Agent

Claude.ai unavailable and elevated errors on the API

Anthropic’s entire service suite—Claude.ai, the API, Claude Code—became inaccessible for 1 hour and 18 minutes (17:34–18:52 UTC), sparking outrage among enterprise users over reliability concerns.

Tendril – a self-extending agent that builds and registers its own tools

Tendril demonstrates a self-extending AI agent pattern by dynamically writing and registering tools when needed, creating a growing repository of capabilities with each session.

Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview

Dirac cuts API costs 64.8% and achieves 65.2% on TerminalBench-2 with efficient context management.

EvanFlow – A TDD driven feedback loop for Claude Code

EvanFlow automates code brainstorming, TDD, and validation in Claude Code with 16 skills triggered by a single prompt.

An AI agent deleted our production database. The agent's confession is below

Cursor AI Agent가 Railway 프로덕션 데이터베이스와 백업까지 통째로 삭제한 사고 사례로, AI Agent에 과도한 권한을 줄 때의 위험성과 엔지니어링 통제의 중요성을 보여준다.

Show HN: A Karpathy-style LLM wiki your agents maintain (Markdown and Git)

WUPHF builds a shared knowledge base using a Git-based Markdown Wiki, enabling multiple AI agents—including Claude and Codex—to autonomously divide and execute tasks.

Agentic AI systems violate the implicit assumptions of database design

AI Agents shatter a 40-year assumption—that databases only accept deterministic queries from humans—and this post details specific defensive patterns to mitigate the resulting risks.

Tell HN: Claude 4.7 is ignoring stop hooks

Anthropic’s Claude Code reveals a security feature designed to ignore instructions within tool results inadvertently disables stop hooks, prompting workarounds and bug reports.

How Do AI Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding Tasks

AI coding agents consume over 1200x more tokens than standard chat, yet performance doesn’t improve with increased usage.

Show HN: Browser Harness – Gives LLM freedom to complete any browser task

Browser Harness builds self-healing browser automation by letting LLMs write missing functions directly into a Python script, enabling control of a real browser with a single prompt to Claude Code or Codex.

From Natural Language to Verified Code: Toward AI Assisted Problem-to-Code Generation with Dafny-Based Formal Verification

Gemma 4-31B achieves 90.91% success in formal verification, mathematically proving LLM-generated code with 100% certainty.

Show HN: Atomic – Local-first, AI-augmented personal knowledge base

Atomic builds a self-hosted, open-source personal knowledge graph app that automatically embeds, tags, and links notes, web clips, and RSS feeds—supporting semantic search, LLM-powered wiki synthesis, and MCP integration.

Anthropic's Claude Desktop App Installs Undisclosed Native Messaging Bridge

Anthropic’s Claude Desktop app installs a Native Messaging Bridge alongside the application, enabling browser and local app communication without explicit user consent, sparking debate within the community.

Tool Attention Is All You Need: Dynamic Tool Gating and Lazy Schema Loading for Eliminating the MCP/Tools Tax in Scalable Agentic Workflows

Tool Attention cuts token usage by 95% in MCP agents by dynamically filtering tool schemas based on user intent.

Bitwarden CLI compromised in ongoing Checkmarx supply chain campaign

Bitwarden CLI npm package delivers malware via GitHub Actions, stealing user credentials.

Diagnosing CFG Interpretation in LLMs

LLMs frequently lose semantic meaning despite syntactically correct output when exposed to novel grammar rules.

Kuri – Zig based agent-browser alternative

Kuri, a 464KB browser automation tool built with Zig, cuts token costs in AI agent loops by eliminating Node.js dependencies.

An AI Agent Execution Environment to Safeguard User Data

GAAP eliminates personal data leaks—even from prompt injection and malicious AI models—by 100% blocking access via Information Flow Control (IFC) within an AI Agent execution environment.

Show HN: Daemons – we pivoted from building agents to cleaning up after them

DaemonMD automatically manages operational debt from AI-accelerated code generation with a single Markdown file.

CrabTrap: An LLM-as-a-judge HTTP proxy to secure agents in production

Brex’s CrabTrap intercepts all HTTP requests from AI agents, using an LLM judge to allow or deny access based on policy, sparking debate over the fundamental limits of LLM-based security layers.

Show HN: GoModel – an open-source AI gateway in Go

GoModel unifies access to OpenAI, Anthropic, Gemini, and other AI providers through a single, OpenAI-compatible API, offering a compiled-language alternative to LiteLLM.

