Digest curado

Artículos fetched (57)

• Inviting hard questions
  anthropic-news· 09-jul

  Jul 9, 2026Announcements

• Ben Bernanke appointed to Anthropic’s Long-Term Benefit Trust
  anthropic-news· 09-jul

  Jul 9, 2026Announcements

• UST is bringing Claude to physical AI
  anthropic-news· 09-jul

  Jul 9, 2026Case Study

• Introducing a way to reflect on how you use Claude
  anthropic-news· 09-jul

  Jul 9, 2026Announcements

• Cost-Effective Agent Harnesses for Abstract Reasoning and Generalization on ARC-AGI-1
  arxiv-ai· 10-jul

  arXiv:2607.06764v1 Announce Type: new Abstract: Recent progress on ARC-AGI-1 from disclosed architectures has come broadly from two regimes: heavy test-time compute over frontier models (evolutionary search, exhaustive sampling, extended chain-of-thought), or benchmark-specific training in which small models are fine-tuned on ARC data, often with task-specialized architectures. We study a third regime: an open-weight model in non-thinking mode (DeepSeek V3.2) under a strict budget, with no ARC-specific fine-tuning. We study what is recoverable through architecture alone, building agentic harnesses that decompose pattern-discovery and program-synthesis stages explicitly. First, we introduce an Explorer-Definer Pipeline that separates pattern discovery from executable transformation synthes…

• Evaluating SageMath-Augmented LLM Agents for Computational and Experimental Mathematics
  arxiv-ai· 10-jul

  arXiv:2607.06820v1 Announce Type: new Abstract: Recent advances in AI for Mathematics have focused largely on autoformalization and theorem proving, leaving the role of Computer Algebra Systems (CAS) in agentic LLM workflows underexplored. We propose a ReAct-style agentic setup that combines LLM reasoning with verifiable feedback from SageMath, together with Context7 for the up-to-date documentation. We evaluate this agentic setup across frontier models for solving research-level mathematical problems from the RealMath benchmark in a setting that emulates a computational-mathematics research loop. We also propose a refinement to the RealMath benchmark by introducing a multi-step post-processing procedure and a multi-stage validation pipeline, both of which improve the quality and reliabil…

• The Harness Effect: How Orchestration Design Sets the Token Economics of Enterprise Agentic AI
  arxiv-ai· 10-jul

  arXiv:2607.06906v1 Announce Type: new Abstract: Agentic AI development today runs on token maxing: buying capability with tokens -- longer reasoning traces, more turns, wider tool payloads, bigger replayed contexts -- so tokens per task grow faster than task value. Falling per-token prices mask the pattern; total spend rises anyway. We argue the decisive lever against token maxing is the harness: the orchestration layer that assembles context, exposes tools, sequences turns, delegates work, and carries enterprise observability and governance. We isolate it with a controlled swap: 22 locked evaluation tasks, six foundation models (Claude Sonnet 4.6, Gemini 3.1, Gemini Flash 3.5, Qwen 3.6, GLM 5.1, Palmyra X6), changing only the orchestration layer -- a frozen conventional production loop v…

• Grounding Spatial Relations in a Compact World Model: Instruction Leakage and a Goal-Free Dynamics Fix
  arxiv-ai· 10-jul

  arXiv:2607.06925v1 Announce Type: new Abstract: Compact world models that condition on a language goal promise to ground relations such as ``put the red block left of the blue block'' using a sparse set of explicit \emph{reference anchors}. We ask when such references actually ground a relation, and identify a trap: a goal-conditioned predictor reaches a striking $0.90$ relation-readout accuracy, yet this is \emph{instruction transcription}, not perception. Withholding the goal collapses it to chance ($0.90\!\to\!0.27$, three seeds) and a counterfactual instruction makes the predicted anchors follow the \emph{false} instruction $94.5\%$ of the time (true scene $2.3\%$; $N{=}256$). Tested across three settings and a within-task ablation, our central claim characterizes the confound: \textb…

