'It was for Bimstec chair, not PM': Dhaka declines Brics invite

Rust developer builds social preview cards without headless Chrome using resvg

A developer building Cloud Cost Analyzer (CCA) needed shareable links to unfurl into rich social cards on Slack, LinkedIn, and X, but the app is a client-side single-page app that social crawlers cannot parse. To solve the crawler problem, a CloudFront function detects bot user-agents and redirects them to a lightweight server-rendered page carrying Open Graph meta tags, while human visitors continue to receive the full interactive dashboard. For the card image itself, the developer chose to avoid headless Chrome due to its memory, cold-start, and maintenance overhead in a Rust backend. Instead, cards are generated by building an SVG string and rasterizing it in-process using the pure-Rust libraries resvg and tiny-skia. Fonts are compiled directly into the binary via include_bytes!, ensuring consistent rendering across local and containerized environments without any external font dependencies.

EU AI Act Sets Fines Up to 35M EUR or 7% of Turnover From August 2026

The EU AI Act introduces a three-tier penalty system under Article 99, with fines ranging from 7.5 million EUR to 35 million EUR or 1% to 7% of global annual turnover, depending on the nature of the violation. The steepest penalties apply to prohibited AI practices, while lower tiers cover high-risk system obligations, transparency failures, and the supply of misleading information to authorities. For each violation, companies face whichever is higher — the fixed euro amount or the turnover-based percentage. The European AI Office is set to begin enforcement on August 2, 2026, initially using a graduated approach involving information requests and corrective orders before imposing financial penalties. However, intentional non-compliance or failure to cooperate could trigger direct penalty proceedings, with factors such as repeat violations and evidence concealment able to increase the final fine amount.

1 in 9 YouTube Videos Lacks maxresdefault.jpg, and All Missing Files Return HTTP 404

A developer tested thumbnail availability for 8,664 live YouTube videos on 7 September 2026, finding that 11.7% had no maxresdefault.jpg file — roughly one in nine, even in a sample skewed toward established channels. The three smallest thumbnail sizes were universally present, but availability of the 1280x720 formats dropped noticeably. Older videos are the primary cause: uploads from 2008–2009 had a 0% maxresdefault rate, while videos from 2023 onward had one over 90% of the time, and YouTube does not back-fill thumbnails for older content. A widely repeated claim that missing thumbnails return HTTP 200 with a grey placeholder was found to be false — all 1,014 missing files returned a proper HTTP 404 status. However, the 404 response body is itself a valid 120x90 grey JPEG, which causes browsers to silently render a small grey box and fire a load event rather than an error, misleading developers who do not explicitly check the HTTP status code.

Slack Launches AI Feature to Build Interactive Reports and Dashboards in Chats

Slack is rolling out a new feature called Slackforce Surfaces that lets users create interactive reports, polls, dashboards, and presentations directly within chats. Users can describe their needs to Slackbot, which uses AI to pull relevant information from conversations and connected apps such as Google Drive and Salesforce. The generated Surfaces can be shared with colleagues, pinned to channels, and opened for comments and interaction. The feature was demonstrated with an example of a user requesting an arcade-themed visualization of AI token usage data.

Developer Finds 9 Bugs in His Own AI Eval Tool — All Made Results Look Better

A developer building a mutation-testing evaluation harness discovered nine separate bugs after completing roughly 40 hours of work. Every single bug skewed results in the same direction, making the tool's performance appear better than it actually was. The author argues this is a structural flaw in self-built evaluations: engineers tend to investigate disappointing results and fix those bugs, while positive results go unscrutinised. This selective debugging means flattering errors survive to publication not through dishonesty, but simply because they never trigger the instinct to investigate. The author warns this asymmetric auditing process is likely a widespread problem, not an individual failing.

Insufficient content to generate a reliable headline

The source provided contains no readable article text — only metadata such as a URL, a Hacker News comments link, a score of 5 points, and zero comments. Without the actual article content, it is not possible to accurately summarize the claims or arguments made. Fabricating details about Navier-Stokes or formal methods would risk spreading misinformation. Please provide the full article text for a proper summary.

Texas engineer collects 5,000 condiment packets, including 400 Taco Bell designs

Why Returning Pointers to Local C Structs Causes Segmentation Faults

In C programming, local variables declared inside a function are allocated on the stack and destroyed the moment that function returns. If a pointer to such a local struct is returned or misused across function boundaries, it becomes a dangling pointer referencing invalid memory. Accessing this pointer triggers a segmentation fault, a common pitfall for developers transitioning from memory-managed languages like JavaScript. The safe alternative is to allocate the struct in the caller function and pass its address down to any function that needs to modify it, keeping the memory lifecycle tied to the caller's scope. This pattern ensures crash-free execution and is considered standard practice in systems programming.

Developer cuts LLM token usage 42x with deterministic code-edit pipeline D-Engine

Software developer Sergi Corruchaga published findings in September 2026 showing his open-source tool D-Engine completed the same coding task using 2,552 tokens, compared to over 107,000 tokens consumed by DeepSeek's official agent under default settings. Corruchaga attributed the gap not to the underlying model or thinking mode, but to architectural differences in how each tool manages context. Conventional agentic coding tools re-send the full conversation history on every loop turn, causing token costs to grow quadratically with the number of steps taken. D-Engine separates responsibilities by having the LLM only generate structured patch blocks, while a local deterministic runtime applies changes to an isolated copy of the repository and runs compiler checks before merging. The tool is available under the MIT licence as D-Engine v0.2.2.

