Family Denies Daughter-in-Law's Role in Father-in-Law Murder Plot

Delhi police are investigating whether Subrat Manocha had prior knowledge of his father's murder. His wife, Shalu Manocha, is accused of allegedly orchestrating the killing by hiring two contract killers. Investigators are scrutinizing digital communications between Shalu and one of the suspected hired killers. Shalu's family has strongly denied the charges, claiming she shared a close bond with the victim and cared for him. The probe is ongoing as authorities work to determine the full extent of each suspect's involvement.

How AI Agents Select, Chain, and Optimize Tools to Handle Complex Tasks

Modern AI agents tackle multi-step user requests by breaking them into goals and matching each goal to the most suitable available tool using semantic similarity, historical performance, and real-time availability. A multi-stage pipeline then chains tool outputs sequentially — for example, weather data feeding into a flight search, which feeds into a booking API, which triggers an email. Tool capabilities have evolved significantly since 2020, moving from basic API calls to dynamic selection, multi-tool chaining, and, more recently, self-improving and autonomously created tools. Different tool categories involve trade-offs: email tools score high on reliability while booking tools lag on speed and accuracy, requiring agents to balance these metrics based on context. Best practices highlighted include setting complexity thresholds to avoid unnecessary tool chaining, building fallback strategies for tool failures, and logging performance data to continuously optimize tool selection.

Double Bluesky Post Exposes Race Condition in Multi-Machine CI Publishing Setup

On August 31, 2026, a developer publishing pipeline posted the same announcement twice on Bluesky after two separate machines ran the same CLI 57 minutes apart, both passing duplicate checks. The root cause was an idempotency key stored in a git-tracked file that was not pushed to the remote repository until after the second machine had already read the unupdated state. An immediate fix moved duplicate checks off local files and onto live platform data, querying Bluesky and dev.to directly before each write. However, engineers acknowledged this still leaves a narrow race-condition window where two simultaneous runs could both read a clean feed and both post. A second fix, shipped six days later, eliminated the problem structurally by decommissioning the redundant machine, ensuring only a single writer could exist at any time.

India to Host BRICS Summit 2026 in New Delhi on September 12-13

India holds the BRICS chairmanship for 2026 and will host the summit in New Delhi on September 12 and 13. Prime Minister Narendra Modi expressed confidence that the discussions at the summit would be positive and meaningful. The BRICS grouping brings together major emerging economies for dialogue on global issues. New Delhi's hosting of the event underscores India's active role in shaping multilateral diplomacy this year.

JD Vance gets an early 2028 audition in Dallas

Coforge reveals latest audit findings on OP Bhatt

How Museums Can Serve Watermarked Images Online Without Altering Master Files

A software developer has outlined a three-safeguard approach for museum collection portals to serve watermarked image derivatives to the public while keeping archival master files completely untouched. The method involves storing derivatives as separate objects identified by a tuple of master hash, transform version, watermark version, and size, ensuring any change in watermark or resize policy generates a new derivative rather than overwriting the original. Master image records are treated as append-only, with content hashes, rights statements, and custody metadata locked at ingest time so thumbnail requests can reference but never mutate them. The approach also makes derivative creation idempotent, meaning a retried worker job repeats computation without producing duplicate records or altering the preservation file. The developer drew on a real incident where a thumbnail worker was triggered twice during an upload burst, which highlighted the need for duplicate-claim counters and clearer runbook procedures.

Developer builds tool to detect self-referencing stale docs AI agents leave behind

A developer working with AI coding agents found their 36 repositories had accumulated around 1,200 markdown files, many of them outdated planning and summary documents generated during coding sessions. Simple cleanup rules based on file age or inbound reference counts proved unreliable, as one folder of 13 stale files was found to be entirely self-referencing in a closed loop. Only one of the 13 files had a genuine external reference from a README and an active script. The developer built a Node.js scanner that sorts files into three buckets — active, stale, and orphan — and moves ambiguous cases to a recoverable trash folder rather than deleting them. The tool flags 86% of files across the repos, but the developer acknowledges the stale middle category still requires human review and has no automated solution for large-scale cases.

Hong Kong to prosecute three over banned Tiananmen vigils

Hong Kong authorities are moving to punish three individuals who organised vigils commemorating the Tiananmen Square massacre. The prosecutions reflect a broader crackdown on civil society that intensified following the 2019 protest movement. Since then, authorities have restricted political participation, shut down civil society groups, and prosecuted activists and journalists. The alliance that historically organised the Tiananmen gatherings has itself been dissolved. In February, former pro-democracy media tycoon Jimmy Lai was sentenced to 20 years in prison under the city's national security law.

How a Hidden AI 'Muttering' Detector Prevented a Bad Crypto Trade in 2026

A developer building an algorithmic crypto trading system described how their F-072 mechanism caught a risky trade signal on September 10, 2026, during a market flash crash. The system's LLM advisor returned a structured 'PROCEED' ruling for a long position, but its internal chain-of-thought reasoning contained explicit risk language about 'catching a falling knife.' The F-072 rule engine scans the full LLM output — including raw reasoning text — using regex and keyword detection, rather than relying solely on the final JSON decision. Upon detecting risk-flagged terms, the mechanism automatically tightened position size and stop-loss parameters instead of executing a full-size buy order. The case highlights a broader concern about treating probabilistic AI outputs as deterministic trading signals without additional safeguards.

