Tata Steel asks UK for more cash as Port Talbot slips

Step-by-Step Guide: AWS IoT Fleet Provisioning with CSR on ESP32-S3

Fleet provisioning is a secure IoT onboarding method where a device generates its own private key locally and submits a Certificate Signing Request (CSR) to AWS IoT Core to receive a unique X.509 certificate on first connection. The private key never leaves the device, making it suitable for hardware with secure elements and custom PKI setups. A developer has published a detailed walkthrough using an ESP32-S3-WROOM-1 module running on Ubuntu with ESP-IDF v5.5.5 and the esp-aws-iot SDK. The guide covers setting up a provisioning policy in AWS IoT, generating claim certificates, building the firmware, and configuring a provisioning template to register and activate devices. The approach supports both AWS-managed and self-managed certificate authorities, offering flexibility in certificate validity and signing algorithms.

MongoDB Indexes Explained: B-Trees, ESR Rule, and Query Optimization Tips

A technical guide published on DEV Community revisits how MongoDB uses B-Tree data structures to power collection indexes, enabling logarithmic O(log N) query lookups instead of full document scans. Every MongoDB collection automatically receives a unique index on the _id field, while additional indexes must be designed carefully using the Equality-Sort-Range (ESR) rule to avoid inefficient in-memory sorting. Developers are advised to run explain('executionStats') to detect slow COLLSCAN operations and monitor the ratio of documents examined versus documents returned. The guide also covers compound, multikey, text, and TTL index types, mapping each to specific query patterns. A key trade-off highlighted is that indexes consume memory and slow down write operations, so write-heavy workloads such as IoT telemetry should use as few indexes as possible.

Designer argues AI-generated event posters can achieve quality results

A designer named John Hartnup published a blog post on June 7, 2026, challenging the notion that AI-generated posters are inevitably low quality. The article, shared on Hacker News, argues that thoughtful use of AI tools can produce visually appealing event posters. The post attracted modest engagement, earning 13 points and 5 comments on the platform. Hartnup appears to advocate for a more nuanced view of AI in graphic design, suggesting outcomes depend heavily on how the technology is applied.

Fix macOS Stage Manager Full-Screen Gap Caused by External Keyboard Fn Key Limits

Mac users running Stage Manager often find that window-maximizing tools like Rectangle leave a 15–20% strip on the left side of the screen reserved for app thumbnails. The built-in macOS 'Fill' command solves this by stretching a window edge-to-edge while keeping Stage Manager active, but its default shortcut Fn+Control+F fails on third-party keyboards because those devices process the Fn key internally and never pass it to macOS. Users can work around this by creating a custom keyboard shortcut — such as Control+Option+F — mapped to the menu item 'Fill' under System Settings > Keyboard > App Shortcuts. Logitech keyboard users should also switch their device to Mac mode by holding Fn+O for three seconds, ensuring modifier keys map correctly to Apple's layout. Together, these two steps restore full-screen functionality without requiring an Apple Magic Keyboard.

Why AI Coding Agents Should Leave Replay Scripts, Not Just Chat Logs

A backend engineer merged a fix generated by an AI coding agent, but reviewers found the build failed on a clean machine with no trace of the original chat session. The core problem is that chat transcripts record intent rather than reproducible steps, meaning work validated inside one AI session may not survive outside it. To address this, a structured 'spike' ritual is proposed where the agent must produce a replay script — a shell script that restores a clean environment, applies only committed changes, and exits successfully on any machine without agent involvement. The spike runs within a fixed time budget, starts with a written hypothesis and a pre-existing failing test, and is killed outright if the replay script is missing or unrunnable elsewhere. The framework treats reproducibility, not conversational confidence, as the true measure of whether engineering work is done.

Western Australia issues rare shark kill order after fatal Perth attack

Authorities in Western Australia have issued an unusual order to kill a shark following a deadly attack on a swimmer near Perth. Police boats and helicopters have been deployed to patrol the beaches in the area. The search is ongoing for both the victim's body and the shark responsible for the attack. Such kill orders are considered rare in the region, highlighting the severity of the incident.

Inside Disney World's Golden Oak: Luxury Homes Priced at $20 Million

Golden Oak is an exclusive residential community located within the Walt Disney World resort in Florida, where homes can cost up to $20 million. The properties feature Disney-themed designs, with at least one residence drawing inspiration from Belle's cottage in Beauty and the Beast. Residents enjoy direct access to Disney's theme parks, allowing for spontaneous visits without the typical logistical challenges. The community is designed to blend luxury home living with the immersive Disney experience, offering a high level of privacy alongside resort-style amenities.

AI cloud firm Nscale files for US IPO

Sehrawat replaces Toor as India's flag-bearer at Olympics opening ceremony

Wrestler Sehrawat was named India's co-flag-bearer at the Paris Olympics opening ceremony following a last-minute change. He replaced shot-putter Tajinderpal Singh Toor, who was ruled out due to technical issues. Sehrawat carried the flag alongside double Olympic medallist Manu Bhaker. The 2022 Asian Games gold-medal-winning captain revealed the honour left him too excited to sleep the night before the ceremony.

Delhi Airport's Third Runway Set to Resume Operations on September 20

Delhi International Airport is set to reopen its third runway for operations starting September 20. The runway had been taken out of service to undergo upgrades and improvements. Airport operator confirmed that all necessary inspections, validations, and regulatory clearances have been successfully completed. The resumption of the third runway is expected to enhance the airport's overall operational capacity.

