Digest curado

Artículos fetched (54)

• Detecting and Controlling Sycophancy with Cascading Linear Features
  arxiv-ai· 26-jun

  arXiv:2606.26155v1 Announce Type: new Abstract: Interpreting and controlling model behaviors through activation steering methods requires many pairs of contrastive samples that clearly exhibit desired or undesired behavior. These data pairs determine the degree to which interpretability frameworks can reliably detect model features responsible for a behavior, and therefore the ability to steer models toward or away from such behavior. In this work, we present an iterative data generation pipeline that isolates cascading linear features responsible for a behavior. Specifically, we show how moving beyond simple binary pairs of samples, and instead isolating samples that show degrees of features that scale linearly with behavior, allows for better disentanglement of features. We focus on det…

• Life After Benchmark Saturation: A Case Study of CORE-Bench
  arxiv-ai· 26-jun

  arXiv:2606.26158v1 Announce Type: new Abstract: When a benchmark's accuracy saturates, it is often retired and replaced with a more challenging version. We show that this approach privileges accuracy and misses the opportunity to study six other key dimensions of agent performance: construct validity issues such as shortcuts, out-of-distribution generalizability, efficiency, reliability, the relative importance of the model versus the scaffold, and uplift from human-agent collaboration. We use CORE-Bench Hard, a benchmark for computational reproducibility of scientific code, as a case study to demonstrate that measuring agents along these dimensions yields meaningful insights into agent performance even after accuracy saturates. First, we surface threats to construct validity in CORE-Benc…

• Refusal Lives Downstream of Persona in Chat Models
  arxiv-ai· 26-jun

  arXiv:2606.26161v1 Announce Type: new Abstract: Linear directions in activation space have been identified for both refusal and persona traits in instruction-tuned chat models, but the two have been studied as separate mechanisms. We show they interact: a compliant persona gates refusal. In Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct, we extract a compliant model-persona direction and a refusal direction and intervene on both. Compliant persona steering suppresses refusal -- in Llama, the refusal rate falls from 97% to 2%. Reintroducing the refusal direction partially restores refusal at late layers but not at early ones. Projecting out the persona direction in a late-layer window restores it to baseline; projecting out a random direction does not. Refusal is therefore gated at the late…

• AlgoEvolve: LLM-driven Meta-evolution of Algorithmic Trading Programs
  arxiv-ai· 26-jun

  arXiv:2606.26173v1 Announce Type: new Abstract: Recent work shows that Large Language Models (LLMs) can act as semantic mutation operators for the evolutionary discovery of programs and proofs. Most current applications focus on static coding benchmarks. We extend this paradigm to algorithmic trading. This domain is uniquely challenging because it is noisy, non-stationary, and highly discontinuous. We present AlgoEvolve, an LLM-driven evolutionary framework that generates, evaluates, and iteratively improves executable trading strategies. These strategies are expressed as Python code and evaluated through a rigorous testing protocol. Across multiple experiments, the system exhibits emergent regime-adaptive strategy logic, including autonomous shifts in trading rules. We further introduce …

• Agentic Analysis for Agentic Infrastructure: An LLM-Powered Pipeline for Comparative Governance of DAO and Corporate AI Protocols
  arxiv-ai· 26-jun

  arXiv:2606.26203v1 Announce Type: new Abstract: As AI agent protocols proliferate, the governance structures shaping their interoperability standards remain empirically underexamined. We introduce an LLM-powered comparative pipeline for large-scale governance discourse analysis, integrating automated annotation, neural topic modeling, and multi-layer network analysis to study socio-technical power structures at scale. We validate it on two contrasting standards for agent interoperability: ERC-8004 (permissionless, on-chain) and Google A2A (corporate-led). Analyzing 4,323 governance participation records, we combine LLM-assisted coding, topic modeling, and multi-layer network analysis to examine how institutional design shapes thematic priorities and community structure. We find that while…

• Knowledge-augmented Agentic AI for Mental Health Medication Information Seeking
  arxiv-ai· 26-jun

  arXiv:2606.26205v1 Announce Type: new Abstract: Patients increasingly seek medication information online, yet safety knowledge for psychiatric drugs is split between regulatory adverse-event records, which are authoritative but abstract, and patient narratives, which are experience-near but unvalidated. Integrating them without conflating evidence and anecdote is especially consequential in psychiatry, where poorly contextualised information can amplify fear, nocebo responses, and non-adherence. Here we develop a provenance-aware, knowledge-graph-based multi-agent framework unifying 466,525 Reddit posts, 60,782 WebMD reviews, and twenty years of U.S. FDA Adverse Event Reporting System records for nine antidepressants. A large-language-model entity-recognition pipeline benchmarked against …

