SShortSingh.

Programming

0
ProgrammingDEV Community ·

Developer Launches KIRA, a Local-First AI Agent That Verifies Its Own Actions

A developer has released KIRA Superapp, an open-source AI agent runtime designed exclusively for Apple-silicon Macs that prioritizes on-device processing and verifiable task completion. Unlike conventional AI assistants, KIRA only marks a task as done when its runtime receives confirmed output from the tool that executed it, rather than relying on the model's generated text. The system's decision layer, Orchestrator V1, runs locally via MLX and is built to plan tool use rather than produce conversational responses. Local storage covers chat history, memory, logs, and downloaded models by default, while voice features are handled on-device through Whisper and Kokoro ONNX. The project's code and installation instructions are publicly available on GitHub, and the developer is actively seeking technical feedback on real-world failures and edge cases.

0
ProgrammingDEV Community ·

Developer Rebuilds Failed Social App as CrowdWide to Boost Visibility for Small Creators

A developer has launched CrowdWide, a new social platform built from scratch after abandoning an earlier project called VibeConnect due to security gaps and incomplete features. CrowdWide is designed around communities, creators, and content discovery, with a focus on giving smaller or newer accounts a fairer chance at initial exposure. The platform experiments with a feed distribution model that allocates 40% of reach to new or smaller communities, 55% to established ones, and 5% to trending content. It targets a broad range of creators including developers, writers, artists, bloggers, and YouTubers, and already includes features such as scheduled publishing, nested replies, community moderation tools, and privacy controls. The developer acknowledges the algorithm is still evolving and expects it to change significantly once real usage data becomes available.

0
ProgrammingDEV Community ·

Go Developer Builds Ethereum Wallet CLI to Demystify Crypto Key Management

A software developer built a command-line Ethereum wallet in Go to understand the real mechanics behind cryptocurrency wallets, documenting key findings in a technical post on DEV Community. The project revealed that wallets are less about storing keys and more about a series of precise cryptographic decisions, where small implementation errors can break security or compatibility. Following the BIP39 standard, the wallet generates a 12-word mnemonic by combining 128 bits of random entropy with a 4-bit SHA-256 checksum, split into 11-bit groups mapped to a 2,048-word list. The mnemonic then seeds a hierarchical deterministic wallet tree via BIP32, allowing thousands of keys to be derived from a single master key, with hardened derivation requiring a parent private key rather than just a public key. The developer emphasized the project is intended for learning purposes only and not for production use.

0
ProgrammingDEV Community ·

OpenCompany Open-Sources 22 AI-Run Business Templates, but Core AI Remains Proprietary

A GitHub project called tinyhumansai/opencompany, created in July 2026 and written in Rust under a GPL-3.0 license, offers 22 ready-to-run company templates designed to let AI agents handle most business operations. The project markets itself as an "operating layer for solo businesses," where agents manage tasks like accounting, marketing, legal drafting, and software development, while the human owner retains final decision-making authority. Each company type is defined in a plain-text company.toml file, making it easier to configure roles without writing code. However, the system's core orchestrator is a proprietary hosted AI model called Medulla, developed by TinyHumans, which requires a paid API key from tinyhumans.ai to function. This means that while the structural framework is open-source, the intelligence driving it remains a closed, external dependency.

0
ProgrammingDEV Community ·

Developer Builds Automated Brand Abuse and DMCA Takedown System Using n8n

A developer created a fully automated brand protection pipeline using the n8n workflow tool after a client discovered a cloned app on the Play Store that had gone unnoticed for weeks. The system uses three chained workflows to scan Google, Bing, and app stores for unauthorized brand usage, leveraging Monitorank for web search and DataForSEO for app store data. A separate workflow enriches each detection with domain registrar and hosting information via RDAP lookups. Once a live infringing page is confirmed, the pipeline automatically generates and sends legally structured DMCA takedown notices to the appropriate targets without any human intervention. The solution was built over a few weekends as a cost-effective alternative to hiring an IP law firm or subscribing to expensive brand protection software.

