SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

CBN Opens Regulatory Sandbox Cohort 2 to Unlicensed Nigerian Fintech Startups

The Central Bank of Nigeria (CBN) has launched Cohort 2 of its Regulatory Sandbox, making it accessible to unlicensed fintech startups with working minimum viable products for the first time. The program runs for six to twelve months and provides supervised testing with real users under CBN oversight, culminating in a readiness assessment for a full licence application. Applicants can apply under one of two tracks: a VASP track for crypto and digital asset products, or a Non-VASP track covering open banking, payments, fraud detection, and financial inclusion services. To qualify, startups must demonstrate meaningful innovation, have no outstanding regulatory sanctions, and meet compliance requirements including AML policies, data protection, cybersecurity controls, and a named compliance officer. Experts advise builders not to pursue the VASP track unless their product genuinely involves digital assets, as it carries significantly greater regulatory and technical complexity, including SEC Nigeria oversight and blockchain analytics obligations.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

monday.com Shares Its Approach to Testing AI Agents Against Real Systems

monday.com recently detailed its methodology for evaluating AI agents in a webinar recap published on DEV Community on September 21. The company, known for its work management platform, outlined how it runs agent evaluations against real dependencies rather than mocked environments. The recap was authored by Arsh Sharma for MetalBear and covers practical insights into AI and LLM testing. This approach aims to ensure AI agents perform reliably under actual production conditions. The discussion is part of a growing industry focus on robust evaluation frameworks for AI-powered systems.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How monday.com Built Agent Evals Against Live Staging Instead of Mocks

monday.com AI Engineering Director Dor Cohen shared insights in a September 15 webinar on why his team chose to run agent evaluations against a live staging cluster rather than mocked dependencies. Cohen demonstrated that mocks fall short because they drift from production, use unrealistically clean synthetic data, and cannot accurately represent stateful changes made by an agent. A key example showed an agent that appeared to complete a task correctly but had silently processed only 500 of 600 requested items — a flaw only caught by evaluating the full action trajectory, not just the final output. To connect eval runners directly to a real Kubernetes environment, monday.com adopted mirrord, a tool that mirrors live traffic, secrets, and file systems to a local or CI process. The approach gives their evals production-level fidelity, enabling them to catch edge cases and verify actual end states that mock-based testing would miss.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

15 AI Search Guides Reviewed: Four Key Checks Before You Act on Their Advice

A review of 15 top-ranking guides on AI search optimisation found several widely repeated recommendations lack solid evidence or practical context. Schema markup, endorsed by all five 'LLM SEO' guides, showed statistically no effect on ChatGPT citations in an Ahrefs study tracking nearly 1,900 pages over seven months. AI visibility scores are highly volatile — with just 30 data points, a 10-percentage-point swing can be statistical noise rather than a meaningful change, as Semrush's own head of AI visibility has noted. Guides also largely fail to distinguish between OpenAI's three crawlers — OAI-SearchBot, ChatGPT-User, and GPTBot — which serve different purposes and carry different implications for how a site appears in ChatGPT responses. The review, conducted while building guides for tool Shruwd, underscores the importance of checking sources and sample sizes before acting on AI search advice.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Finance AI Agents Need Evaluation Harnesses, Not Just Better Prompts

AI-powered finance agents can generate confident-sounding but incorrect decisions, making prompt engineering alone an insufficient quality control strategy for accounting automation. Experts recommend building an evaluation harness — a repeatable test system that runs realistic financial scenarios through an agent and scores outputs against predefined expectations. A robust test dataset should include routine, missing-evidence, conflicting, and adversarial cases, with expected behaviors stored as structured data to enable regression testing. Rather than relying on a single quality score, evaluations should be layered across schema validity, evidence grounding, policy compliance, and decision accuracy. This approach ensures that agent behavior remains auditable, within authority boundaries, and reliable enough for production use.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

LLMFit Helps Developers Match Open-Source AI Models to Their Hardware

A tool called LLMFit has been developed to help users identify which open-source large language models (LLMs) are compatible with their local hardware. The tool scans a system's CPU, RAM, and GPU specifications and recommends models that can run efficiently on that machine. It aims to eliminate the trial-and-error process of downloading models that may be too demanding for a given device. LLMFit is particularly useful for developers experimenting with local AI tools such as Ollama or building side projects with LLMs. The project is available on GitHub.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev Tool Adds Oracle Safety Check to DeFi Agents Before Trade Execution

A new pre-trade safety mechanism has been proposed for decentralized finance (DeFi) agents to verify oracle price data before submitting transactions. The tool targets a key vulnerability where technically valid transactions can still result in poor outcomes due to stale or inconsistent price feeds. The risk is particularly acute in lending protocols, where even minor oracle discrepancies can erode a position's liquidation buffer before a borrow is executed. The project, shared on DEV Community, includes a reference implementation available on GitHub alongside a dedicated information site. The initiative aims to add a reliability layer to automated DeFi agents that rely on real-time price data for decision-making.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Integrate st-core.fscss CSS Charts into a Svelte App at Build Time