PM Modi: We invite France and the world to be co-travellers in India’s next space frontier

DNS Lookup Per Request, Not Pingora, Caused 6x Slowdown in Proxy Benchmark

A developer benchmarking Cloudflare's Rust proxy library Pingora against nginx in a containerized 2-vCPU environment initially recorded a six-fold performance gap, with Pingora handling only 21,000 requests per second versus nginx's 126,000. Investigation revealed the culprit was a DNS resolution call — getaddrinfo — being triggered synchronously on every single request inside a Tokio worker thread. In containers, this meant each request fired a live DNS query to Docker's resolver, blocking the thread and inflating latency from 0.45 ms to 4.78 ms. Resolving the upstream address once at startup and reusing the SocketAddr brought Pingora's throughput up to 89,000 requests per second. The developer concluded that the true performance difference between Pingora and nginx in this setup is just 1.49x, and warned that blocking operations hidden on the hot path can easily be misattributed to the framework itself.

Djibouti to Become 72nd Nation to Sign NASA's Artemis Accords

The Republic of Djibouti is set to sign the Artemis Accords on Monday, September 14, at 11 a.m. EDT at NASA Headquarters in Washington. The signing will make Djibouti the 72nd country to join the international space cooperation agreement. NASA Deputy Administrator Matt Anderson will host the ceremony alongside Djibouti's Ambassador to the United States, Mohamed Siad Douale. Representatives from the U.S. State Department are also expected to participate in the event.

Amid West Asia conflict, Iran president to meet PM Modi today

How to Build a Crash-Style Browser Demo With Clean State Design

A developer tutorial on DEV Community outlines best practices for building an educational crash-style multiplier interface in the browser. The guide emphasizes defining an explicit state model — separating game phase, multiplier value, and timing — rather than inferring state from animation frames, which prevents common bugs. It also covers responsible history display, advising against labels like 'hot' or 'due' that could mislead users into thinking past rounds predict future outcomes. The article stresses that virtual credits must be clearly labeled as non-monetary on the first screen, with disclosures placed near the balance rather than buried in fine print. Accessibility is also addressed, including keyboard navigation, stable button labels, and limiting screen reader announcements to meaningful state transitions rather than every animation frame.

Developer builds YouTube video downloader using Python and yt-dlp library

A developer has shared a Python-based project that allows users to search and download YouTube videos directly from the command line. The tool uses yt-dlp to handle YouTube searches and downloads, while FFmpeg merges separately streamed video and audio into a single MP4 file. Users can specify a search query and choose how many results to download, with the program automatically saving files to a designated folder. The project relies on Python's pathlib module for file path management alongside yt-dlp and FFmpeg as its core dependencies. The developer noted that users must ensure they only download content they are permitted to access and must comply with YouTube's Terms of Service and copyright laws.

ECC: Open-Source OS Framework Brings Structured Workflows to AI Coding Agents

Developer affaan-m has released ECC (Everything Claude Code), an open-source performance harness designed to give AI coding assistants like Claude Code and Codex a structured engineering environment. Rather than relying on single-turn prompting, ECC enforces a defined lifecycle — plan, test, implement, review, verify, and improve — directly within the terminal agent's runtime. The framework bundles pre-configured roles for architecture planning, test-driven development, self-review, and domain-specific tooling covering frontend, backend, and DevOps tasks. A built-in security module called AgentShield audits prompt inputs, MCP configurations, and credentials to guard against vulnerabilities as agents gain broader system access. ECC also manages context window efficiency by retaining key architectural lessons and project conventions across sessions, and can be installed via a guided setup using npx or Claude Code's plugin manager.

OpenAI Temporarily Pauses Pro Subscription Sign-Ups Amid Surging Demand

OpenAI has temporarily halted new sign-ups for its Pro subscription tier due to overwhelming demand. The company cited Pro subscriptions as placing the greatest strain on its infrastructure. OpenAI stated it is pausing onboarding while it works to expand its system capacity. The move is intended to ensure service quality for existing subscribers during the high-demand period.

Model Context Protocol Offers a Unified Standard for Connecting AI to Domain Systems

The Model Context Protocol (MCP), introduced by Anthropic in November 2024, is an open standard designed to give large language models controlled access to external data sources and tools. Before MCP, every AI application required its own custom integration with domain systems, creating a fragmented tangle of duplicate tool descriptions, varied authentication methods, and hard-to-reuse security logic. MCP addresses this by defining a shared host-client-server architecture using JSON-RPC 2.0, enabling AI hosts such as chat clients, IDEs, and agent tools to connect to domain systems through a single protocol. A developer documenting their own MCP server build — built atop a custom ticket system — illustrates how the standard allows an AI to query current, permission-scoped data rather than relying on guesswork or copied context. The article, first in a series, focuses on foundational concepts, with a follow-up planned to cover interface design and security considerations for a responsible MCP server implementation.

Anthropic Report Accuses Chinese AI Firms of Persistent Model Distillation Attacks

Anthropic released a report on Thursday alleging that China-based AI companies have been conducting repeated distillation attacks against its models. The companies named in the report include Alibaba, Moonshot AI, and DeepSeek. Distillation attacks involve extracting knowledge from a proprietary AI model to train a competing one. Anthropic says these attacks have grown more frequent as competition in the AI industry has intensified in recent months.