0
ProgrammingDEV Community ·

38% of Key Affiliate Program Terms Go Unpublished, Study of 252 SaaS Tools Finds

A researcher manually reviewed affiliate program pages for 252 developer and SaaS tools, recording five key terms for each, including commission rate, cookie window, and payout method. Of the 1,260 total data fields collected, 483 — roughly 38% — were not publicly disclosed by vendors. The most commonly hidden details were minimum payout thresholds (missing for 65% of programs), cookie windows (52%), and payout methods (45%), which the researcher notes are the terms that determine whether earnings are actually received. Cookie windows ranged from 14 days to lifetime attribution among those that disclosed them, while minimum payout thresholds varied from $5 to $450. The full dataset has been published at affiliateprogramterms.com, with source URLs and check dates included for each field to ensure verifiability.

0
ProgrammingDEV Community ·

TrueCharts Guide: Deploy AdGuard and V2Ray on Kubernetes via Helm Charts

Developers can self-host privacy and security tools like AdGuard Home and V2Ray on Kubernetes using TrueCharts, a curated catalog of Helm charts that simplifies complex deployments. AdGuard Home provides network-wide DNS-based ad filtering, while V2Ray enables encrypted proxy tunneling for secure access to internal resources. Both services are deployed into a dedicated Kubernetes namespace using custom Helm values files that configure persistence, ports, and service exposure. After installation, AdGuard can be accessed via port-forwarding for initial setup, and other cluster workloads can route DNS traffic through it using its internal service address. The setup requires a running Kubernetes cluster, Helm v3 or later, and basic familiarity with Kubernetes concepts such as namespaces and ingress.

0
ProgrammingDEV Community ·

Rehost, Refactor or Rewrite: How to Choose the Right Mainframe Modernisation Path

Mainframe modernisation is not a one-size-fits-all decision, with organisations typically choosing between rehosting, replatforming, refactoring, rearchitecting, replacing, or retiring legacy systems. Each strategy carries a different cost, timeline, and risk profile, making the choice highly dependent on specific business drivers such as cost reduction, agility needs, and available skills. A key pressure across UK enterprises is the shrinking pool of COBOL and mainframe expertise, alongside high operating costs and difficulty integrating legacy systems with modern cloud and AI platforms. Rehosting is the fastest and cheapest option but does little to address underlying code problems, while refactoring — often involving automated COBOL-to-modern-language conversion — is considered the sweet spot for most organisations with sound but hard-to-maintain logic. Most large modernisation programmes blend multiple strategies across their application estate rather than applying a single approach uniformly.

0
ProgrammingHacker News ·

AI Coding Tools Raise Questions About Ethics and Accountability in Software Development

The rise of AI-powered coding tools has made software development accessible to virtually anyone, prompting debate over who bears responsibility for what gets built. As barriers to entry fall, traditional gatekeepers like professional developers and engineering teams no longer serve as the sole filters for harmful or irresponsible software. The question of ethical oversight becomes more pressing when non-experts can deploy functional applications with minimal technical knowledge. Commentators are calling for clearer frameworks around accountability, consent, and harm prevention in an era of democratized software creation.

0
ProgrammingDEV Community ·

Hermes Agent Skill System Uses Trust Levels and Safety Scans to Block Risky Installs

Nous Research's Hermes agent platform includes a skill management system where skills are loaded as knowledge documents only when needed, using a progressive disclosure model to conserve token usage. Official documentation outlines three access levels, from listing skill names to loading full content or specific reference files, and skills automatically become slash commands upon installation. Every skill installed from the Hermes hub passes through a security scanner that checks for prompt injection, data theft, destructive commands, and supply chain threats. The system assigns trust levels based on a skill's source, ranging from built-in and official skills to third-party community registries, with dangerous-rated skills blocked even if a user attempts to force-override. A hash-based mechanism also protects user-modified skills from being silently overwritten during updates, requiring an explicit force flag to proceed.

