In eye of storm, Rajasthan Chief Justice opts out of hearing cases

Valuables inside treasure trove safe, match 1978 inventory: Jagannath Temple officials

Why LLMs Should Plan but Not Decide in AI Model Routing Systems

A software architecture article published on DEV Community outlines why relying solely on a large language model to route user prompts to AI generation models is unreliable in production environments. The author argues that LLMs should act as planners within a constrained pipeline, with final decisions enforced by deterministic code that validates model capabilities, user tier, and cost rules. The proposed system classifies generation tasks — such as text-to-image, image editing, multi-reference fusion, and video — before the LLM ever sees the request, narrowing the candidate model pool significantly. A hybrid routing approach is recommended, where strong intent signals are handled by rules and only genuinely ambiguous prompts are passed to an LLM classifier. The architecture also maintains a server-side model catalog to prevent issues like deprecated model IDs, unsupported resolutions, and malformed LLM output from reaching production.

Plastic waste homes built in 7 days for $300

Open-Source Tool Helps AI Agents Automate Flutter Split Package Migration

Flutter has restructured its architecture by splitting Material Design and Cupertino libraries into two standalone packages, material_ui and cupertino_ui, ending their long-standing bundling within the core SDK. This change enables independent release cycles and lighter framework builds, but requires significant code migration across existing Flutter applications. A key obstacle is that popular AI coding assistants like GitHub Copilot and Cursor are unaware of this change due to training data cutoffs, often generating outdated import paths. The open-source dart-sdk-skills repository addresses this gap by providing AI agents with updated knowledge and step-by-step migration runbooks. Developers can now trigger the full migration process through a single prompt by equipping their AI agent with these skills.

NCLT Deadlock Delays Resolution of Subhash Chandra's Rs 22,000 Crore Debt

A tribunal dispute has emerged over the proposed restructuring of media mogul Subhash Chandra's Rs 22,000 crore debt. The matter has reached a deadlock at the National Company Law Tribunal due to a difference of opinion among its members. The case has been referred back to the NCLT President to resolve the impasse. The President holds the authority to either appoint a third member or issue a decisive order to establish a majority view. The development adds further uncertainty to the debt resolution process involving Chandra.

ravynOS: Open-Source OS Built on Darwin and FreeBSD Enters Pre-Alpha

ravynOS is a new open-source operating system currently in pre-alpha stage of development. It is built upon Darwin, FreeBSD, and Apple's open-source components, combining established technologies into a new platform. The project has attracted early attention from the developer community, generating discussion on Hacker News. While details remain limited at this early stage, the OS is publicly accessible via its official website. The release signals a growing interest in alternative Unix-like operating systems drawing from Apple's open-source ecosystem.

Security Analysis Flags Critical Flash Loan Vulnerabilities in Steakhouse Financial Protocol

A security research report dated August 31, 2026, identified multiple high-severity flash loan attack vectors in Steakhouse Financial, a multi-chain yield optimization platform with approximately $3 billion in total value locked. The most critical finding involves the potential for a governance takeover using flash-loan-funded voting power, which could theoretically allow a full protocol drain. Researchers also flagged oracle price manipulation as a high-risk vector, estimating losses of up to 15% of TVL in a single epoch if an attacker moves over $200 million in liquidity on a single AMM. Additional vulnerabilities include a re-entrancy flaw in the vault's deposit and withdrawal logic, where balances are updated after external calls rather than before. The report assigned an overall risk score of 7.4 out of 10 and called for immediate mitigations to the oracle and vault modules, with governance hardening recommended in the next upgrade cycle.

Farm.js compiler bypasses React rendering by writing state changes directly to DOM

Developer Farm.js is an experimental full-stack framework featuring an ahead-of-time compiler for React components. When the compiler can statically prove at build time which DOM nodes a given piece of state affects, it skips React's render cycle and applies updates directly to the DOM via dependency bindings. Components where this cannot be proven fall back to standard React behavior, requiring no new syntax or mental model from developers. The compiled output for direct-binding-only components is reported to be 73.6% smaller than shipping the full runtime, with unused feature modules tree-shaken from production bundles. The framework is open-source on GitHub and can be explored interactively through a browser-based playground on StackBlitz.

Misconfigured CDN rule silently blocked OpenAI and Anthropic SDKs for all developers

A bot-blocking rule deployed at the CDN edge inadvertently matched the User-Agent strings of official OpenAI and Anthropic SDKs, returning 403 errors before requests ever reached the application. Browser-based human traffic was unaffected, making the issue invisible to casual monitoring, while developers using the official SDKs and AI crawlers trying to access llms.txt assets were silently blocked. The problem went undetected because the internal test suite used its own allowed User-Agent, meaning it was structurally incapable of catching the block. The issue was discovered by accident when a developer cross-checked a streaming result using the official SDK and received a 403 response. A targeted fix to the Cloudflare rule — with no application code changes — resolved the problem, restoring access for all seven tested SDK User-Agents and five AI crawler agents as confirmed in a retest on 2026-08-31.