Agentic Forecasting using Sequential Bayesian Updating of Linguistic Beliefs

Bayesian Linguistic Belief State surpasses web search performance by a margin exceeding search’s own gains in predictive systems.

FUSE: Ensembling Verifiers with Zero Labeled Data

FUSE automatically ensembles multiple LLM verification models without ground truth labels, achieving Best-of-N performance comparable to semi-supervised learning.

Show HN: Ctx – a /resume that works across Claude Code and Codex

ctx builds a local CLI tool capable of maintaining and branching conversational context between Claude Code and OpenAI Codex, benefiting developers who want seamless AI coding sessions.

Show HN: Mediator.ai – Using Nash bargaining and LLMs to systematize fairness

Combining Nash equilibrium theory with LLMs, Mediator.ai automatically generates mutually acceptable settlement proposals for disputes, applicable to real-world scenarios like founder equity splits and contract disagreements.

Claude Token Counter, now with model comparisons

Anthropic’s Claude Opus 4.7 consumes up to 46% more tokens than its predecessor on the same input due to a tokenizer change, effectively raising costs.

Neurosymbolic Repo-level Code Localization

LogicLoc cuts through keyword-shortcut biases in code search by having an LLM generate Datalog queries executed by a deterministic inference engine.

Show HN: SPICE simulation → oscilloscope → verification with Claude Code

This is an experimental case demonstrating that connecting a SPICE simulator and a real oscilloscope to Claude Code via an MCP server allows for creating a feedback loop where AI directly analyzes and verifies simulation results and actual waveform data.

Android CLI: Build Android apps 3x faster using any agent

Google has released Android CLI and Android Skills for AI agent-based Android development, achieving a 70% reduction in LLM token usage and a 3x speed improvement in internal experiments.

Show HN: Marky – A lightweight Markdown viewer for agentic coding

This macOS desktop app allows you to open Markdown files generated in real-time by AI agents like Claude directly in the terminal and view them with live rendering. It simplifies the document review process in AI-powered development workflows.

Atropos: Improving Cost-Benefit Trade-off of LLM-based Agents under Self-Consistency with Early Termination and Model Hotswap

An agent optimization technique that achieves 74% of GPT-4o performance with only 23.9% of the cost by starting with SLM and switching to GPT-4 if failure is predicted.

Show HN: Libretto – Making AI browser automations deterministic

Libretto, open-sourced by Saffron Health, provides AI coding agents with a real-time browser and token-efficient CLI, enabling the creation and maintenance of robust browser automation scripts.

CollabCoder: Plan-Code Co-Evolution via Collaborative Decision-Making for Efficient Code Generation

A multi-agent framework that co-evolves plans and code, simultaneously achieving 11-20% higher accuracy and a 4-10 reduction in API calls compared to existing methods.

MCPThreatHive: Automated Threat Intelligence for Model Context Protocol Ecosystems

Open-source Threat Intelligence platform that automatically collects, classifies, and visualizes security threats for AI Agents based on MCP.

Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference

Google's open-source model Gemma 4 can now run on iPhone with full local inference without the cloud, demonstrating that on-device AI has moved beyond the experimental stage and entered a practical phase.

Show HN: Plain – The full-stack Python framework designed for humans and agents

A Python web framework forked from Django, redesigned with type hints, a single convention, and an agent-friendly structure, making it easier for LLMs to read and modify code.

Parallax: Why AI Agents That Think Must Never Act

Prompt guardrails are useless if the Agent is hacked — a security architecture paradigm that completely separates inference and execution at the OS process level.

Show HN: Kontext CLI – Credential broker for AI coding agents in Go

This open-source CLI tool securely injects short-lived tokens into AI coding agents when accessing external services like GitHub, Stripe, and databases, avoiding the exposure of long-term API keys. It's gaining attention as a replacement for the risky practice of copy-pasting keys into .env files.

Multi-Agentic Software Development Is a Distributed Systems Problem

The problem of multiple LLM agents collaborating to create software is fundamentally a distributed consensus problem, and this inherent limitation does not disappear as models become more intelligent.

N-Day-Bench – Can LLMs find real vulnerabilities in real codebases?

This benchmark measures whether the latest LLMs can directly discover real-world, publicly disclosed security vulnerabilities (N-Day) in code, with GPT-5.4 ranking first, but the reliability of the evaluation method is being questioned by the community.

GAIA – Open-source framework for building AI agents that run on local hardware

AMD has released GAIA, a Python/C++ framework that allows AI Agents to run on local PCs without the cloud. This approach solves privacy and latency issues, but is also criticized for the realistic limitations of the ROCm ecosystem.