0
ProgrammingDEV Community ·

LLM Model Routing Explained: How Teams Cut AI Costs by Up to 85% in 2026

As the number of viable large language models has grown from a handful to dozens, relying on a single LLM for every task has become both inefficient and costly. Model routing is a middleware layer that sits between an application and a pool of LLMs, directing each request to the most suitable model based on factors like task complexity, cost, latency, and safety requirements. Organizations that have adopted routing report cost reductions of 40–85% with no measurable drop in output quality, according to data from platforms including Requesty, OpenRouter, and InferenceHub. Three main architectural patterns have emerged: simple fallback chains, task-based classification routing, and hybrid approaches combining both. The shift is being driven partly by rising frontier model costs, with companies like Cognition — maker of the Devin AI agent — publicly acknowledging that single-model strategies are becoming financially unsustainable for engineering teams.

0
ProgrammingDEV Community ·

Hetzner Launches Free AI Inference API in Europe with Qwen 35B Model

German hosting firm Hetzner launched a free, experimental AI Inference API on July 24, 2026, allowing users to run large language model queries at no cost during the trial period. The service is OpenAI-compatible and hosted entirely on Hetzner's infrastructure in Germany and Finland, requiring only a Hetzner account and API token to access. It currently serves a single model, Alibaba's Qwen3.6-35B-A3B-FP8, a Mixture-of-Experts architecture with 35 billion total parameters, a 256K-token context window, and vision support. Hetzner has cautioned that the service carries no SLA or uptime guarantees and is not suitable for production use. The company has committed to notifying users by email before making any changes to the free-access status.

0
ProgrammingDEV Community ·

Weftgate Open-Source Tool Catches Broken Env Vars, Imports, and Routes in Python Code

Weftgate is an open-source static analysis tool designed to verify environment variable reads, dependency imports, and FastAPI route references against contracts defined in a repository. It runs locally via a CLI, MCP, hooks, and CI integrations, with no runtime dependencies or network calls on its default path. The tool flags issues such as misspelled environment variable names and missing dependency declarations, returning actionable suggestions or advisory review findings depending on how confidently it can resolve a reference. A GitHub Actions workflow configuration is available for pull request checks, and MCP clients can invoke the same gate functions without a separate server or API account. Developers are cautioned that a passing result does not confirm an application works correctly, and normal testing, type checking, and security reviews should continue alongside its use.

0
ProgrammingDEV Community ·

Wipro Redeployed 20,000-Employee Capacity Freed by AI Without Any Layoffs

Wipro's Chief Technology Officer Sandhya Arun announced that the company's AI programmes have generated productivity gains equivalent to the output of around 20,000 employees, with none of those staff being let go. The freed capacity has been redeployed internally across the company, which had a total workforce of approximately 243,000 as of June 2026. More than 100,000 Wipro employees are currently undergoing advanced AI training and certification as part of what the company calls a human-AI operating model. Under this model, engineers shift from direct task execution to supervising AI agents, while freed staff are directed toward new projects or client-facing roles. Wipro has not disclosed AI-specific revenue figures, and the company said its forward-deployment plans would be broadly in line with peers such as TCS and Infosys.

0
ProgrammingDEV Community ·

EKS Cluster Storage Cut After Audit Finds 200Gi Nodes Far Exceed Actual Usage

An infrastructure review of an Amazon EKS cluster found that worker nodes provisioned with 200Gi of root storage were consuming significantly less disk space than allocated under observed operating conditions. Engineers examined key storage paths including container images, kubelet data, and system logs to understand real consumption rather than relying on free-space metrics alone. The investigation was part of an ongoing series of capacity optimizations that had previously addressed CPU bottlenecks and oversized resource requests, which reduced the cluster from five worker nodes to four. Unlike earlier findings where capacity needed to be increased, this audit pointed in the opposite direction, suggesting the storage allocation could be safely reduced. The conclusion was not that 200Gi would never be needed, but that the current provisioning exceeded what the workload and operating conditions actually required.

