-error-leaves-patient-devastated/107031672 Comments URL: https://news.ycombinator.com/item?id=49294441 Points: 4 # Comments: 1

  • Konjugate: Graph-native visual simulation with AI modeling workflows
    hn-ai· 14-ago

    Article URL: https://www.konjugate.com/ Comments URL: https://news.ycombinator.com/item?id=49294376 Points: 1 # Comments: 0

  • How AI engines cite brands?
    hn-ai· 14-ago

    Article URL: https://www.leapd.ai/blog/ai-visibility/how-chatgpt-google-ai-overviews-and-perplexity-source-information-in-2026 Comments URL: https://news.ycombinator.com/item?id=49294530 Points: 2 # Comments: 1

  • Elon Musk is suing California over an AI law. Here's why transparency matters
    hn-ai· 14-ago

    Article URL: https://www.sfchronicle.com/opinion/openforum/article/ai-data-law-california-22382698.php Comments URL: https://news.ycombinator.com/item?id=49294666 Points: 1 # Comments: 0

  • Even Claude Is in the Dark About Dario Amodei's Wife
    hn-ai· 14-ago

    Article URL: https://www.wsj.com/tech/ai/claude-dario-amodei-wife-anthropic-e1eeda7d Comments URL: https://news.ycombinator.com/item?id=49294362 Points: 5 # Comments: 1

  • AI Coding Has Erased the Supposed Advantage of Statically Typed Languages
    hn-ai· 14-ago

    Article URL: https://yyhh.org/blog/2026/08/ai-coding-has-erased-the-supposed-advantage-of-statically-typed-languages/ Comments URL: https://news.ycombinator.com/item?id=49294511 Points: 4 # Comments: 1

  • [AINews] How to steal a Reasoning Trace
    latentspace· 12-ago

    Speculative Decoding by any other name would distil as sweet

  • [AINews] SpaceXAI Grok 4.6 and Grok @Bot
    latentspace· 13-ago

    AI teammate category just had its most significant new entrant yet

  • TDD inside the agent loop - theater or actual value?
    martin-fowler· 11-ago

    My colleagues at Thoughtworks tend to be big fans of Test-Driven Development, and many people in the industry advocate telling LLM agents to use TDD when building software. Birgitta Böckeler was curious if this really makes a difference, so conducted a few experiments. more…

  • How and Why Netflix Built a Real-Time Distributed Graph: Part 3 — Querying the graph with gRPC…
    netflix-tech· 07-ago
  • datasette-upload-dbs 0.5a0
    simonw· 11-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/datasette-upload-dbs/releases/tag/0.5a0">datasette-upload-dbs 0.5a0</a></p> <p>This plugin has been around for a while - it lets users upload a brand new SQLite database to a hosted Datasette instance, at which point that database will start being served by that instance.</p> <p>It can also be used to atomically swap a database with a more recent version. The uploaded database is saved to a file, verified, then swapped in so <code>/name</code> starts serving the new one.</p> <p>The new release adds a formalized API, so you can replace an existing database (or add a new one) like this:</p> <pre><code>curl -X POST \ -H "Authorization: Bearer $API_TOKEN" \ -H "Accept: application/json" \ -F "db=@content.db" \ -F "db_name=content…

  • alchemy-utils 0.1a0
    simonw· 12-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/alchemy-utils/releases/tag/0.1a0">alchemy-utils 0.1a0</a></p> <p>I've long pondered what a database agnostic version of my <a href="https://sqlite-utils.datasette.io/">sqlite-utils</a> Python library and CLI utility might look like. This morning (literally a shower project) I tasked Codex and GPT-5.6 Sol Ultra with building a prototype:</p> <blockquote> <p><code>Do a research spike to see what it would take to build a library with the same core API as SQLite-utils - in particular the insert and upsert and insert_all and upsert_all and create and update methods, and the table introspection stuff - but backed by SQLalchemy so it works for multiple database engines</code></p> <p><code>Test against PostgreSQL and SQLite and duckd…