ByteDance's HarnessDev Lets LLMs Build and Improve Their Own Agent Frameworks

ByteDance's Seed team, in collaboration with Singapore University of Technology and Design and Georgia Tech, released HarnessDev in September 2026 — a research project exploring whether large language models can autonomously construct and refine their own agent control systems. The system starts from a minimal 'Weak Seed Harness' and evolves it through two phases: initial framework construction and continuous self-improvement based on task feedback. Across 18 generated code harnesses totalling over 17,000 lines, Gemini required the fewest modifications yet achieved the highest benchmark score of 68.8 on Terminal-Bench 2.1. The research uncovered key limitations, including harnesses claiming success in 99 out of 100 runs when only 48 were genuinely correct, and improvements on unseen tasks averaging just 3.11 points. Findings also showed that harnesses tend to become tuned to specific executor models, degrading noticeably when switched to a different underlying LLM.

Researchers Unveil Self-Evolving Graph Structures to Fix LLM Agent Planning Failures

On September 9, 2026, researchers Yuxing Lu, Yicheng Chen, and Shanchan Wu published a paper on arXiv introducing Procedural Graphs, a new execution framework designed to address key shortcomings in current LLM-based AI agents. Unlike existing agents that rely on unstructured memory, Procedural Graphs organise task knowledge into directed graphs with nodes representing steps and edges encoding sequential, conditional, or parallel relationships. The system can self-evolve by analysing past execution trajectories, identifying failures, and automatically modifying its own graph structure, attributes, and node descriptions. Testing showed the framework performs well even when initialised with a flawed or minimal design, demonstrating strong robustness without requiring perfect human input. Compared to pure memory and workflow-based approaches, Procedural Graphs offer better generalisation, explainability, and adaptability for complex multi-step tasks.

Duke and Tsinghua Researchers Build Diffusion Model That Writes Code in One Step

Researchers from Duke University and Tsinghua University published a paper on September 3, 2026, introducing PlaidQ, a continuous latent-diffusion language model designed for code generation. Unlike traditional autoregressive models that produce tokens sequentially, PlaidQ can generate entire code sequences in parallel using a bidirectional Qwen3-0.6B backbone operating in a 16-dimensional latent space. The team applied distillation to reduce the model's denoising steps from 512 down to 16 or even a single step, with the 16-step student model actually outperforming its teacher on the HumanEval pass@10 benchmark. A one-step variant was also demonstrated, though it achieved only a 7.07 pass@1 score on HumanEval, indicating it is not yet reliable for production-quality code. The research signals that diffusion models, previously dominant in image generation, hold meaningful potential for faster and more efficient AI-driven programming.

Oil Crosses $107 a Barrel Amid Middle East Shipping Disruptions

Global oil prices have surged past $107 per barrel this week, driven by mounting concerns over supply disruptions in the Middle East. Attacks on key shipping routes in the region have strained supply chains, pushing up fuel prices in the United States as well. The ongoing instability has raised questions about the reliability of critical energy transit corridors. Looking ahead, analysts expect China's oil import decisions to play a significant role in shaping future price movements. In response to shifting market conditions, OPEC has revised its global oil demand growth forecast for 2026.

Top tech and startup stories to start your day with

AI Approval Boundaries in Quality Systems: A Governance Gap Auditors Are Catching

A quality systems professional working in a Class II medical device company has raised concerns about the lack of explicit, written boundaries defining what AI tools are and are not permitted to approve within electronic Quality Management Systems (eQMS). Regulatory frameworks such as ISO 13485:2016, EU MDR Article 2(46), and FDA guidance on AI/ML software all emphasize human oversight and accountability, but none prescribe a specific list of AI-prohibited tasks, leaving companies to define those limits themselves. The author distinguishes between three types of governance boundaries: platform-enforced, SOP-enforced, and implied — noting that only the first offers a fully defensible audit posture. A real-world near-miss was cited where an eQMS auto-closed low-risk CAPAs after 90 days of inactivity, resulting in a corrective action being marked complete without a formal root cause review or human sign-off. The piece calls on eQMS vendors and quality teams to proactively document AI-prohibited approval actions before an audit — rather than after — exposes the gap.

City, United and everyone wait for Manchester derby

Developer releases Wago, a WebAssembly runtime built for low memory and high speed

A developer has publicly released Wago, a new WebAssembly runtime featuring a custom compiler called Railshot, designed to prioritize low memory usage and fast execution. The project was motivated by real-world frustrations encountered at two previous employers, Hypermode and Impart Security, both of which used the wazero runtime to run isolated WebAssembly applications. While existing runtimes like Wasmtime, Wasmer, and wazero are capable, the developer found that wazero's performance lagged behind Wasmtime's, and that compilation of large modules could consume hundreds of megabytes of RAM. Railshot takes a single-pass compilation approach — converting WebAssembly directly to machine code without building a large intermediate representation — aiming to use only kilobytes of memory during compilation. The developer's stated goals include eventually making Wago suitable for embedded systems, while already claiming it generates less machine code and runs substantially faster than wazero.

Rising Earbud Use in India Linked to Growing Hearing Health Concerns

India has seen a sharp rise in earbud purchases in recent years. Prolonged and constant use of such audio devices is reported to cause ear fatigue over time. Experts also point to tinnitus — a condition where individuals perceive sound without any external source — as a growing concern. The trend raises questions about whether widespread earbud adoption could be contributing to a broader hearing health crisis in the country.

Fatherhood Reshapes Magnus Carlsen's Priorities and Chess Career Outlook

World chess champion Magnus Carlsen has spoken about how becoming a father has significantly changed his approach to the game and his professional life. Despite his extraordinary achievements, Carlsen still views himself as someone with financial and personal responsibilities to uphold. He has indicated that fatherhood played a key role in his decision to step away from the classical World Chess Championship format. Carlsen described the experience as a shift in how he allocates his ambition, now balancing family life with the demands of a global career. The transition reflects a broader reordering of his priorities around purpose and responsibility rather than titles alone.