0
ProgrammingDEV Community ·

How Nginx Cache-Control Headers Can Delay Critical React Hotfixes Reaching Users

When developers deploy a React hotfix, some users may continue seeing the old version due to browser-side caching behavior that silently works against urgent updates. Browsers apply heuristic caching — defined in RFC 7234 — when no explicit Cache-Control headers are set, calculating a time-to-live based on a file's Last-Modified timestamp provided automatically by Nginx. Immediately after a fresh deploy, this TTL is near zero, so everything appears fine, but as days pass the TTL grows, meaning browsers serve stale files without ever contacting the server. The root cause is that browsers only send a conditional revalidation request to Nginx after the TTL expires, so ETags and updated files are never checked while the cache is still considered valid. The fix involves explicitly setting Cache-Control response headers in Nginx configuration, giving developers direct control over how long browsers store static assets.

0
ProgrammingDEV Community ·

AI Hits Physical Limits: GPU Shortages, 151M Token Theft, and a 38GW Power Crisis

The rapid scaling of AI systems is now running into real-world physical and security constraints, as highlighted by three developments this week. OpenAI paused new Pro subscriptions after demand for its high-reasoning GPT-6 Astra architecture overwhelmed available GPU capacity, since advanced reasoning models require far more compute per query than standard models. In a separate incident, attackers used 3,500 compromised accounts to extract over 151 million tokens from Anthropic's Claude, disguising data-harvesting prompts as Japanese translation requests to evade detection and train a cheaper rival model. Meanwhile, Microsoft's reported request for 38 gigawatts of electricity to power future data centers underscored the staggering energy demands of AI infrastructure — a figure that surpasses the total grid capacity of countries like Ireland or New Zealand. Experts now argue that the next competitive edge in AI will depend not just on model quality, but on advances in energy efficiency, inference optimization, and API security.

0
ProgrammingDEV Community ·

React's Standalone startTransition API Enables Concurrent Scheduling Outside Components

React exports a standalone startTransition function that schedules low-priority state updates without requiring component scope or hooks. Unlike useTransition, which must be called inside a React component, the standalone API can be used in third-party stores, synchronous callbacks, and non-component modules. The key trade-off is that startTransition does not return an isPending flag, meaning loading state indicators must still be handled via useTransition inside components. Developers can combine both approaches — using startTransition in store logic and useTransition in components — to achieve concurrent scheduling across the entire application. This prevents rendering jank caused by blocking state updates in code that sits outside the React component tree.

0
ProgrammingDEV Community ·

How Redis Pub/Sub and WebSockets Solve Real-Time Scaling Challenges

Modern applications increasingly demand real-time interactivity, making low-latency communication a core engineering requirement rather than an optional feature. WebSockets, standardized under RFC 6455, have become the dominant protocol for persistent, bidirectional client-server communication, replacing inefficient older approaches like HTTP polling. However, WebSocket connections are stateful, meaning each client is tied to a specific server instance, which creates serious obstacles when scaling horizontally across multiple nodes. Without a synchronization layer, messages can become siloed per server, causing fragmented user experiences, file descriptor exhaustion, and memory overflows. The article outlines how Redis Pub/Sub and Redis Streams can serve as a message broker backbone to bridge these isolated server instances and enable reliable, high-concurrency real-time applications.

0
ProgrammingDEV Community ·

Agentic Synthetic Data Generation Emerges as Key Shift in AI Development

As publicly available web data approaches saturation, AI researchers and developers are turning to agentic synthetic data generation as a new approach to training models. This method involves autonomous agents operating within simulated environments to produce high-quality behavioral datasets, rather than depending on scraped real-world data. The approach addresses two significant challenges in AI development: privacy compliance and the declining availability of clean, usable training data. Proponents argue that systems capable of generating and learning from their own high-fidelity environments represent the next frontier in large language model and agent development. The shift is prompting debate within the AI community over whether synthetic data fine-tuning offers advantages over retrieval-augmented generation pipelines.

← NewerPage 1400 of 3334Older →