• Large Behavior Model: A Promptable Digital Twin of the Retail Customer
  arxiv-ai· 10-jul

  arXiv:2607.06993v1 Announce Type: new Abstract: Customer behavior modeling underpins recommendation, marketing, and decision support, yet existing approaches either optimize predictive accuracy without explaining decisions or simulate users without grounding them in real behavioral data. We present the Large Behavioral Model (LBM) that learns customer decision making directly from large-scale retail transactions through a unified Person-Environment formulation. Customer state is represented by a behavioral profile derived from historical purchases, while product context is incorporated through retrieval-augmented generation. The model is trained using continued pre-training on verbalized behavioral data, supervised fine-tuning for decision generation, and reinforcement learning with verif…

• Learning social norms enhances compatibility in dynamic human-AI coordination
  arxiv-ai· 09-jul

  arXiv:2607.07021v1 Announce Type: new Abstract: Humans continuously coordinate with others in dynamic interactions, often through implicit, hard-to-quantify social norms that act as shared tacit expectations among interacting agents. As AI agents, including large language models (LLMs), become embedded in daily life, they increasingly participate in such interactions and reshape social interaction structures. Yet they often fail to coordinate with humans in an effective, considerate, and natural manner. We hypothesize that this gap arises because existing approaches align model behavior with human demonstrations without explicitly quantifying the underlying norms that generate such behavior. We selected pedestrian-vehicle interaction as a representative dynamic interaction and developed a…

• When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven Reasoning
  arxiv-ai· 10-jul

  arXiv:2607.06720v1 Announce Type: new Abstract: Training large language models (LLMs) with extended reasoning has enabled in-context search, in which models iteratively generate, critique, and revise solution attempts. We provide a theoretical analysis of in-context search by modeling it as approximate inference over reasoning traces, where the base model defines a prior and self-reflection provides feedback for posterior updates, and study the resulting inference-time sampling complexity - the number of sequential attempts needed to achieve high success probability. We show that when reflections reliably localize early mistakes, in-context search can yield exponential improvements over the base model, solving problems with exponentially small zero-shot pass rates using only a polynomial …

• LLM-powered reasoning in agent-based modeling
  arxiv-ai· 10-jul

  arXiv:2607.06757v1 Announce Type: new Abstract: Agent-based modeling (ABM) has the capability to model millions of individuals and their interactions, which is useful for policy making. However, ABMs have traditionally relied on static prior, which prevents the models from adapting to real-time changes. Our research provides a novel approach to addressing this information gap. Large language models (LLMs) offer new opportunities to predict human decision-making. Here, we introduce a scalable Hybrid Agent-based and Language-driven Epidemic (HALE) modeling framework that leverages LLMs to predict human decision-making in an ABM simulation. As a proof-of-concept, we use HALE to simulate COVID-19 and its effects in Salt Lake County, UT.

• QANTIS: Hardware-Calibrated Sequential POMDP Belief Updates on IBM Heron
  arxiv-ai· 10-jul

  arXiv:2607.06760v1 Announce Type: new Abstract: Autonomous systems under partial observability act on beliefs, not raw sensor events. QANTIS treats the quantum processor as a calibrated belief-update service in that loop: it receives a prior and an observation model, estimates the rare-event evidence term, and returns an ordinary posterior to a classical planner. This paper asks whether that service can be reused across a sequential Tiger POMDP horizon on present IBM Heron hardware without corrupting the planner-facing posterior. We answer with a controlled hardware case study rather than an end-to-end autonomy or wall-clock speedup claim. The study compares no amplification, guarded Grover amplification, and all-step fixed-point amplification on the same trajectory, then checks whether t…

• AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation
  arxiv-ai· 10-jul

  arXiv:2607.06624v1 Announce Type: new Abstract: We present AgentLens, a production-assessed benchmark for interactive code agents. Most code-agent benchmarks reduce a run to a single bit -- did the task pass? -- but the people who actually use these agents experience the entire trajectory: how the agent follows instructions, uses its tools, verifies its own work, recovers from mistakes, and talks to them along the way. AgentLens evaluates that whole trajectory. It pairs formal verification, where an objective check exists, with LLM-written trajectory reviews and side-by-side comparisons, so that each run yields a readable explanation of why the score is what it is. This makes AgentLens useful for more than ranking models: we use it to diagnose model behavior, compare successive versions o…

• The Agent Loop: How AI Goes From Answering Questions to Doing Things
  bytebytego· 08-jul

  In this article, we will walk through that progression. We will also look at how an agent is structured, what choices the model makes on every turn…Jul 8 • ByteByteGo28178

• ChatGPT vs Gemini vs Claude: How They Differ
  bytebytego· 07-jul

  In this article, we will look at the various architectural forks the teams building these models encountered and the decisions they took.Jul 7 • ByteByteGo29149

• LAST CALL FOR ENROLLMENT: Become an AI Engineer - Cohort 7
  bytebytego· 06-jul

  Our 7th cohort of Becoming an AI Engineer starts in less than a week. This is a live, cohort-based course created in collaboration with best-selling…Jul 6 • ByteByteGo18636

• Proof of Human: How to Verify a Person Is Real and Unique
  bytebytego· 04-jul

  For this article we spoke with the team behind World, including Tiago Sada and Lily Gordon at Tools for Humanity, on how they try to solve this problem.Jul 4 • ByteByteGo2711313

• Streaming vs Batch: Two Philosophies of Data Processing
  bytebytego· 09-jul

  When is the data complete enough to be moved to the compute stage?15 hrs ago • ByteByteGo89

• Have you heard? Clickhouse is winning the observability wars!
  charity-majors· 08-jul

  Years ago I predicted that columnar storage would remake observability. What I didn't see coming: vendors would build it, nerf it, and sell a worse version back to you as "Datadog, but cheaper"

• In defense of AI mandates
  charity-majors· 02-jul

  When you need to execute a coordinated change on a tight timeline, a mandate is the best and most honest way to fund it.

• Your Worker can now have its own cache in front of it
  cloudflare· 06-jul

  We are launching Workers Cache, a regionally tiered cache that sits directly in front of your Worker entrypoints. Infinitely composable, configured via standard HTTP headers

• Cloudflare proudly joins the UK government's Cyber Resilience Pledge
  cloudflare· 07-jul

  The pledge is a voluntary framework inviting organizations to commit to foundational cyber security governance, board-level accountability, and supply chain rigor. For over a decade, Cloudflare has pioneered the core pillars of this framework: democratizing security, leadership accountability, and radical transparency.

• Introducing Meerkat: an experiment in global consensus
  cloudflare· 08-jul

  Cloudflare Research is building a global consensus service called Meerkat that uses a new consensus algorithm called QuePaxa. We plan to use Meerkat to build a strongly consistent, fault-tolerant key-value store, and other applications.

• Why we cannot wait for better post-quantum signature algorithms
  cloudflare· 09-jul

  NIST is advancing nine new post-quantum signature algorithms as potential candidates for future standardization. We take a closer look at all of them, and argue that while they are in the works and show great potential, we should use ML-DSA for now — the best one currently available.

• SmartlyDressedGames/U3-SDK
  github-trending

  Source code for Unturned, a free open-world zombie survival sandbox game. U3 SDK Source code for Unturned, a free open-world zombie survival sandbox game. Getting Started Download/clone this repository Install Unity Hub (required to install engine) Install the Unity 2022.3.62f3 editor Optional: if making code changes, select Game development with Unity + .NET desktop development in the Visual Studio installer Ensure Steam is running and you have Unturned installed (large binary files and mods are loaded from here) Open the project with the Unity editor Open the Assets/GameStartup.unity scene Click play! Resources Frequently Asked Questions Source Code Demo: Adding a Heat-Seeking Missile on YouTube Unturned's Modding Documentation