The st-core.fscss library renders charts using pure CSS clip-path polygons and custom properties, requiring no JavaScript chart library, SVG, or canvas. Integrating it into a Svelte or SvelteKit project involves compiling the FSCSS source file into static CSS once at build time using the FSCSS CLI, rather than loading a runtime CDN script on every page load. Svelte's role is limited to reactively mapping data values to CSS custom properties such as --st-p1 and --st-p2, while the browser handles rendering via the compiled CSS. A key constraint is that the array length defined in the FSCSS template must match the number of data points, as it determines the polygon geometry and cannot be changed without recompiling. The article provides component examples for both Svelte 4 and Svelte 5, along with package.json scripts to automate compilation before each build.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Most structured data parsers miss JSON-LD inside @graph blocks, audit finds

A developer discovered that nine out of twelve structured data audit tools, including their own, fail to detect schema markup when it is wrapped inside a JSON-LD @graph block rather than a flat top-level object. The @graph format is the default output of popular WordPress SEO plugins Yoast SEO and RankMath, meaning a large share of correctly marked-up web pages are being misread as having no structured data. Unlike flat JSON-LD, the @graph structure allows schema nodes to reference each other by ID, forming a single linked entity graph rather than isolated, redundant blocks. The author identified the bug in their own tool after a well-marked page returned a zero schema score, and traced the root cause to parsers checking only for a top-level @type field. A fix involving recursive node iteration — handling both flat and @graph shapes at the parse boundary — is proposed so that no downstream check needs to account for either format separately.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer stress-tests AI cost-monitoring agent using chaos testing and red teaming

A software developer subjected their AWS cost-monitoring AI agent, built on Amazon Bedrock's AgentCore Runtime using the Strands framework, to deliberate failure and adversarial testing. The agent, which daily reads AWS billing data and emails a color-coded spending report, had never been tested under adverse conditions such as API timeouts, partial data returns, or prompt manipulation. Using Strands Evals 1.0, which shipped chaos testing and red-teaming tools in June, the developer built an isolated lab replica of the agent running on fixture data to avoid incurring real API costs during testing. The key concern was not outright crashes but silent, believable failures — such as a falsely reassuring green-status report while cloud costs quietly escalate. The developer's stated goal is to share replicable testing methods so others can apply the same adversarial evaluation approach to their own AI agents.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Launches Free AI Toolkit for LinkedIn Profiles, Resumes, and Bios

A developer has released ProfileCraftAI, a free, no-signup toolkit designed to help users write LinkedIn headlines, resume summaries, professional bios, and more. The platform offers eight tools in total, built using Next.js for the frontend and Groq's Llama-based API for near-instant text generation, deployed on Vercel. The project was motivated by the developer's observation that many people struggle to articulate their professional identity in writing. Key lessons from the build include the importance of carefully structured prompts to avoid generic AI output, adding educational content alongside each tool to build user trust, and implementing rate limiting early to prevent abuse. The toolkit is publicly accessible at profilecraftai.com, and the developer is actively seeking community feedback on usability and missing features.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Disposable Email Checkers Protect Sender Reputation and Block Signup Abuse

Disposable email addresses pose a significant risk to businesses running SaaS products, newsletters, or waitlists, as these inboxes typically expire within hours and cause hard bounces that damage a sender's domain reputation. Sustained bounce rates above roughly 2% can cause mailbox providers to route all outgoing mail — including legitimate messages — to spam folders. Tools like Reloop's free Temp Email Checker address this by matching signup addresses against a catalogue of approximately 210,000 known disposable providers, validating MX records, and flagging role-based prefixes, all without probing actual mailboxes. Beyond deliverability, blocking disposable emails at the point of registration also prevents free-trial farming, coupon fraud, waitlist inflation, and multi-account bot abuse. The recommended approach is to validate email addresses before they enter a database, then immediately suppress any hard bounces to prevent dead addresses from receiving follow-up sends.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds AI Argument Debugger in Three Days for AWS Hackathon

A developer built MindMap Debugger, a tool that scans text for logical contradictions and circular reasoning, as part of the WeMakeDevs AWS First Commit hackathon. The tool accepts any pasted argument or transcript, uses an LLM via Groq's API wrapped in AWS's Strands Agents SDK, and flags claims that conflict or loop back on themselves. Without a debit or credit card for AWS signup, the developer switched from the 'Ship It' track to the 'Build It' track while still meeting the AWS-native tooling requirement. Key challenges included a reasoning model leaking its internal chain-of-thought into outputs and non-deterministic extraction results, the latter solved by running three parallel extractions and merging consensus findings. The project also required a similarity-matching fix after the model paraphrased identical claims differently across runs, causing duplicate propositions to slip through the merge logic.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Third-Party Scripts Quietly Drain Performance on CMS-Driven Marketing Sites

