Show HN: MailPilot – Freedom to Go Anywhere While Your Agents Work
MailPilot is a terminal‑based user interface that wraps local large language model agents and forwards each interaction to an email thread. The tool lets users and their AI agents communicate asynchronously via email, freeing the user from a constant desk presence. A built‑in CC capability turns a single‑user session into a collaborative, multiplayer conversation, and it works out of the box with Claude, Codex, Gemini, Copilot and OpenCode. The creator positioned it as a productivity‑first product for remote and mobile professionals.
ChromaDB Explorer
Chroma Explorer is a native macOS desktop client designed for ChromaDB, enabling users to browse collections, perform semantic searches, and manage vector embeddings through a polished interface. The app supports multi‑profile connections to local, remote, or Chroma Cloud instances and...
Native ZFS VDEV for Object Storage (OpenZFS Summit)
At the OpenZFS Developer Summit 2025, Zettalane unveiled MayaNAS and MayaScale, highlighting objbacker.io—a native ZFS VDEV that eliminates FUSE overhead and streams data directly from S3, GCS, and Azure Blob. The solution aligns 1 MB ZFS blocks with individual objects, delivering...
Show HN: Tabstack – Browser Infrastructure for AI Agents (by Mozilla)
Mozilla’s Tabstack launches an API that handles the web‑layer for AI agents, turning raw URLs and intents into clean, structured data. It dynamically escalates from lightweight fetches to full headless browser sessions only when JavaScript rendering is required. The service...
HTTP RateLimit Headers
An IETF draft proposes standard HTTP RateLimit-Policy and RateLimit headers to inform clients about quota and window constraints, helping avoid 429 errors. The draft originally assumes quota‑reset algorithms, but the article argues that headers can work with any rate‑limit strategy,...
Show HN: Webctl – Browser Automation for Agents Based on CLI Instead of MCP
Webctl is a new command‑line tool that lets AI agents and developers automate Chromium browsers via simple CLI commands instead of the traditional Model‑Controlled‑Prompt (MCP) approach. By shifting control of what data enters the LLM context to the user, it...
Show HN: Run LLMs in Docker for Any Language without Prebuilding Containers
The open‑source tool agent‑en‑place lets developers spin up on‑demand Docker containers that match the exact language runtimes defined in a project’s configuration files. It automatically parses .tool‑versions, mise.toml and other idiomatic version files, builds a Debian‑based image with the required tools,...
I Love You, Redis, but I'm Leaving You for SolidQueue
Rails 8 removes Redis from its default stack, replacing it with SolidQueue, SolidCache, and SolidCable that run on the existing relational database. The change eliminates the need for a separate key‑value store for job queuing, caching, and real‑time messaging, simplifying infrastructure...
Show HN: OpenWork – an Open-Source Alternative to Claude Cowork
OpenWork is an open‑source, desktop‑first alternative to Claude Cowork that wraps the OpenCode framework in a guided visual workflow. It lets knowledge workers pick workspaces, start runs, watch real‑time progress, and approve permissions without using a command line. The initial...
Exa-D: How to Store the Web in S3
Exa‑d is Exa’s in‑house data‑processing framework that stores the entire web on S3 using Lance and manages billions of derived artifacts such as embeddings, tokenized text, and metadata. It treats data as typed columns with declarative dependencies, automatically building a...
Open Sourcing Dicer: Databricks's Auto-Sharder
Databricks has open‑sourced Dicer, its proprietary auto‑sharding system that dynamically assigns key ranges to service pods. The platform replaces static sharding and pure‑stateless designs, eliminating downtime during restarts, autoscaling, and hot‑key spikes. In production, Dicer drives over 90% cache‑hit rates...
Every GitHub Object Has Two IDs
A developer building a link feature for Greptile discovered that GitHub’s GraphQL node IDs and the integer IDs used in URLs are two distinct identifier systems. By decoding the Base64‑encoded portion of the node ID, he found the database ID...