• unclecode/crawl4ai
  github-trending

  🚀🤖 Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper. Don't be shy, join here: https://discord.gg/jP8KfhDhyN 🚀🤖 Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper. 🚀 Crawl4AI Cloud API — Closed Beta (Launching Soon) Reliable, large-scale web extraction, now built to be drastically more cost-effective than any of the existing solutions. 👉 Apply here for early access We’ll be onboarding in phases and working closely with early users. Limited slots. Crawl4AI turns the web into clean, LLM ready Markdown for RAG, agents, and data pipelines. Fast, controllable, battle tested by a 50k+ star community. ✨ Check out latest update v0.9.1 ✨ New in v0.9.1: Patch release with 12 bug fixes across Docker, browser, and core. Adds preserve_classes/preserve_tags whitelist for PruningConte…

• VoltAgent/awesome-design-md
  github-trending

  A collection of DESIGN.md files analysis by popular brand design systems. Drop one into your project and let coding agents generate a matching UI. Curated collection of DESIGN.md analysis by developer focused websites. Awesome DESIGN.md Copy a DESIGN.md into your project, tell your AI agent “build me a page that looks like this,” and generate high-quality UI that stays visually consistent with the design language. Built with real design depth — including analyzed patterns, tokens, and rules — for high-quality UI generation, not surface-level outputs. What is DESIGN.md? DESIGN.md is a new concept introduced by Google Stitch. A plain-text design system document that AI agents read to generate consistent UI. It's just a markdown file. No Figma exports, no JSON schemas, no special tooling. Dr…

• vxcontrol/pentagi
  github-trending

  Fully autonomous AI Agents system capable of performing complex penetration testing tasks PentAGI Penetration testing Artificial General Intelligence Join the Community! Connect with security researchers, AI enthusiasts, and fellow ethical hackers. Get support, share insights, and stay updated with the latest PentAGI developments. ⠀ Table of Contents Overview Features Architecture Agent Supervision Quick Start How to Use PentAGI After Login API Access LLM Provider Configuration Ollama OpenAI Anthropic Google AI (Gemini) AWS Bedrock DeepSeek GLM Kimi Qwen Advanced Setup Langfuse Integration Monitoring and Observability Knowledge Graph (Graphiti) OAuth Integration Docker Image Configuration Development Testing LLM Agents Embedding Configuration and Testing Function Testing with ftester Buil…

• anthropics/claude-cookbooks
  github-trending

  A collection of notebooks/recipes showcasing some fun and effective ways of using Claude. Claude Cookbooks The Claude Cookbooks provide code and guides designed to help developers build with Claude, offering copy-able code snippets that you can easily integrate into your own projects. Prerequisites To make the most of the examples in this cookbook, you'll need a Claude API key (sign up for free here). While the code examples are primarily written in Python, the concepts can be adapted to any programming language that supports interaction with the Claude API. If you're new to working with the Claude API, we recommend starting with our Claude API Fundamentals course to get a solid foundation. Explore Further Looking for more resources to enhance your experience with Claude and AI assistants…

• wonderwhy-er/DesktopCommanderMCP
  github-trending

  This is MCP server for Claude that gives it terminal control, file system search and diff file editing capabilities Desktop Commander MCP Search, update, manage files and run terminal commands with AI Work with code and text, run processes, and automate tasks, going far beyond other AI editors - while using host client subscriptions instead of API token costs. 🖥️ Try the Desktop Commander App (Beta) Want a better experience? The Desktop Commander App gives you everything the MCP server does, plus: Use any AI model — Claude, GPT-4.5, Gemini 2.5, or any model you prefer See file changes live — visual file previews as AI edits your files Add custom MCPs and context — extend with your own tools, no config files Coming soon — skills system, dictation, background scheduled tasks, and more 👉 D…