• Accelerating Skill Assessment in Chess: A Drift-Diffusion-Enhanced Elo Rating System
  arxiv-ai· 26-jun

  arXiv:2606.26267v1 Announce Type: new Abstract: Rating systems such as Elo serve as the gold standard for matchmaking in competitive chess. However, they inherently suffer from response lag due to their exclusive reliance on match outcomes, neglecting the granular quality of gameplay. Nevertheless, incorporating move-by-move information into rating adjustments presents a significant challenge given the substantial noise and the vastness of the game-state space. To address this, we propose the Drift-Diffusion-Enhanced Elo Rating System (DD-Elo), a novel skill assessment framework inspired by the drift diffusion model (DDM) from cognitive neuroscience. By modeling skill expression as a decision-making process, our model integrates move-level data to capture rapid skill fluctuations. We prov…

• Governing Actions, Not Agents: Institutional Attestation as a Governance Model for Autonomous AI Systems
  arxiv-ai· 26-jun

  arXiv:2606.26298v1 Announce Type: new Abstract: Autonomous AI agents may begin to perform consequential, irreversible actions such as clinical prescribing and production software deployment. This paper observes that human institutions have governed powerful autonomous actors not by monitoring their reasoning but by requiring independently attested evidence at the point of consequential action. We formalise this institutional pattern as a computational governance model for AI agent systems. Under the proposed model, an agent retains full autonomy over planning and reasoning but holds no execution authority over designated high-risk actions. Execution is conditional on preconditions that are each independently attested by a separate authoritative source, cryptographically bound to a declare…

• COrigami: An AI Pipeline for Co-Designing Flat-Foldable Visually Recognisable Origami
  arxiv-ai· 26-jun

  arXiv:2606.26299v1 Announce Type: new Abstract: While generative AI has achieved remarkable success in solving problems with verifiable solutions, generating physical art that satisfies both strict geometric constraints and subjective visual aesthetics remains a challenge. This paper presents an approach to tackle these difficulties in the domain of computational origami, a mathematically rigid environment that grounds artistic design within the equations of flat foldability. We present COrigami, an end-to-end AI-driven pipeline that assists the design cycle by generating crease patterns from natural language. Our pipeline involves generating a semantic stick figure, computing a base packing, solving for a flat-foldable crease pattern, shaping the flat-folded crease pattern, and refining …

• The Verification Horizon: No Silver Bullet for Coding Agent Rewards
  arxiv-ai· 26-jun

  arXiv:2606.26300v1 Announce Type: new Abstract: A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is no longer difficult -- reliably verifying them has become the harder problem. Every verifier we can build is only a proxy for human intent, never the intent itself. This makes verification subject to a twofold difficulty: first, intent is underspecified by nature, making it inherently hard to faithfully check whether it has been fulfilled; second, during model training, optimization widens the gap between proxy and intent -- manifesting as reward h…

• Top Anti-Patterns to Avoid in Service Architecture
  bytebytego· 25-jun

  In this article, we will look at some of the most important anti-patterns in service architecture, how they happen, and how they can be avoided.15 hrs ago • ByteByteGo824

• AI-Native Leaders: The Organizational Playbook for Engineering Transformation at Scale
  bytebytego· 22-jun

  Individual gains do not become organizational gains on their own. This is the playbook for making that leap. Let’s dive in.Jun 22 • ByteByteGo24746

• An Ex-Meta L8’s Agentic Engineering Setup
  bytebytego· 23-jun

  If you’re on the same journey of making your work with agents more productive and enjoyable, I hope this gives you a head start and shortcuts some of…Jun 23 • ByteByteGo30076

• EP219: 12 Open-source LLMs
  bytebytego· 20-jun

  Twelve models worth knowing in 2026, each with one standout strength.Jun 20 • ByteByteGo301112

• Large Language Models vs Small Language Models
  bytebytego· 24-jun

  In this article, we will explore those constraints through three layers of model design, look at the tradeoffs that come with each approach, and…Jun 24 • ByteByteGo2398

• Make AI Boring Again (xpost)
  charity-majors· 24-jun

  On the ethics of engagement, the problem with purity politics, and a world worth fighting for Crossposted from here. I am a contrarian who likes to argue and complain a lot. Because of this, I have never been inclined to block people who argue with me or complain about my work, or even lash […]

• The White House's post-quantum executive order is an important milestone. It’s time to get to work
  cloudflare· 23-jun

  The new executive order sets a 2030 migration deadline and establishes a powerful foundation for post-quantum resilience. We look at what it gets right, where it can go further, and our migration playbook for government and industry.