Apple Creator Studio
Apple introduced Apple Creator Studio, a subscription bundle that unites Final Cut Pro, Logic Pro, Pixelmator Pro, Motion, Compressor, MainStage and AI‑enhanced features for Keynote, Pages, Numbers, and future Freeform. The suite launches on Jan 28, 2026, priced at $12.99 per...
Show HN: SnackBase – Open-Source, GxP-Compliant Back End for Python Teams
SnackBase is an open‑source Python backend that provides instant CRUD APIs, built‑in GxP compliance, and immutable audit logs. It offers multi‑tenancy, row‑level security, and a comprehensive permissions engine, allowing SaaS developers to isolate data across organizations. The platform includes out‑of‑the‑box...
Provenance Is the New Version Control
AI‑driven code synthesis is turning traditional version control on its head. When a component can be regenerated from a formal specification, the program text ceases to be the primary source of truth, and the reasons—requirements, constraints, and design decisions—become the...
Cowork: Claude Code for the Rest of Your Work
Anthropic unveiled Cowork, a research‑preview extension of Claude Max that gives the AI direct access to user‑selected folders on macOS. The tool lets Claude read, edit, create, and organize files, effectively acting as a virtual coworker for both coding and...
Why We Built Our Own Background Agent
Ramp unveiled its proprietary background coding agent, Inspect, which automates code generation and verification within fully sandboxed VMs. The platform integrates with tools like Sentry, Datadog, LaunchDarkly, GitHub, Slack, and Buildkite, supporting both backend and frontend workflows. Inspect sessions launch...
Computational Complexity of Schema-Guided Document Extraction
Pulse discovered that while JSON‑schema guided extraction simplifies data pipelines, enforcing complex, nested schemas during LLM generation incurs significant compute overhead. Naïve constrained decoding either explodes in state space or silently degrades extraction quality. Experiments on thousands of varied invoices...
Why Ontario Digital Service Couldn't Procure '98% Safe' LLMs (15M Canadians)
Ontario’s Digital Service discovered that high‑stakes institutions cannot adopt AI based solely on probabilistic safety claims such as "98% safe." Decision‑makers need governance mechanisms that can be defended if a failure occurs. The team built an "Authority Boundary Ledger" that...
Show HN: Agent-of-Empires: OpenCode and Claude Code Session Manager
Agent of Empires (aoe) is an open‑source terminal session manager written in Rust that wraps tmux to orchestrate AI coding agents like Claude Code and OpenCode. It provides a TUI dashboard, hierarchical grouping, multi‑profile workspaces, and automatic status detection, allowing...
Zorgdomein Integration: A Guide to Secure .NET and Azure Architecture
Zorgdomein integration demands a “Double‑Lock” security model that couples mutual TLS with specialized JWTs, forcing .NET teams to reconfigure IIS for client‑certificate authentication. The project built custom JWT middleware to enforce Dutch‑specific claim rules, ensuring that only authorized providers can...
Unauthenticated Remote Code Execution in OpenCode
OpenCode, an open‑source AI coding assistant, shipped an unauthenticated HTTP server that ran silently on startup in versions prior to 1.1.10. The server exposed endpoints for arbitrary shell execution, PTY sessions, and file reads, allowing any local process or web...
FUSE Is All You Need – Giving Agents Access to Anything via Filesystems
The article proposes using FUSE to expose any data source as a sandboxed filesystem for AI agents, illustrated with an email‑assistant example. By implementing standard file operations, agents can browse, read, move, and flag items using familiar Unix commands, eliminating...
Elasticsearch Was Never a Database
Elasticsearch was designed as a Lucene‑based search engine, not as a system of record. Over the past decade many teams have promoted the index to a primary datastore, only to encounter consistency gaps, immutable mappings, and limited query capabilities. The...
Show HN: PrintReadyBook
PrintReadyBook is an AI‑driven platform that creates complete books in minutes. Users supply a concept, choose settings, and receive a professionally formatted manuscript, an editable DOCX, AI‑generated cover art, and print‑ready PDFs for both interior and cover. The service highlights...