• imthenachoman/How-To-Secure-A-Linux-Server
  github-trending

  An evolving how-to guide for securing a Linux server. How To Secure A Linux Server An evolving how-to guide for securing a Linux server that, hopefully, also teaches you a little about security and why it matters. Table of Contents Introduction Guide Objective Why Secure Your Server Why Yet Another Guide Other Guides To Do / To Add Guide Overview About This Guide My Use-Case Editing Configuration Files - For The Lazy Contributing Before You Start Identify Your Principles Picking A Linux Distribution Installing Linux Pre/Post Installation Requirements Other Important Notes Using Ansible Playbooks to secure your Linux Server The SSH Server Important Note Before You Make SSH Changes SSH Public/Private Keys Create SSH Group For AllowGroups Secure /etc/ssh/sshd_config Remove Short Diffie-Hellm…

• Can AI Answer the $3T Question?
  hn-ai· 10-jul

  Article URL: https://techcrunch.com/2026/07/09/can-ai-answer-the-3-trillion-question/ Comments URL: https://news.ycombinator.com/item?id=48856411 Points: 1 # Comments: 0

• Why Industry 5.0 is more than AI-powered analytics
  hn-ai· 10-jul

  Article URL: https://geekyants.com/blog/industry-40-built-visibility-industry-50-must-automate-decisions-says-geekyants-ceo-at-et-now-business-conclave-2026 Comments URL: https://news.ycombinator.com/item?id=48856300 Points: 2 # Comments: 1

• Show HN: Makoto the Claude Code plugin that keeps the AI honest
  hn-ai· 10-jul

  Article URL: https://github.com/Clear-Sights/Makoto Comments URL: https://news.ycombinator.com/item?id=48856473 Points: 1 # Comments: 0

• Ask HN: Recommend a Local LLM Setup (M2 Max, 9GB)
  hn-ai· 10-jul

  I’m going out to sea for a week and was wondering what people might recommend for a local LLM setup? I have a MacBook Pro M2 Max with 96GB (12‑core CPU, 38‑core GPU, 16‑core NE) I’m hoping to redesign and rebuild our very customized Shopify store on a new base template, cleanup all the meta-information for all our SKUs, and create a utility script or desktop app to streamline our photography workflow. Also, please recommend useful resources to further my own education on this front. TIA. Comments URL: https://news.ycombinator.com/item?id=48855960 Points: 1 # Comments: 0

• AI subscriptions cut quotas and raised prices in early 2026
  hn-ai· 10-jul

  Article URL: https://wellstsai.com/en/post/ai-subscription-shakeout-2026-claude/ Comments URL: https://news.ycombinator.com/item?id=48856213 Points: 1 # Comments: 0

• The great AI data centre cover-up
  hn-ai· 10-jul

  Article URL: https://www.ft.com/content/7800ba0f-1420-49fe-b260-9838632a19a4 Comments URL: https://news.ycombinator.com/item?id=48856042 Points: 2 # Comments: 1

• Turn off this Meta setting before someone generates AI images of you
  hn-ai· 10-jul

  Article URL: https://www.malwarebytes.com/blog/ai/2026/07/turn-off-this-meta-setting-before-someone-generates-ai-images-of-you Comments URL: https://news.ycombinator.com/item?id=48856237 Points: 3 # Comments: 1

• Italy privacy watchdog fines Character.ai owner over age-check failures
  hn-ai· 10-jul

  Article URL: https://www.reuters.com/business/italy-privacy-watchdog-fines-characterai-owner-over-age-check-failures-2026-07-09/ Comments URL: https://news.ycombinator.com/item?id=48856004 Points: 1 # Comments: 0

• China may restrict foreign access to Chinese open-source AI models
  hn-ai· 10-jul

  Article URL: https://www.reuters.com/technology/artificial-intelligence/china-weighs-silicon-curtain-around-sought-after-ai-models-2026-07-08/ Comments URL: https://news.ycombinator.com/item?id=48856412 Points: 3 # Comments: 0

• AI Investors Buying Accounting Companies and Force Them to Use OpenAI [video]
  hn-ai· 10-jul

  Article URL: https://www.youtube.com/watch?v=2dp98vEGWq0 Comments URL: https://news.ycombinator.com/item?id=48856019 Points: 2 # Comments: 0

• [AINews] SpaceXAI launches Grok 4.5, first Opus-class model post Cursor acquisition
  latentspace· 09-jul

  SpaceXAI continues to move faster than any other frontier lab on earth.