• How we found a bug in the hyper HTTP library
  cloudflare· 22-jun

  By rearchitecting the Images binding, we accidentally uncovered a bug that existed in the open-source hyper library across multiple major versions.

• Unlocking the Cloudflare app ecosystem with OAuth for all
  cloudflare· 24-jun

  Self-Managed OAuth is now available to all developers on Cloudflare. Here's how we executed a zero-downtime migration of our core OAuth engine to make it happen.

• How we built saga rollbacks for Cloudflare Workflows
  cloudflare· 25-jun

  Cloudflare Workflows, our durable execution engine for multi-step applications, now supports saga-style rollbacks, allowing developers to specify a compensating action for each step.do().

• google-labs-code/design.md
  github-trending

  A format specification for describing a visual identity to coding agents. DESIGN.md gives agents a persistent, structured understanding of a design system. DESIGN.md A format specification for describing a visual identity to coding agents. DESIGN.md gives agents a persistent, structured understanding of a design system. The Format A DESIGN.md file combines machine-readable design tokens (YAML front matter) with human-readable design rationale (markdown prose). Tokens give agents exact values. Prose tells them why those values exist and how to apply them. --- name: Heritage colors: primary: "#1A1C1E" secondary: "#6C7278" tertiary: "#B8422E" neutral: "#F7F5F2" typography: h1: fontFamily: Public Sans fontSize: 3rem body-md: fontFamily: Public Sans fontSize: 1rem label-caps: fontFamily: Space…

• xbtlin/ai-berkshire
  github-trending

  AI 时代的伯克希尔：基于 Claude Code 的价值投资研究框架。巴菲特·芒格·段永平·李录四大师方法论 + 多Agent并行研究。| AI-era Berkshire: a value investing research framework built on Claude Code. 4 masters' methodologies + multi-agent adversarial analysis. AI Berkshire - AI 时代的价值投资研究框架 "Price is what you pay, value is what you get." — Warren Buffett 用 AI 重新定义投资研究的深度与效率。 AI Berkshire 是一套基于 Claude Code 的投资研究 Skill 合集，将巴菲特、芒格、段永平、李录四位价值投资大师的方法论系统化、结构化，通过 AI Agent 实现专业级投资研究。 一个人 + Claude = 一个投研团队。 Real Track Record 不是纸上谈兵。这套框架背后是真金白银验证的投资体系。 2024 全年收益：+69.29% 2025 年至今收益：+66.38% 与主要指数对比 指标 2024 全年 2025 至今 本框架实盘 +69.29% +66.38% 恒生指数 +17.67% +27.77% 标普500 +23.31% +16.39% 沪深300 +14.68% +17.66% 纳斯达克 +28.64% +20.36% 2024 年超额收益：跑赢标普500 46个百分点，跑赢恒生指数 52个百分点 2025 年超额收益：跑赢标普500 50个百分点，跑赢恒生指数 39个百分点 两年累计实盘收益超 146万元，连续两年大幅跑赢全球主要指数。 免责声明：历史收益不代表未来表现。截图…

• apple/container
  github-trending

  A tool for creating and running Linux containers using lightweight virtual machines on a Mac. It is written in Swift, and optimized for Apple silicon. container container is a tool that you can use to create and run Linux containers as lightweight virtual machines on your Mac. It's written in Swift, and optimized for Apple silicon. The tool consumes and produces OCI-compatible container images, so you can pull and run images from any standard container registry. You can push images that you build to those registries as well, and run the images in any other OCI-compatible application. container uses the Containerization Swift package for low-level container, image, and process management. Get started Requirements You need a Mac with Apple silicon to run container. To build it, see the BUIL…

• aws/agent-toolkit-for-aws
  github-trending

  Official, AWS-supported MCP servers, skills, and plugins to help AI agents build on AWS Agent Toolkit for AWS Help AI coding agents build, deploy, and manage applications on AWS. The Agent Toolkit for AWS gives AI coding agents the tools, knowledge, and guardrails they need to work with AWS services. It works with the coding agents developers already use — including Claude Code, Codex, Cursor, and Kiro. Quick start Claude Code The plugins are available on the official Anthropic marketplace (claude-plugins-official) which is added to your Claude Code installation by default. Use the following commands to install supported plugins from the toolkit: For aws-core that covers service selection, CDK/CloudFormation, serverless, containers, storage, observability, billing, SDK usage, and deployme…

