DevHunt Daily
Friday, November 20, 2026
- 1LanguagesHacker News
PHP 8.6 released with partial function application, Duration class, and polling API
PHP 8.6 adds partial function application, a new Time\Duration class, a scalable polling API, a clamp() function, and other type‑safe features. These changes improve code expressiveness, type safety, and I/O performance, which developers can leverage in modern PHP applications.
- 2DatabasesHacker News
PostgreSQL 19 released with REPACK, plan advice and improved autovacuum
PostgreSQL 19 adds a REPACK (CONCURRENTLY) command for non‑blocking table cleanup, a pg_plan_advice extension for stabilizing query plans, and enhancements to autovacuum and monitoring. These changes aim to simplify administration and improve performance for large‑scale workloads, which matters to developers and DBAs.
- 3SecurityHacker News
OpenSSL fixes buffer overflow in EVP_EncodeBlock()
A patch returns -1 for invalid inputs to EVP_EncodeBlock(), addressing CVE-2026-121481. The change adds guards for null buffers, negative lengths, and size overflows, and includes tests and documentation updates. Developers using OpenSSL will receive safer base64 encoding functions.
- 4SecurityLobsters
Java4Shell vulnerability requires JVM flag change and upgrade
A flaw called Java4Shell allows the JVM to replace Java code with native code, creating a security risk. Developers can mitigate it by adding the -XX:-UseBASE64Intrinsics flag and updating to a patched JVM version. The issue highlights concerns about native code in widely used runtimes.
- 5SecurityHacker News
Synology releases critical security update for DSM and BeeStation OS
Synology issued an advisory (SA-26:21) for a critical vulnerability (CVE-2026-120382) that allows remote attackers to read/write files and cause denial‑of‑service. The fix is available for DSM 7.4, 7.3 and recommended upgrades for older DSM versions; BeeStation OS 1.5 also affected. Developers using Synology devices should apply the updates promptly.
- 6AIGitHub Trending
MiniMind lets you train a 64M LLM from scratch in 2 hours for under $3
The open‑source MiniMind project provides code to train a 64‑million‑parameter language model on a single GPU in about two hours, costing roughly three dollars in cloud GPU rental. It includes full training pipelines—from preprocessing to RLHF—and supports inference via standard APIs, helping developers explore LLM internals without large resources.
github.com238 stars today - 7AIGitHub Trending
Firecrawl adds local binary and cloud API for web scraping and AI agent integration
Firecrawl now offers a local command‑line tool that scrapes, crawls and maps websites into markdown, plus a hosted cloud service with structured JSON, search and 550+ data sources. The local tool can be added to AI coding assistants like Claude Code, Cursor, Windsurf and Gemini CLI, letting developers run scraping on‑premise or via the cloud.
github.com825 stars today - 8Coding agentsGitHub Trending
CLI-Anything adds hub for installing community-built CLIs
CLI-Anything provides a pip‑installable hub that lets developers browse, install, and manage community‑created command‑line interfaces for AI agents. The hub supports quick setup and contributions, enabling software to become agent‑native across tools like Cursor and Claude Code.
github.com898 stars today - 9AIGitHub Trending
OpenVINO toolkit released for optimizing and deploying AI inference
OpenVINO is an open‑source toolkit that helps developers optimize deep‑learning models and run inference on CPUs, GPUs and Intel accelerators. It supports models from PyTorch, TensorFlow, ONNX and other frameworks, and can be installed via pip. The project provides documentation, tutorials and sample notebooks for various AI tasks.
github.com246 stars today - 10SecurityCloudflare
Cloudflare shields origins from five new nginx memory‑corruption vulnerabilities
Five memory‑corruption bugs in nginx (nginx2root) allow unauthenticated code execution, but Cloudflare's edge proxy blocks the most severe attacks for sites using its HTTP proxy. Developers should still update nginx to the patched versions, as some configurations remain vulnerable.
blog.cloudflare.com - 11AIGoogle DeepMind
Google DeepMind launches Gemini 4 Neon model with 1M token context
Google DeepMind announced Gemini 4 Neon, a new AI model that supports up to 1 million tokens and is priced at $3 per million input tokens. The model is available through the Gemini API, Google AI Studio, and Gemini for Business, offering strong performance on enterprise, analytical, and scientific benchmarks, which may benefit developers building complex workflows.
deepmind.google - 12AIGoogle AI
Gemini API adds preview Tool Retrieval to reduce token usage
Google announced Tool Retrieval in the Gemini Interactions API, letting agents load tool schemas on demand instead of sending the full library each request. This reduces input tokens, latency, and improves tool selection accuracy for agents handling many tools. The feature is available in preview with Gemini 3.6+ models.
blog.google - 13AIDevHunt launch
Robonito launches AI agent that writes, runs and fixes tests across multiple platforms
Robonito introduces an AI-driven QA agent that converts requirements into executable test cases for web, mobile, API, desktop, and IVR systems. It can self‑heal tests after UI changes and provides root‑cause analysis for failures, integrating with common development tools. A free tier offers limited test minutes, with paid plans for higher usage.
27 points - 14Dev toolsDevHunt launch
LaunchCrate releases Rust SaaS starter kit with Axum, SQLx, and React
LaunchCrate offers a Rust-based SaaS starter kit that includes an Axum API, SQLx database layer, and a React front‑end, preconfigured for organizations, SSO, billing, background jobs, and an admin panel. The kit is sold for a one‑time fee and can be deployed via Docker with minimal setup, appealing to developers building SaaS products in Rust.
22 points - 15Dev toolsDevHunt launch
Pacecar launches as a GitHub App that reviews pull requests in about 30 seconds
Pacecar is a GitHub App that can be installed on an organization to automatically comment on open, updated, and draft pull requests. It runs six standard checks—security, intent, scope, logic, reliability, and contracts—without nitpicking, and charges per review rather than per seat. The service costs $0.20 per PR open and $0.08 per push, with a $20 starter credit.
19 points