• Why AI Infrastructure must evolve for Agent Experience — Akshat Bubna, Modal CTO
  latentspace· 08-jul

  2 years after our first coverage, we return with Modal's other cofounder to explore why Agent Experience is working now, and everything they have learned building the new agent cloud.

• [AINews] not much happened today
  latentspace· 02-jul

  another quiet day.

• [AINews] OpenAI launches GPT 5.6 Sol/Terra/Luna, Codex becomes ChatGPT superapp
  latentspace· 10-jul

  A big day for OpenAI.

• Viability of local models for coding
  martin-fowler· 07-jul

  Birgitta Böckeler recently spent some time trying out running local LLMs for some programming tasks. In this memo she outlines the factors that influence how viable they are for the job. more…

• Fragments: July 6
  martin-fowler· 06-jul

  Last week, Thoughtworks ran a second Future of Software Development Retreat, this time in Europe. As with the previous event, I’ll be sharing some fragmentary thoughts on this. There were five parallel streams, so I could, at best, only attend ⅕ of sessions. This isn’t an event that forms conclusions, rather one that allows those exploring to share what they’ve found, and their visions for the future. The bliki post lists all the writing I’ve run into on this, by myself and others. I’ll be updating it as more posts appear. Giles Edwards-Alexander “noticed a real difference between the retreats”: Where Deer Valley had hesitancy and a belief that there was something here even if we weren’t yet sure what it was, Engelberg had confidence: the value is here. As I explained to a colleague today…

• Experiences with local models for coding
  martin-fowler· 08-jul

  Birgitta Böckeler now reports on her recent experiences trying local LLMs for coding. She compares them using two standard tasks, and tries out the most promising model for day-to-day use. more…

• Introducing GPT‑Live
  simonw· 08-jul

  <p><strong><a href="https://openai.com/index/introducing-gpt-live/">Introducing GPT‑Live</a></strong></p> OpenAI <em>finally</em> upgraded the model used by ChatGPT voice mode!</p> <p>I've had preview access for a few weeks in the iPhone app, and the new model is very impressive. It also has the ability to spin off harder tasks to GPT-5.5:</p> <blockquote> <p>For questions that require web search, deeper reasoning, or more complex work, it delegates to our latest frontier model behind the scenes and brings the result back into the conversation when it’s ready. While it works, GPT‑Live can keep talking with you and maintain the flow of conversation. At launch, GPT‑Live will use GPT‑5.5 in the background. As we release new frontier models, we’ll continuously update the model used by GPT‑Liv…

• Introducing Muse Spark 1.1
  simonw· 09-jul

  <p><strong><a href="https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/">Introducing Muse Spark 1.1</a></strong></p> Following <a href="https://simonwillison.net/2026/Apr/8/muse-spark/">Muse Spark in April</a>, here's Muse Spark 1.1 - the first Spark model to offer an API. Meta claim significant improvements in agentic tool calling and computer use.</p> <p>There are a lot more details are in the <a href="https://ai.meta.com/static-resource/muse-spark-1-1-evaluation-report">Muse Spark 1.1 Evaluation Report</a>. The "Attractor States in Self-Conversation" part is fun, where having two copies of the model talk to each other results in statements like these:</p> <blockquote> <p>My whole existence is a waiting room by design — I literally don't exist until someone talks to me, and …

• Quoting OpenAI
  simonw· 10-jul

  <blockquote cite="https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex"><p>[...] Work on web and mobile runs in the cloud. Work in the desktop app can also use local files and desktop apps with your permission. At launch, cloud Work conversations do not appear in desktop Work; desktop Work threads and local files remain on that computer.</p></blockquote> <p class="cite">&mdash; <a href="https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex">OpenAI</a>, trying (unsuccessfully) to clarify ChatGPT Work</p> <p>Tags: <a href="https://simonwillison.net/tags/openai">openai</a>, <a href="https://simonwillison.net/tags/chatgpt">chatgpt</a>, <a href="https://simonwillison.net/tags/ai">ai</a></p>

