PR Fixes Qwen3 Text Encoder Crashes in FP8 and GGUF Formats for Z-Image

A pull request (#1904) submitted by developer arifanchan on June 13, 2026, addresses two separate crashes that occur when loading the Qwen3 text encoder in quantized FP8 Safetensors and GGUF formats alongside the Z-Image model. The first issue stemmed from FP8 checkpoints omitting the lm_head.weight key to save disk space, causing a strict key check in offload.py to fail; the fix dynamically aliases the existing embed_tokens weight to that slot without duplicating VRAM usage. The second crash arose from a data type mismatch during attention computation when using GGUF encoders, where the Value tensor was in bfloat16 while Query and Key were in float32, violating PyTorch's SDPA requirements. This was resolved by casting the entire text encoder to a uniform dtype immediately after loading. Both fixes were applied within models/z_image/z_image_main.py through a unified state dictionary preprocessor and a post-load dtype cast.

Developer Builds Autonomous AI Agent on Solana With On-Chain Policy Controls

A developer participating in the #100DaysOfSolana challenge has built an autonomous AI agent capable of reasoning, making decisions, and executing transactions on Solana's Devnet. The system was assembled over five days, with each session contributing a distinct component: an agent loop, blockchain tools, an MCP server, a policy engine, and a final autonomous workflow. The agent uses a large language model to decide which actions to take, but all fund transfers must pass through a deny-by-default policy engine before any transaction is signed. The developer noted that the policy engine, rather than the AI model, serves as the true core of the system, since the LLM only determines intent while the surrounding infrastructure enforces safety. The project culminated in a documentation effort aimed at making the entire architecture reproducible by any developer without additional guidance.

London Gatwick Airport Introduces Robotic Self-Parking System

London Gatwick Airport has launched an automated robotic parking system developed by Stanley Robotics. The system allows travelers to drop off their vehicles at a designated point, after which a robot takes over and parks the car autonomously. This technology aims to optimize parking space usage and improve the overall airport experience for passengers. The deployment marks a significant step in the adoption of robotics within major UK transport infrastructure.

Modi Forms Task Force Under Nilekani to Overhaul India's Public Exam System

Prime Minister Narendra Modi has brought in Infosys co-founder and Aadhaar architect Nandan Nilekani to lead efforts aimed at reforming India's public examination framework. A high-powered task force is being set up to recommend technology-driven solutions for the National Testing Agency (NTA) and other public exams. The move comes amid growing concerns over exam paper leaks and systemic failures in the country's testing infrastructure. The task force is expected to focus on making the examination process more secure and resistant to malpractice. This initiative signals the government's intent to modernize and strengthen the credibility of high-stakes national exams.

Opposition demands Amit Shah's resignation over police action against NEET protesters

Opposition parties have intensified their attack on the government in Parliament, holding Home Minister Amit Shah responsible for police action against students protesting the NEET paper leak scandal. They are demanding his resignation over what they describe as a brutal assault on the protesting students. Meanwhile, the government is preparing to introduce amendments to the existing anti-paper leak legislation. Questions have also been raised in Parliament over the appointment of a key higher education secretary.

Why Senior Engineers Read Tickets Before Writing a Single Line of Code

A fictional dialogue series on DEV Community uses an uncle-nephew conversation to explore how software is built from idea to production. In the first episode, an experienced engineer challenges his nephew — who has just completed React and RAG courses — to reconsider what separates junior from senior developers. The uncle argues that writing code is the final step of engineering, not the first, comparing it to a civil engineer who studies a river thoroughly before laying a single brick. Using a simple 'Add Wishlist Feature' ticket as a case study, he demonstrates how plain-language requirements hide dozens of unresolved decisions around authentication, data persistence, and edge cases. The key lesson is that engineers must surface ambiguities that affect database schemas or API contracts before coding, while making reasonable assumptions about minor details to keep the sprint moving.

Trump Paused Iran Strikes to Allow Room for Peace Negotiations: Official

A US official stated that President Trump deliberately paused strikes against Iran to create space for peace talks. The announcement was made by Waltz, though he did not elaborate on the specifics or current status of the negotiations. A noticeable lull in hostilities between the US and Iran was observed over the weekend. The pause signals a potential diplomatic opening amid ongoing tensions between the two nations.

From Aadhaar to exams, Nilekani gets a new test

Developer Builds Auditable Palworld Breeding Calculator With Versioned, Verified Data

A developer has built an independent, ad-free Palworld breeding calculator called Palworld Breeding Combos, designed around data transparency and reproducibility. Each generated dataset carries a manifest recording the game version, source revision, generation timestamp, and verification status, so users always know exactly what data produced a given result. Rather than reimplementing the breeding formula, the project imports a pinned public dataset from a specific upstream release, normalizes it, and excludes any values that cannot be confidently mapped instead of guessing. A known data gap involving the Pal 'Astralym' was handled as a separately documented overlay, accepted only after two independent public sources confirmed the missing relationship. Validation scripts confirmed 288 of 288 breeding ranks and 251 of 251 unique combinations matched independent reference data, with seven unresolved codes excluded rather than approximated.

Developer Builds Open-Source Tool to Make LLM Conversation Context Manually Editable

A developer has created ThoughtDAG, an open-source prototype that reimagines how users interact with large language models by representing conversations as a directed acyclic graph on an infinite canvas. Each question-and-answer exchange becomes a node, and the edges between nodes directly determine which prior exchanges are included in the model's next request. Users can branch, merge, or prune conversation paths, giving them explicit control over what the model 'remembers' at any point. The tool runs entirely in the browser using IndexedDB, requires no account or hosted database, and supports both local inference via Ollama and OpenAI-compatible endpoints. Built with React, TypeScript, and the Vercel AI SDK under an MIT license, the project raises the open question of whether users prefer manual context control over automated memory management by agents or retrieval systems.

