Published on April 17, 2026 at 06:01 CEST (UTC+2)
Claude Opus 4.7 (1554 points by meetpateltech)
Anthropic announces the general availability of Claude Opus 4.7, a significant upgrade focused on advanced software engineering and complex, long-running tasks. It features improved vision capabilities, greater creativity for professional outputs, and better performance on benchmarks compared to its predecessor. Notably, it is released with tested cyber safeguards, as part of a cautious strategy following the announcement of the more powerful but restricted Claude Mythos Preview.
Codex for almost everything (739 points by mikeevans)
While the content preview is unavailable, the title "Codex for almost everything" and its high score suggest an OpenAI announcement or article about expanding the capabilities or applications of Codex, their AI system for translating natural language to code. It likely discusses making the technology more general-purpose and accessible for a wider range of tasks beyond pure code generation.
CadQuery is an open-source Python library for building 3D CAD models (63 points by gregsadetsky)
CadQuery is introduced as an open-source Python library for creating parametric 3D CAD models programmatically. This approach allows models to be defined by code, enabling version control, easy sharing, and parameterization without the need for a graphical user interface. It represents a shift towards code-driven design in computer-aided design and manufacturing.
Guy builds AI driven hardware hacker arm from duct tape, old cam and CNC machine (112 points by scaredpelican)
A developer has created "AutoProber," an AI-driven hardware hacking automation stack built from repurposed components like a camera and CNC machine. The system allows an AI agent to autonomously handle the entire process of discovering targets on a circuit board, mapping them with a microscope, and performing safe, controlled pin probing. This project demonstrates the application of AI agents to physical hardware security testing and reverse engineering.
Show HN: Spice simulation → oscilloscope → verification with Claude Code (35 points by fizz_buzz)
The author explores using Claude Code for hardware development by integrating it with a SPICE simulator and an oscilloscope for immediate feedback. This workflow is valuable for validating circuits, embedded programming, and automating tedious data analysis tasks like aligning measurement data. Key lessons include preventing the AI from guessing hardware connections and managing data context efficiently.
A Better R Programming Experience Thanks to Tree-sitter (100 points by sebg)
This blog post explains how the integration of Tree-sitter, a parsing generator, has significantly improved the R programming experience. It enables advanced tooling features such as code reformatting, linting, better IDE autocompletion, and enhanced code search on GitHub. The post details how this underlying technology allows for more sophisticated and reliable analysis of R code structure.
Android CLI: Build Android apps 3x faster using any agent (153 points by ingve)
Google's Android team introduces the Android CLI and a suite of tools designed to integrate with any AI coding agent (like Gemini, Claude Code, or Codex) to speed up Android app development. These tools provide agents with structured knowledge and skills for core Android workflows, aiming to eliminate guesswork and enforce best practices when developing outside of Android Studio.
Substrate AI Is Hiring Harness Engineers (1 points by kunle)
Substrate, a startup building an AI-native service for healthcare revenue cycle management, is hiring a Harness Engineer. The role focuses on building systems around AI agents that process healthcare claims, requiring the engineer to improve the reliability and precision of these systems as they navigate complex healthcare contracts and sensitive financial infrastructure.
Official Clojure Documentary page with Video, Shownotes, and Links (124 points by adityaathalye)
The official page for the Clojure programming language documentary provides the full video, show notes, and references. It traces Clojure's origins, its philosophy rooted in academic research on reducing software complexity, and its impact within the tech industry, notably at large companies like Nubank. The page lists key research papers and books that influenced the language's design.
288,493 Requests – How I Spotted an XML-RPC Brute Force from a Weird Cache Ratio (11 points by taubek)
The author details how they identified a large-scale XML-RPC brute force attack on a WordPress site by noticing an anomalous drop in Cloudflare's cache hit ratio to 0.8%. The attack originated from a single IP making hundreds of thousands of requests to xmlrpc.php. The post explains the technical mechanism of the attack and recommends a defense-in-depth strategy using WAF rules and WordPress plugins to block such exploits.
1. AI Agents Moving from Digital to Physical Domains * Why it matters: Projects like AutoProber (hardware hacking) and the SPICE/Oscilloscope integration demonstrate that AI agents are no longer confined to software and text. They are being equipped with tools to perceive and act upon the physical world, enabling automation in engineering, manufacturing, and security. * Implication: This expands the addressable market for AI and creates a new frontier for R&D. It necessitates the development of robust safety frameworks and precise tool-use APIs to allow agents to interact with real-world systems without causing harm.
2. The Rise of "Harness Engineering" and AI Systems Reliability * Why it matters: Job postings like Substrate's for a "Harness Engineer" highlight a new specialization focused on building the scaffolding, tooling, and control systems around core AI models. The goal is to improve the success rates of probabilistic AI outputs in critical, production-grade applications. * Implication: As AI moves into enterprise cores (like healthcare billing), mere model performance is insufficient. A new engineering discipline is emerging to make AI systems deterministic, auditable, and reliable enough for business-critical operations.
3. Vertical Integration: Platforms Providing Native AI Toolchains * Why it matters: Google's release of Android CLI with "Android skills" shows major platforms are not just providing AI models, but are building official, vertically integrated toolchains that optimize the agentic workflow for their specific domain. * Implication: This lowers the barrier to AI-assisted development within specific ecosystems but may lead to platform lock-in. It forces AI model providers (like Anthropic, OpenAI) to ensure their models can effectively utilize these specialized toolkits.
4. Strategic Model Deployment with Differential Capabilities * Why it matters: Anthropic's release of Opus 4.7, with deliberately reduced cyber capabilities compared to Mythos Preview, illustrates a cautious, staged deployment strategy for high-risk AI features. Safety measures are being tested on less capable models before broader release. * Implication: AI development is becoming more nuanced, with capabilities being fine-tuned and governed per domain. This reflects a growing industry focus on AI safety and security, moving beyond a pure "capability race" to a "responsible capability scaling" approach.
5. AI as a Catalyst for New Developer Tooling Paradigms * Why it matters: The Tree-sitter integration for R and libraries like CadQuery represent a shift towards tools that provide deep, structured understanding of code and models. This is partly driven by the need to feed clean, semantically parsed data to AI assistants and to build better AI-powered developer experiences. * Implication: The infrastructure of software development itself is being upgraded to be more "AI-friendly." This creates opportunities for new tools that abstract complexity and allow both humans and AI agents to reason more effectively about code, designs, and data.
Analysis generated by deepseek-reasoner