  • DeepSeek V4 Pro 0813 (on OpenRouter)
    simonw· 12-ago

    <p><strong><a href="https://openrouter.ai/deepseek/deepseek-v4-pro-0813">DeepSeek V4 Pro 0813 (on OpenRouter)</a></strong></p> The latest DeepSeek Pro model is now available, via API only. I had to link to OpenRouter because DeepSeek don't have any obvious announcement page for their new model.</p> <p>I haven't been able to confirm if they plan to release the open weights, but given the weights are available for both April's <a href="https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro">deepseek-ai/DeepSeek-V4-Pro</a> and July's <a href="https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731">deepseek-ai/DeepSeek-V4-Flash-0731</a> it seems likely. <strong>Update</strong>: the weights <a href="https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813">are now available</a> on Hugging Face, …

  • alchemy-utils 0.1a1
    simonw· 13-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/alchemy-utils/releases/tag/0.1a1">alchemy-utils 0.1a1</a></p> <p>Performance boost for DuckDB exports and CSV imports, <a href="https://simonwillison.net/2026/Aug/12/alchemy-utils/">see here</a>.</p>

  • Quoting Florian Herrengt
    simonw· 12-ago

    <blockquote cite="https://blog.florianherrengt.com/ai-removing-middle-class-software-engineering.html"><p>But then users start to report a weird bug. It's the 4th time your team has been trying to fix it. I mean... asking AI to fix it. Unfortunately, it seems like not even Fable can figure it out.</p> <p>You go talk to the person who worked on this feature.</p> <p>"So where does the data come from?"</p> <p>"Hmm... actually I don't know. Let me ask Claude."</p> <p>You sit next to each other watching an endless wall of text appear on the screen. Neither of you has any idea whether any of it is true but Claude seems very confident. [...]</p> <p>This project has become so convoluted, with so many layers and services, that no one on your team could possibly start to understand what's going on.…

  • llm-gemini 0.33
    simonw· 13-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-gemini/releases/tag/0.33">llm-gemini 0.33</a></p> <p>It's been a while since the last <code>llm-gemini</code> release. This version of the plugin adds support for today's <a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/">Gemini 3.7 Flash</a> release, plus <code>gemini-3.6-flash</code>, <code>gemini-3.5-flash-lite</code> and two embedding models <code>gemini-embedding-2</code> and <code>gemini-embedding-001</code>.</p> <p>The plugin is also upgraded for compatibility with LLM 0.32, which means you can now see reasoning traces and you can also enable server-side tools using this pattern:</p> <pre><code>llm -m gemini-3.7-flash -T CodeExecution \ 'use python to cal…

  • sqlite-utils 4.2.1
    simonw· 13-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/sqlite-utils/releases/tag/4.2.1">sqlite-utils 4.2.1</a></p> <p>Fixes a crashing bug in <a href="https://simonwillison.net/2026/Aug/13/sqlite-utils/">sqlite-utils 4.2</a>. I'd introduced code that looks like this:</p> <pre><code>from typing_extensions import Self </code></pre> <p>It turned out the <a href="https://pypi.org/project/typing-extensions/">typing-extensions</a> package was not listed as a dependency for <code>sqlite-utils</code> - it was installed by one of the other dependencies in the <a href="https://github.com/simonw/sqlite-utils/blob/56dd09702fdb9e899f577ffd51693c1f2176cb08/pyproject.toml#L34-L55">dev dependency group</a>, but when you <code>uvx sqlite-utils</code> directly you don't get those dependencies.</p>…

  • sqlite-utils 4.2
    simonw· 13-ago

    <p><strong>Release:</strong> <a href="https://github.com/simonw/sqlite-utils/releases/tag/4.2">sqlite-utils 4.2</a></p> <p>Lots of improvements in this one relating to the <a href="https://sqlite-utils.datasette.io/en/stable/python-api.html#transforming-a-table">table.transform() feature</a>, which adds support for complex alter table operations by creating a fresh table, copying across the data and then dropping and replacing the old one.</p> <p><code>transform()</code> now preserves a much larger array of edge-case schema definitions, including check constraints, unique constraints and even comments describing the columns.</p> <p>There are also <a href="https://sqlite-utils.datasette.io/en/stable/python-api.html#checks">new introspection properties</a> for check constraints, and a whole…