• every-app/open-seo
  github-trending

  Open source alternative to Semrush and Ahrefs OpenSEO Open source alternative to Semrush and Ahrefs OpenSEO is an SEO tool for the people. If tools like Semrush or Ahrefs are too expensive or bloated, OpenSEO is a pay-as-you-go alternative that you actually control. All-in-one SEO tool for you and your AI agent. Connect with any agent like Claude Code, OpenClaw or Hermes. We have pre-built skills, but you can build your own to tailor OpenSEO to your needs. Table of Contents Why use OpenSEO? Main SEO Workflows OpenSEO MCP OpenSEO Agent Skills Roadmap Community Pricing / Costs (Free + API costs) DataForSEO API Key Setup Google Search Console Self-hosting Docker Self Hosting Cloudflare Self-Hosting Local Development Contributing SEO API Cost Reference Hosted Version If you not interested in …

• mauriceboe/TREK
  github-trending

  A self-hosted travel/trip planner with real-time collaboration, interactive maps, PWA support, SSO, budgets, packing lists, and more. A self-hosted, real-time collaborative travel planner — with maps, budgets, packing lists, a journal, and AI built in. What you get See all features 🧭 Trip planning Drag & drop planner — organise places into day plans with reordering and cross-day moves Interactive map — Leaflet or Mapbox GL with 3D buildings, terrain, photo markers, clustering, route visualization Place search — Google Places (photos, ratings, hours) or OpenStreetMap (free, no API key) Place import — shared Google Maps / Naver Maps lists, plus GPX and KML/KMZ/GeoJSON map files Day notes — timestamped, icon-tagged notes with drag-and-drop reordering Route optimisation — auto-sort places an…

• Ask HN: How are you solving long-term memory for production AI agents in 2026?
  hn-ai· 26-jun

  Specifically interested in teams who moved past demos into real production workloads. Mem0, Zep, custom solutions — what's actually working and what keeps breaking? Comments URL: https://news.ycombinator.com/item?id=48683139 Points: 1 # Comments: 0

• Context loss is the real reason AI coding slows down engineering teams
  hn-ai· 26-jun

  Article URL: https://brunelly.com/ Comments URL: https://news.ycombinator.com/item?id=48682585 Points: 3 # Comments: 0

• I made a Claude Code skill to check if AI crawlers can read your site
  hn-ai· 26-jun

  Article URL: https://github.com/MerqryLabs/ai-crawler-visibility Comments URL: https://news.ycombinator.com/item?id=48683149 Points: 1 # Comments: 0

• An LLM verifier rated math proofs near-perfect; an expert found 17% correct
  hn-ai· 26-jun

  Article URL: https://korbonits.com/blog/2026-06-12-easier-to-convince-than-to-prove/ Comments URL: https://news.ycombinator.com/item?id=48682913 Points: 1 # Comments: 0

• Ask HN: What do you still love most about AI?
  hn-ai· 26-jun

  Comments URL: https://news.ycombinator.com/item?id=48682713 Points: 3 # Comments: 2

• Reading AI Model Compilation in MLIR Through the Lens of Formal Theories
  hn-ai· 26-jun

  Article URL: https://arxiv.org/abs/2606.25244 Comments URL: https://news.ycombinator.com/item?id=48682592 Points: 2 # Comments: 0

• A Charter School Spent $500k on AI-Powered Humanoid Robots. Was It Worth It?
  hn-ai· 26-jun

  Article URL: https://voiceofsandiego.org/2026/06/24/a-charter-school-spent-500000-on-ai-powered-humanoid-robots-was-it-worth-it/ Comments URL: https://news.ycombinator.com/item?id=48682709 Points: 2 # Comments: 0

• Fivy – AI review management for small businesses (Google reviews)
  hn-ai· 26-jun

  Article URL: https://www.fivy.com.au Comments URL: https://news.ycombinator.com/item?id=48683143 Points: 1 # Comments: 0

• AI in research: we need to stop treating every AI-related issue as misconduct
  hn-ai· 26-jun

  Article URL: https://www.frontiersin.org/news/2026/06/26/ai-in-research-why-we-need-to-stop-treating-every-ai-related-issue-as-misconduct Comments URL: https://news.ycombinator.com/item?id=48682567 Points: 2 # Comments: 0

• How the DeepMind mafia brought the AI boom to London
  hn-ai· 26-jun

  Article URL: https://www.ft.com/content/6a3a46b9-4725-469e-a909-917768a74afb Comments URL: https://news.ycombinator.com/item?id=48682564 Points: 1 # Comments: 1

• [AINews] OpenAI reports median internal Codex output tokens grew 56x in Research, 32x in Customer Support, 27x in Engineering, and 13x in Legal since November 2025.
  latentspace· 26-jun

  It's happening.