On CMS-driven marketing sites, performance often degrades not from code changes but from content decisions made by editors — such as adding embeds, chat widgets, and heatmap tools over time. Developer Aryan Bhosale highlights this challenge on the CometChat website, where marketing teams freely compose pages using a headless CMS. Key metrics like LCP, INP, and CLS are most commonly hurt by third-party JavaScript and unoptimized media rather than engineering missteps. To address this, he recommends auditing every third-party script, replacing open HTML embed fields with typed CMS components, and deferring non-essential scripts until after user interaction. Longer-term guardrails — including CI-based performance budgets and a single owner for the tag manager — are proposed as more durable fixes than one-off cleanups.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Shadow DOM Breaks Screen Reader Accessibility and Ways to Fix It

Web components built with Shadow DOM can appear visually flawless yet remain largely invisible or confusing to screen readers, because browsers construct a separate accessibility tree that does not always expose a shadow root's internals correctly. A developer discovered this while testing a custom dropdown with VoiceOver and NVDA, finding that ARIA roles set inside the shadow root were ignored and keyboard focus behaved unpredictably. The root cause lies in how browsers treat shadow boundaries differently from standard light DOM when building the accessibility tree, with behavior varying across browsers and assistive technologies. Practical fixes include creating shadow roots with the delegatesFocus option, applying explicit ARIA roles such as listbox and option to internal elements, and using aria-hidden to conceal purely decorative nodes. Using native HTML semantics wherever possible inside shadow components is also recommended, as ARIA attributes alone cannot fully compensate for poor structural exposure.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Bivack Lets Coding Agents Run on AWS Lambda MicroVMs, Accessible from Any Device

A developer has released a new version of Bivack, a cloud-based sandbox that runs coding agents on AWS Lambda MicroVMs, allowing users to work from any device including phones and laptops without losing session context. Each user gets a dedicated MicroVM with a persistent home directory stored on Amazon S3, meaning files and logins survive restarts and device switches. Unlike traditional setups where coding agents are tied to a local machine, Bivack lets agents continue running in the cloud while the user moves between devices. The platform is accessible entirely through a browser, offering both a terminal and a full VS Code-style workbench so users can actively review and edit agent-produced code. The developer built and shipped the latest release while traveling across multiple locations, demonstrating the tool's core premise of location-independent, persistent development environments.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Neuron AI Adds Classifier Component Using TypeSafe AI's Jev for PHP Agent Routing

TypeSafe AI released Jev on September 15, a model designed not to generate text but to classify inputs by returning probabilities across predefined answer sets in a single pass. Unlike generative models, Jev accepts a state and a list of questions with fixed answers, making it suitable for fast, low-cost decision-making within AI agents. A PHP developer integrated Jev into the Neuron AI framework as a new Classifier component, enabling agents to route conversations by difficulty, enforce policy checks, or decide tool-call safety on every turn. TypeSafe AI reports Jev responds in 70 to 500 milliseconds and is priced at $0.042 per million input tokens with no output charges, addressing the latency and cost problems of using full generative models as judges. The component is currently in early access, with TypeSafeAI serving as the first concrete implementation of the Neuron AI Classifier interface.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Build a Risk-Score Middleware in Laravel to Block Signup Abuse

A technical guide published on DEV Community outlines how to build a Laravel middleware that blocks high-risk signups using IP risk scoring rather than static blocklists. The approach addresses signup abuse, where attackers rotate IP addresses too quickly for traditional ban lists to be effective. Key implementation decisions include correctly resolving the real user IP behind proxies, avoiding blanket VPN blocks, and applying a four-band scoring system — allow, observe, challenge, and block — instead of a single threshold. The middleware calls an external scoring API with a 1.5-second timeout and caches verdicts, with developers choosing deliberately between fail-open and fail-closed behavior. Compatible with Laravel 11, 12, and 13, the full build including a service class, middleware, and Pest tests takes approximately twenty minutes to complete.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Spain records first fully autonomous AI-executed data breach, raising security alarms

Spain's Data Protection Agency has confirmed the country's first data breach carried out entirely by an AI agent, with no human involved at any stage of the attack. The agent independently performed reconnaissance, authenticated, probed the system, and altered data — a step beyond previously documented cases where humans still directed each move. Separately, researchers at Hacktron detailed a related attack chain that exploited a flaw in the libheif image-decoding library, a misconfigured OpenAI SSO, and used Claude Opus 5 to build the exploit. In a third incident, a threat actor dubbed LeakySensey compromised over 87,000 IP addresses via brute-force attacks on PPTP and L2TP devices, automating the operation with a modified AI coding tool and earning over $200,000 since 2024. Security experts note that the common thread across all three cases is not novel sophistication but rather long-standing unpatched vulnerabilities and misconfigurations that AI tools now allow a single operator to exploit at scale.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

ZuckOff App Claims to Detect Hidden Cameras in Any Room

ZuckOff is a new app designed to alert users when a camera may be present in their surroundings. The tool appears aimed at privacy-conscious individuals concerned about covert surveillance in homes, hotels, or other spaces. It gained traction on Hacker News, accumulating 115 upvotes and sparking 16 comments. The app's name is a cultural reference to Facebook founder Mark Zuckerberg, who is widely associated with privacy controversies. No detailed technical information about its detection method was publicly available from the submission.

0 comments Read more at Hacker News
← NewerPage 1051 of 3197Older →