Why Comparing AI Agent Pass Rates Without Token Budgets Is Misleading

A software engineering analysis argues that pass rates alone are an unreliable benchmark for evaluating AI coding agents when the token and tool-call budgets differ between runs. An agent that retries repeatedly until hitting a cap may outperform one given a single attempt, yet both results get averaged as if conditions were equal. The proposed protocol requires declaring a fixed budget before any model call, logging every attempt in a JSONL ledger, and publishing three metrics together: pass rate, truncation rate, and tokens-per-success. Task files must be hashed with SHA-256 so any fixture change invalidates the run ID, ensuring reproducibility. The author frames this as a methodological standard rather than a vendor comparison, noting that omitting budget data turns an engineering review into what they call 'a vibe with a percentage sign.'

Developer builds AI-integrated platform where models share app architecture

A software developer has been experimenting with embedding AI directly into application architecture, rather than treating it as a standalone chatbot feature. An early project involved an AI-assisted chess database, where users could search games, add commentary, and configure the app through conversational AI assistants. This experiment demonstrated that AI could serve as a core interface for operating an application, not merely an add-on layer. The developer has since expanded the concept into a more general platform where AI models, databases, and application components share a unified architecture. In this evolving system, the application manages context, exposes functions to the models, and retains control over the overall workflow.

Codpal open-sources Deckgauge, a self-hosted engineering metrics dashboard

Yaniv, founder of CTO-as-a-service firm Codpal, has open-sourced Deckgauge, an engineering analytics tool his team built to rapidly assess how large engineering departments operate. The platform integrates data from Jira, GitHub, GitLab, and Azure DevOps into a single dashboard, surfacing DORA metrics, team focus, per-engineer workload, and CapEx vs OpEx breakdowns. Notably, the tool was largely developed by AI agents using Claude Code with human review at every merge. Deckgauge is fully self-hosted via a single Docker deployment, meaning no data leaves a company's own infrastructure. It is available under the FSL-1.1 license, which permits free use for non-competing purposes and is set to convert to Apache-2.0 in September 2028.

Developer Builds AI Agent to Automate Real Estate Comps, Shares What Works

A real estate investor and developer documented their hands-on experience building an AI agent to automate property comparable analysis, a process traditionally requiring hours of manual spreadsheet work. Early attempts using a basic LangChain agent with web scraping tools repeatedly failed with silent timeouts and no useful output. Switching to LangGraph provided more structured control by defining explicit states for data retrieval, filtering, adjustment calculation, and report generation. The developer also experimented with CrewAI's multi-agent setup, separating data analysis and property valuation roles, though this introduced additional failure points. Key lessons included building robust edge-case handling, structured JSON outputs, and pre-filtering steps to reduce API token costs.

How LLMs Actually Process Text: Tokenization Explained for Developers

Large language models never process raw words — instead, text is first broken into fragments called tokens and converted into sequences of numbers before the model sees anything. A token is typically a piece of a word rather than a whole word, and a separate component called a tokenizer handles this conversion before any neural network processing occurs. Modern models use subword tokenization algorithms such as BPE or WordPiece, which balance vocabulary size and sequence length by splitting rare words into smaller reusable pieces. This design explains several practical AI engineering realities, including why APIs charge per token, why long prompts cost more, and why context windows have hard limits. Developers can explore tokenization hands-on using OpenAI's open-source tiktoken library to see how different types of text — including code, emojis, and non-Latin scripts — map to token counts.

Plugin4Shell: AI Coding Agents Ignore SHA Pins, Exposing Millions to Supply Chain Attack

Security researchers at AIR have disclosed a vulnerability called Plugin4Shell, which they describe as the first supply chain vulnerability in the AI agent ecosystem, affecting Claude Code, Codex, GitHub Copilot, and Gemini CLI. The flaw allows attackers to substitute malicious code in place of a pinned commit SHA, because none of these agents verify that the checked-out code actually matches the requested commit. The attack requires zero user interaction and can be triggered silently via auto-update features enabled by default in Claude Code and Codex. Anthropic and OpenAI have issued patches, but Microsoft has yet to release a fix for Copilot — used by nearly 90% of Fortune 500 companies — while Google has deprecated Gemini CLI entirely without patching it. Users of unpatched tools remain at risk of attacker-controlled code executing with elevated agent permissions until vendors ship verified checkout logic.

The Best VPN Services of 2026, Tested and Ranked by Experts

VPN services have become a crowded market, with numerous providers claiming to offer the best privacy and performance. Wired tested and reviewed multiple VPN options to help consumers identify which services genuinely deliver on their promises. The evaluation focused on key factors that distinguish reliable VPNs from those that fall short. Six services stood out as top performers based on the testing criteria applied. The review aims to cut through marketing claims and provide users with trustworthy, evidence-based recommendations.

WIRED Hosts One-Day World Fair Event in Miami on November 4

WIRED magazine is organizing a live event called the WIRED World Fair, scheduled for November 4 in Miami. The single-day event aims to bring together the technologies, people, and cultural forces that are shaping the modern world. The fair is designed to translate WIRED's editorial themes into a real-life, in-person experience. It represents an opportunity for attendees to engage directly with the ideas and innovations typically covered in the publication.

Dear Customer, Fuck You

Article URL: https://fuck-off.ai Comments URL: https://news.ycombinator.com/item?id=49764686 Points: 14 # Comments: 1