Show HN: Librario, a Book Metadata API that Aggregates G Books, ISBNDB, and More
Librario is a pre‑alpha, open‑source book‑metadata API written in Go that consolidates information from Google Books, ISBNDB, Hardcover and other sources into a single response. It merges conflicting data using a priority‑based strategy and stores results in a PostgreSQL database...
Anthropic Blocks Third-Party Use of Claude Code Subscriptions
Anthropic has begun blocking third‑party access to Claude Code subscriptions, causing widespread failures for developers using the Opencode plugin. The outage stems from a new request‑signature enforcement that the plugin did not meet, prompting users to downgrade, cancel, or switch...
ChatGPT Health Is a Marketplace, Guess Who Is the Product?
OpenAI unveiled ChatGPT Health, a service that links users’ medical records, wellness apps, and insurance data to generate personalized health profiles. The rollout partners with b.well Connected Health, a firm that primarily serves insurers, turning the platform into a data‑driven...
Show HN: DeepDream for Video with Temporal Consistency
The deepdream-video-pytorch fork adds RAFT optical flow and occlusion masking to the original PyTorch DeepDream implementation, enabling temporally consistent video hallucination. A new CLI, video_dream.py, warps previous dream frames into the current frame, producing smooth transitions and reducing ghosting. The...
ChatGPT Health
OpenAI unveiled ChatGPT Health, a secure, dedicated interface that merges personal health data with ChatGPT’s conversational AI. The service adds purpose‑built encryption, isolated storage, and a strict no‑training policy to protect sensitive medical information. Users can link electronic medical records...
NPM to Implement Staged Publishing After Turbulent Shift Off Classic Tokens
GitHub postponed its new billing model for self‑hosted Actions after strong developer pushback, while still implementing price cuts for hosted runners on Jan 1, 2026. In parallel, a sophisticated spearphishing campaign leveraged 27 compromised npm packages to harvest credentials from manufacturing, automation...
Show HN: 30k IKEA Items in Flat Text
A new open dataset titled "IKEA US – CommerceTXT" provides a complete snapshot of 30,511 IKEA products in a lightweight, line‑oriented text format. The CommerceTXT protocol cuts token usage by about 24% compared with traditional JSON, saving roughly 3.6 million tokens...
Everyone Hates OneDrive, Microsofts Cloud App that Steals and Deletes Files
Microsoft’s OneDrive is now being accused of employing dark‑pattern tactics that automatically sync users’ files to the cloud and, in many cases, delete the local copies without clear warning. Users report that after enabling the service, OneDrive silently uploads data,...
Elo – A Data Expression Language Which Compiles to JavaScript, Ruby, and SQL
Elo is a new data‑expression language that compiles to JavaScript, Ruby, and SQL while preserving identical semantics. It targets no‑code and low‑code tools, letting non‑technical users manipulate data safely through a pure‑data model and built‑in schema validation. Features include first‑class...
CI-Hush
The open‑source ci‑hush tool introduces lightweight scripts that let GitLab CI users collapse noisy command output into expandable sections. By wrapping commands with a start and end script, developers see a concise label instead of hundreds of lines, yet can...
Laylo (YC S20) – Head of Growth (Organic and Partners and Loops and AI) – Remote US
Laylo, the Y Combinator‑backed Drop CRM for musicians and live events, is hiring a Head of Growth to lead its organic, influencer, partner and product‑loop initiatives. The remote, full‑time role offers a $150K‑$200K salary plus 0.01%‑0.03% equity and reports directly to...
High-Performance DBMSs with Io_uring: When and How to Use It
The paper investigates how Linux io_uring can be harnessed by modern database management systems to lower I/O overhead. It evaluates two scenarios: a storage‑bound buffer manager and a network‑bound analytical shuffling workload. Advanced io_uring features such as registered buffers and passthrough...