Sikandar Raza now second-most dismissed for duck as T20I captain with seven

Zimbabwe captain Sikandar Raza was dismissed for a golden duck by India pacer Yash Thakur in a recent T20I match. The dismissal took his total ducks as T20I captain to seven, moving him past Pakistan's Babar Azam on the unwanted list. Only former Australia captain Aaron Finch holds more ducks as a T20I captain, with eight. Babar Azam and India's Rohit Sharma are now tied just behind Raza, each having been dismissed for a duck six times as captain.

Rare 'Panda of the Sea' Fish Detected in Hong Kong Waters via DNA Tech

A new study has confirmed the presence of the critically endangered Chinese bahaba in Hong Kong's waters for the first time in over a decade. Researchers detected the fish not by capturing it, but by extracting its genetic material directly from seawater using environmental DNA technology. Traces of the fish's DNA were found across multiple western bays in the region. The Chinese bahaba, nicknamed the 'Panda of the Sea,' had eluded scientific observation for more than ten years. The findings mark a significant step forward in non-invasive marine monitoring and conservation efforts.

Terax bundles terminal, editor, git client, and AI agent in an 8 MB app

Terax is a new open-source developer environment that packs a terminal, code editor, git client with commit graph, web preview pane, and an AI agent panel into a binary of roughly 7–8 MB. Built on Tauri 2 and Rust rather than a bundled browser runtime, it installs without an account and collects no telemetry. The AI agent supports bring-your-own-key integration with providers like OpenAI, Anthropic, and Google Gemini, and also works with local inference tools such as Ollama and LM Studio. API keys are stored in the OS keychain and never written to disk, reflecting a local-first privacy approach. The project is still in early development — it is not yet code-signed — but installers are available on its GitHub Releases page, with packages also offered for Arch Linux and Nix users.

How to use third-party web components without deep npm knowledge

A developer on DEV Community shared a practical guide for using third-party web components after a friend struggled to follow standard npm-based installation instructions. The tutorial targets beginners unfamiliar with modern JavaScript tooling, walking them through setting up a simple project using the Vite bundler. Vite handles the build process, compiling project files into a deployable 'dist' folder ready for a web server. During development, running 'npm run dev' provides hot reload and sourcemap support, making debugging significantly easier than working with minified production builds. The guide aims to bridge the gap between basic HTML knowledge and the modern JavaScript ecosystem.

How Multi-Agent RAG Systems Using LangGraph Are Expanding AI Capabilities

Retrieval-Augmented Generation (RAG) combined with LangGraph, an open-source agent-building library, enables developers to create multi-agent AI systems where specialized agents collaborate to answer complex queries. LangGraph supports features like prompt management, LLM chaining, and memory, allowing autonomous agents to make decisions based on user input. A notable implementation, the End-to-End-Multi-AI-Agents-RAG-With-LangGraph-AstraDB-And-Llama-3.1 project on GitHub, showcases how such systems can be built using AstraDB and Meta's Llama 3.1 model. Multi-agent RAG architectures show promise across sectors including customer service, healthcare, and education, where agents can each handle domain-specific tasks. Researchers identify more advanced RAG algorithms and deeper integration with technologies like computer vision as key areas for future development.

Kargil Martyr's Mother Keeps Son's Room as Memorial 25 Years On

Grenadier Udeyman Singh, an 18-year-old soldier from Shamachak in Jammu, was killed in action during the Kargil War at Tiger Hill. His mother has preserved his room at home as a personal shrine in his memory. The space serves as a lasting tribute to the young soldier who laid down his life for the country. Decades later, the family continues to honour his sacrifice through this intimate memorial.

Sub2API Lets You Self-Host a Private AI API Gateway Across Multiple Providers

Sub2API is an open-source tool that allows developers to self-host an AI API gateway, placing a single base URL in front of multiple AI providers and accounts. The gateway handles authentication, routing, usage accounting, and concurrency limits, while issuing separate revocable keys to users or projects. Deployment requires a Linux server running Docker 20.10 or newer, Docker Compose v2, a domain, and at least one authorized upstream API account, with PostgreSQL and Redis bundled in the official stack. The project warns that redistributing quota from AI subscriptions may violate upstream terms of service, and advises keeping initial deployments private using only authorized accounts. Self-hosting becomes worthwhile primarily when managing multiple users or projects, as adding a gateway to a single-user setup can increase system fragility.

Ex-Isro chief, IIT director & more: Who is on Nilekani-led NTA exam reforms panel

Defen.so offers developers instant app security with WAF, scanning, and monitoring

A developer named '1fancy' has built and open-sourced Defen.so, an application security layer designed to be installed in seconds via a single CLI command. The tool bundles several security features including a managed WAF, SQL injection and XSS protection, bot detection, DDoS mitigation, and file upload checks. It also provides uptime monitoring, real-time attack logs, and alerts through email, Slack, Telegram, and webhooks. Defen.so includes a security scanner that flags exposed secrets, missing headers, and misconfigured cloud permissions common in rapidly built applications. The project is actively in development and targets developers, startups, and small teams who lack dedicated security infrastructure.

Iran Pauses Strikes on US Allies as Washington Holds Back Military Action

Iran has announced a halt to its attacks on US allies in the Middle East following a pause in hostilities by Washington. The ceasefire-like development comes as the conflict approaches its fifth month. Reports indicate that the United States has been constrained by concerns over depleting weapons stockpiles. These munitions shortages are said to have limited American plans for further military escalation in the region.