Zimbabwe beat Namibia by 5 runs in tense Africa T20 Cup thriller

Zimbabwe narrowly defeated Namibia by five runs in a gripping Africa T20 Cup encounter. Namibia were chasing a target of 196 but ultimately fell short of the mark. A remarkable seventh-wicket partnership of 110 runs between Green and Busing-Volschenk gave Namibia a real chance of pulling off the upset. However, Zimbabwe held their nerve in the final stages to secure a hard-fought victory.

Civic Tech Fails Citizens When Feedback Loops Are Hidden, Developers Argue

A commentary from DEV Community argues that most municipal software projects fail not due to poor code performance, but because citizens receive no meaningful updates after submitting reports. The author contends that transparency should be treated as a core architectural requirement rather than an optional feature added late in development. To address this, the piece proposes event-driven notification systems that map directly to real-world municipal workflows, including public audit logs triggered whenever an inspector updates a ticket's status. A code example illustrates how backend status changes can be published to a public event stream, making municipal workers' actions visible and potentially reducing bureaucratic delays. The author concludes that the true measure of civic software success is not uptime or throughput, but whether residents feel their input has genuinely reached the people responsible for their city.

spfx-factory Update Adds Evidence Ledger to Track SPFx Project Readiness

The latest release of spfx-factory, an open-source SharePoint Framework project generator, introduces an evidence ledger that records the outcome of key checks during the generation workflow. Six evidence categories — scaffold, codex, policy, local, assurance, and tenant — each report a status of verified, warning, blocked, or not-tested. The update adds a parallel evidence field to preserve backward compatibility with the existing factory report contract. A new inspect command allows teams to review recorded evidence without re-running generation, and an optional flag lets teams block builds on warnings. The release explicitly notes that a passing local report does not confirm SharePoint permissions, App Catalog approval, or tenant deployment.

Centre defers Census in Manipur following ‘NRC first’ demand by civil society groups

Lionel Messi Retires from International Football at 39 After Two-Decade Career

Lionel Messi has announced his retirement from international football at the age of 39, bringing an end to an era spanning nearly two decades with the Argentine national team. During his tenure, Messi led Argentina to victory in the 2021 Copa America and the historic 2022 FIFA World Cup. He also guided the team to the final of the 2026 World Cup before stepping away from international duty. His exit marks an emotional conclusion to one of the most celebrated careers in football history.

1,800 Volunteers Relocate 400-Ton Japanese Castle Keep Using Traditional Methods

Around 1,800 volunteers in Japan successfully moved a historic castle keep weighing approximately 400 tons using traditional techniques. The relocation marks a significant milestone in a restoration project that has been underway for over a decade. The castle keep has been repositioned closer to its original stone foundation as part of the ongoing work. The process uncovered archaeological finds beneath the structure, including ancient drainage systems and remnants of earlier walls. The castle is expected to be fully reopened to visitors by 2033.

India's services exports rise 13.4% in July

Developer launches Corporate Mind Games, a sarcastic office-themed puzzle site

A developer and regular player of NYT and LinkedIn Games has launched Corporate Mind Games, a collection of logic puzzles built as a tongue-in-cheek alternative to LinkedIn Games. The puzzles are designed with a sarcastic corporate theme, aiming to be both challenging and humorous. The creator originally planned to announce the project on Hacker News the previous week but was sidetracked when an older game, Don't Wordle, unexpectedly hit the platform's front page. The launch was deliberately timed for a Monday morning to match the workplace theme of the puzzles.

VLC Media Player Hits 7 Billion Downloads Worldwide

VLC, the free and open-source media player, has surpassed 7 billion downloads globally. The milestone highlights the software's enduring popularity despite the rise of paid streaming services. Unlike subscription-based platforms, VLC operates entirely offline and at no cost to users. The achievement underscores continued demand for free, locally-run media playback solutions.

IRC-A Protocol Aims to Bring Zero-Trust Security and Dynamic Discovery to Enterprise AI Agents

A software architect with experience at Citibank and Bloomberg has developed IRC-A, a protocol and gateway architecture designed to address critical security and scalability gaps in enterprise multi-agent AI systems. The system tackles four core production challenges: dynamic agent discovery without hardcoded URLs, channel-level data sovereignty enforcement, concurrent LLM call management, and distributed audit trails. IRC-A was demonstrated through a reference implementation called the Dr. Cureta Healthcare Fleet, deployed on Google Cloud Run, which enforces strict department-level channel isolation to prevent unauthorized data access across medical records. The fortified version of the system upgrades agent identity verification to Ed25519 challenge-response tokens, adds semantic confidence thresholds for routing, and introduces a dual-LLM fallback mechanism between OpenAI and Google Gemini. The architecture is positioned as an alternative to centralized orchestration frameworks like LangGraph and CrewAI, which the author argues create brittle, monolithic dependencies in distributed agent deployments.