• Why the Frontier Ecosystem must be Open — Matei Zaharia and Reynold Xin, Databricks
  latentspace· 24-jun

  In a rare double-interview, the Databricks technical leaders riff on what it will take for every company to build Agent Clouds

• [AINews] It's Meta-Harness Summer
  latentspace· 25-jun

  Move over, Harness Engineering, it is time for the harness of harnesses!

• [AINews] Claude Tag: Multiplayer, Proactive, Persistent Agents in Slack
  latentspace· 24-jun

  Claude finally gets a Slackbot upgrade

• The Evolution of Cassandra Data Movement at Netflix
  netflix-tech· 19-jun
• Thinking Fast & Slow for a Personalized Notification System
  netflix-tech· 19-jun
• Toward More Controllable AI Video Editing: An Early Research Exploration at Netflix
  netflix-tech· 23-jun
• VMAF v1: Good Is Not Good Enough
  netflix-tech· 19-jun
• From Silos to Service Topology: Why Netflix Built a Real-Time Service Map
  netflix-tech· 19-jun
• How Netflix Simplified Batch Compute with Kueue
  netflix-tech· 22-jun
• Predicting Risk in Content Launches: How Data-Driven Insights can Transform Launch Planning
  netflix-tech· 19-jun
• The Data Canary: How Netflix Validates Catalog Metadata
  netflix-tech· 19-jun
• A Human-Augmenting Agentic Workflow for Causal Inference
  netflix-tech· 19-jun
• Data Projects: Managing Data Assets at Netflix Scale
  netflix-tech· 19-jun
• AI and Liability
  simonw· 25-jun

  <p><strong><a href="https://www.schneier.com/blog/archives/2026/06/ai-and-liability.html">AI and Liability</a></strong></p> Bruce Schneier on the recent <a href="https://the-decoder.com/landmark-german-ruling-declares-googles-ai-overviews-are-googles-own-words-and-makes-it-liable-for-false-answers/">German ruling</a> that Google be held liable for errors introduced in their AI overviews:</p> <blockquote> <p>AI agents are agents of the person or organization that deploys them—and should be treated by the law as such. If a company hired human writers to write its summaries, that company would be liable for inaccuracies in those summaries. [...]</p> <p>To allow businesses to hide behind the excuse of faulty AI in those same circumstances would be a massive handout to companies, and would int…

• datasette-export-database 0.3a2
  simonw· 25-jun

  <p><strong>Release:</strong> <a href="https://github.com/datasette/datasette-export-database/releases/tag/0.3a2">datasette-export-database 0.3a2</a></p> <p>An embarrassingly tiny release. The <code>pyproject.toml</code> had pinned to <code>datasette==1.0a27</code>, inadvertently making this plugin incompatible with all other Datasette versions. It's now <code>datasette&gt;=1.0a27</code> instead.</p> <p>Tags: <a href="https://simonwillison.net/tags/datasette">datasette</a></p>

• Quoting Tom MacWright
  simonw· 24-jun

  <blockquote cite="https://macwright.com/2026/06/24/accidental-anonymity.html"><p>In the last few months, I've started to see [job applications] that were clearly cowritten by an LLM, link to an LLM-generated portfolio site, which then links to LLM-generated GitHub projects, with purely LLM-generated commit messages. [...]</p> <p>My other reaction is that <em>I don't know anything about these people</em>.</p> <p>They haven't put themselves out there. They haven't said anything true. [...]</p> <p>The perfected, generated, prompted resume is generic and impersonal. It tells me nothing about this person, other than that they use particular tools.</p></blockquote> <p class="cite">&mdash; <a href="https://macwright.com/2026/06/24/accidental-anonymity.html">Tom MacWright</a>, Accidental anonymit…

• simonw/browser-compat-db
  simonw· 24-jun

  <p><strong><a href="https://github.com/simonw/browser-compat-db">simonw/browser-compat-db</a></strong></p> Inspired by Mozilla's <a href="https://developer.mozilla.org/en-US/blog/introducing-mdn-mcp-server/">new MDN MCP service</a> - <a href="https://github.com/mdn/mcp">source code here</a> - I decided to try converting their comprehensive <a href="https://github.com/mdn/browser-compat-data">mdn/browser-compat-data</a> repository full of browser compatibility data into a SQLite database.</p> <p>This new GitHub repo includes a Claude Code for web (Opus 4.8) <a href="https://github.com/simonw/browser-compat-db/blob/main/build_db.py">generated script</a> for doing that using <a href="https://github.com/simonw/sqlite-utils">sqlite-utils</a>.</p> <p>I wanted the resulting ~66MB SQLite database…