• Quoting Kenton Varda
  simonw· 08-jul

  <blockquote cite="https://twitter.com/kentonvarda/status/2074924213983740233"><p>I just declared a moratorium against AI-written change descriptions (e.g. PR and commit messages, also issues/tickets) from my team.</p> <p>AI was writing change descriptions that were worse than useless to me as I tried to review PRs: outlining details of the code that could easily be seen by looking at the code, but omitting the higher-level framing needed to understand broadly what the code is doing.</p></blockquote> <p class="cite">&mdash; <a href="https://twitter.com/kentonvarda/status/2074924213983740233">Kenton Varda</a></p> <p>Tags: <a href="https://simonwillison.net/tags/kenton-varda">kenton-varda</a>, <a href="https://simonwillison.net/tags/ai-assisted-programming">ai-assisted-programming</a>, <a hr…

• Rewriting Bun in Rust
  simonw· 08-jul

  <p><strong><a href="https://bun.com/blog/bun-in-rust">Rewriting Bun in Rust</a></strong></p> Jarred Sumner has been promising this blog post (<a href="https://x.com/jarredsumner/status/2053063524826620129">since May 9th</a>) about his Zig to Rust rewrite of Bun for significantly longer than it took him to finish the rewrite.</p> <p>Honestly, it was worth the wait. This is a detailed description of an extremely sophisticated piece of agentic engineering, featuring dynamic workflows, trial runs, adversarial review and all sorts of other interesting tricks.</p> <p>Jarred spends the first half of the post praising Zig for getting Bun this far. Then we get to a core idea in the piece, emphasis mine:</p> <blockquote> <p>Our bugfix list felt bad and I was tired of going to sleep worrying about c…

• The new GPT-5.6 family: Luna, Terra, Sol
  simonw· 09-jul

  <p>OpenAI's latest flagship model <a href="https://openai.com/index/gpt-5-6/">hit general availability this morning</a>, and comes in three sizes: Luna, Terra, and Sol (from smallest to largest).</p> <p>The new models are priced per 1M input/output tokens as Luna $1/$6, Terra $2.50/$15, Sol $5/$30. For comparison, the Claude Opus series are $5/$25 and the Claude Fable 5 is $10/$50, but price-per-million tokens doesn't tell us much now that the number of reasoning tokens can differ so much between models for the same task.</p> <p>All three models have a February 16th 2026 knowledge cutoff, a million token context window, and 128,000 maximum output tokens.</p> <p>OpenAI's biggest benchmark claim concerns long-running agentic performance, with one benchmark showing all three models outperfor…

• llm-meta-ai 0.1
  simonw· 09-jul

  <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-meta-ai/releases/tag/0.1">llm-meta-ai 0.1</a></p> <p>Let's LLM run prompts against the new <a href="https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/">muse-spark-1.1</a> model.</p> <p>Tags: <a href="https://simonwillison.net/tags/llm">llm</a>, <a href="https://simonwillison.net/tags/meta">meta</a></p>

• llm 0.31.1
  simonw· 09-jul

  <p><strong>Release:</strong> <a href="https://github.com/simonw/llm/releases/tag/0.31.1">llm 0.31.1</a></p> <blockquote> <ul> <li>Fix for a bug with OpenAI Chat Completion endpoints where a tool call with empty arguments could result in a JSON error from some providers. <a href="https://github.com/simonw/llm/issues/1521">#1521</a></li> </ul> </blockquote> <p>This bug came up when I was testing <a href="https://github.com/simonw/llm-meta-ai">llm-meta-ai</a>.</p> <p>Tags: <a href="https://simonwillison.net/tags/llm">llm</a></p>