Locating a Photo of a Vehicle in 30 Seconds with GeoSpy
Vehicle theft investigations are gaining a powerful new ally with Graylark's GeoSpy suite. The original GeoSpy tool offered 1‑25 km photo‑geolocation, but its new SuperBolt model now pinpoints locations to within one meter. By pairing a broad‑scale estimation step with a...
Show HN: VaultSandbox – Test Your Real MailGun/SES/Etc. Integration
VaultSandbox is an open‑source Docker gateway that lets developers test real email providers such as MailGun, SES, Postmark, and SendGrid inside their VPC. It auto‑provisions TLS certificates and DNS records, enabling full SPF, DKIM, DMARC, and TLS validation without exposing...
Show HN: Mantic.sh – Search 480k Files in 0.46s without Embeddings
Mantic.sh is a structural code‑search engine that ranks files in under 500 ms across massive repositories, such as Chromium’s 480 k files, without relying on embeddings, vector databases, or external services. Version 1.0.12 resolves scoring bugs, restores proper confidence scores, and improves reliability....
Show HN: DDL to Data – Generate Realistic Test Data From SQL Schemas
DDL to Data is a web‑based tool that creates realistic test datasets directly from SQL CREATE TABLE statements. By parsing schema definitions it automatically respects foreign‑key links, uniqueness constraints, and data types, producing plausible values such as email addresses and...
System: Control Your Mac From Anywhere Using Natural Language
SYSTEM lets users control their Mac from any device using natural‑language prompts, pairing a Cloudflare Workers‑based AI agent with a local bridge that runs AppleScript, shell commands, and Raycast extensions. The split architecture keeps the brain in the cloud while...
LocalFirst: You Keep Using That Word
Local‑first software aims to keep users productive regardless of network availability, attacking vendor lock‑in by ensuring data, formats, and services remain under user control. The article argues that the term describes a spectrum rather than a binary label, highlighting that...
Show HN: Tailsnitch – A Security Auditor for Tailscale
Tailsnitch is a command‑line security auditor for Tailscale that scans a tailnet for over 50 misconfigurations and best‑practice violations. It supports OAuth and API‑key authentication, offers an interactive fix mode, and can export SOC 2‑compatible evidence. The tool integrates with CI/CD...
Observability's Past, Present, and Future
The article maps observability from its birth in the early 2010s—driven by distributed tracing and a need to tame cloud‑native complexity—to its present state as a mandatory yet burdensome engineering layer. It shows how over‑instrumentation, stale dashboards, and noisy alerts...
Show HN: OSS Sustain Guard – Sustainability Signals for OSS Dependencies
OSS Sustain Guard is an open‑source, multi‑language analyzer that evaluates the health of software dependencies using 24 core sustainability metrics. It scans packages across Python, JavaScript, Go, Rust, Java, and many other ecosystems, delivering scores on maintainer activity, community engagement,...
Show HN: An LLM-Powered PCB Schematic Checker (Major Update)
Traceformer is an LLM‑powered PCB schematic checker that analyzes KiCad or Altium projects through a three‑phase multi‑agent pipeline. The Planner creates work orders, up to ten parallel Workers retrieve datasheet evidence and run lightweight models, and the Merger compiles a...
OpenGitOps
OpenGitOps is an open‑source initiative defining standards, best practices, and education for GitOps adoption. It codifies version‑controlled, declarative infrastructure management through four core principles: declarative state, immutable versioning, automatic pulling, and continuous reconciliation. Backed by industry leaders like Google’s Kelsey...
Claude Code On-the-Go
A developer runs six Claude Code agents from an iPhone using Termius, mosh, and a Tailscale‑secured Vultr VM. The VM is started and stopped on demand via iOS shortcuts, costing about $0.29 per hour. Push notifications from a custom webhook...
Show HN: Uvx Ptn, Scan a QR, Get a Terminal in Your Phone
Porterminal is a lightweight tool that creates a Cloudflare tunnel and presents a QR code, letting developers launch a full‑featured terminal on their phone with a single command. It eliminates the need for SSH keys, port‑forwarding, or separate mobile clients,...