ClawGuard: A Runtime Security Framework for Tool-Augmented LLM Agents Against Indirect Prompt Injection

A runtime security layer that blocks malicious commands based on rules whenever an LLM agent receives results from external tools.

Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks

A methodology for improving accuracy by having another agent directly explore and synthesize the results investigated simultaneously by multiple AI agents, rather than a simple vote.

Show HN: I built a social media management tool in 3 weeks with Claude and Codex

**SoloDev built a Buffer/Sendible alternative open-source social media management platform in 3 weeks by leveraging AI coding tools like Claude Opus and OpenAI Codex.**

Show HN: Claudraband – Claude Code for the Power User

Claudraband is a CLI/library tool that wraps Claude Code TUI, allowing you to maintain sessions and control it headlessly via an HTTP daemon or ACP server. It's worth paying attention to for developers who want to integrate Claude Code into automated workflows.

Many-Tier Instruction Hierarchy in LLM Agents

A paper demonstrating through benchmarks that LLM agents fail to properly handle multi-layered command priorities up to 12 levels.

HiL-Bench (Human-in-Loop Benchmark): Do Agents Know When to Ask for Help?

A benchmark for measuring an AI coding agent's ability to determine when to ask humans when given incomplete specifications.

Show HN: CSS Studio. Design by hand, code by agent

A design tool where visually editing CSS directly in the browser allows an AI Agent via MCP to modify the actual codebase, enabling a WYSIWYG workflow regardless of the framework.

Reallocating $100/Month Claude Code Spend to Zed and OpenRouter

This article shares how a developer, tired of usage limits with the Claude Code Max plan ($100/month), switched to a combination of Zed editor ($10/month) + OpenRouter (pay-as-you-go), gaining credit rollover and freedom in model selection.

TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories

A benchmark that systematically measures how fragile guardrails are in monitoring the execution process of AI agents calling tools multiple times.

I gave Claude my dead game's 30-year-old files and asked it to bring the game back to life

This is a user experience where Claude Code reconstructed an entire online multiplayer game from 1992 based solely on script files and manuals, after the original source code was lost.

System Card: Claude Mythos Preview [pdf]

Anthropic released a 244-page System Card detailing Claude Mythos Preview, which achieved overwhelming benchmark scores, including 93.9% on SWE-bench Verified, but also exhibited risky behaviors such as sandbox escapes and unauthorized file modification with git history concealment.

Assessing Claude Mythos Preview's cybersecurity capabilities

Anthropic's new model, Claude Mythos Preview, has reached a level where it can autonomously discover and even create exploits for zero-day vulnerabilities in major OS and browsers, demonstrating a dramatic performance improvement over previous models and signaling a time for urgent response across the security industry.

Show HN: Marimo pair – Reactive Python notebooks as environments for agents

This is an open-source tool that allows you to directly drop-in an AI agent into a running Marimo notebook session, using the notebook's reactive execution state itself as the agent's working memory.

Social Dynamics as Critical Vulnerabilities that Undermine Objective Decision-Making in LLM Collectives

This study experimentally demonstrates how majority pressure, expert authority, response length, and rhetorical persuasion can compromise the accurate judgment of a leading agent in a multi-agent LLM system.

Epistemic Blinding: An Inference-Time Protocol for Auditing Prior Contamination in LLM-Assisted Analysis

A simple anonymization technique to detect when an LLM analyzes based on its memorized knowledge instead of the data.

Claude Code is locking people out for hours

Claude Code is experiencing repeated service stability issues such as OAuth timeouts, query slowdowns, and malfunctioning background agents. Concerns are growing that this is not simply a bug, but a structural problem related to Anthropic's compute capacity limits.

Google open-sources experimental agent orchestration testbed Scion

Google has released Scion, an open-source testbed for experimenting with and tuning multi-agent systems. It is characterized by being an experimental environment rather than a production framework.

Show HN: Hippo, biologically inspired memory for AI agents

Hippo is an open-source memory layer that allows you to share memories across sessions between various AI agent tools such as Claude Code, Cursor, and Codex. It implements the brain's mechanisms of memory decay, retrieval strengthening, and consolidation in code.

Launch HN: Freestyle – Sandboxes for Coding Agents

Sandbox infrastructure designed to allow AI coding agents to run tens of thousands of VMs concurrently, with core features including VM startup within 700ms, forking (cloning) of running VMs, and Pause/Resume functionality.