SShortSingh.

Programming

0
ProgrammingDEV Community ·

How a Metric Data Dictionary Makes Cybersecurity Dashboards Reproducible and Trustworthy

Cybersecurity dashboards often appear precise but fail when two analysts cannot reproduce the same number from the same data, typically because metric definitions are incomplete or scattered. The root cause is an undefined metric contract — missing scope, formula, denominator, source, and exception rules — rather than a flawed charting tool. NIST recommends a structured measurement program that documents scope, numeric formulas, targets, data sources, responsible parties, and reporting formats for each metric. A practical data dictionary consolidates these elements into a single reviewable record per metric, covering fields such as grain, numerator, denominator, validation tests, and exception policies. Security program managers and GRC leads are advised to assign each metric a stable ID and complete all definition fields before the metric enters any recurring report.

0
ProgrammingHacker News ·

Magnitude 7.7 Earthquake Strikes Near Ende, Indonesia

A powerful 7.7 magnitude earthquake struck approximately 68 kilometres north-northwest of Ende, a city located on Flores Island in Indonesia. The seismic event was recorded and reported by the United States Geological Survey (USGS). Indonesia sits within the Pacific Ring of Fire, making it one of the most seismically active regions in the world. Authorities and monitoring agencies were expected to assess the potential for tsunamis or significant damage following a quake of this magnitude.

0
ProgrammingDEV Community ·

Mixedbread AI Launches Toast 1 Embedding Model to Rival OpenAI at Lower Cost

Mixedbread AI has released Toast 1, an open-weight text embedding model that outperforms OpenAI's text-embedding-3-large on the MTEB benchmark with a score of 68.4 versus 67.9. The model supports over 50 languages and offers variable output dimensions, including Matryoshka embeddings that allow vector truncation while retaining most semantic information. Mixedbread attributed the model's performance to high-quality training data curation and transformer architecture improvements rather than sheer data volume. OpenAI's competing model costs $0.13 per million tokens, while Toast 1 can be self-hosted on a consumer GPU with 8GB VRAM, potentially reducing costs to a fraction of that figure. The model is aimed at developers building RAG systems, semantic search, and document classification applications at scale.

0
ProgrammingDEV Community ·

Anthropic Engineers Share 7 Practical Tips for Better Claude Code Sessions

Anthropic has published an official guide with seven tips for developers to get more out of Claude Code, its CLI-based AI coding assistant. The guide, which gained traction on Hacker News with 130 points and 87 comments, emphasizes starting sessions with precise context including project structure, relevant files, and error details. Key recommendations include using a persistent CLAUDE.md configuration file, breaking tasks into small verifiable steps, and leveraging the tool's codebase exploration and planning modes before writing any code. The guide also advises enabling git commits after each step to allow easy rollbacks, and starting fresh sessions when long conversations become cluttered with outdated context. While tailored to Claude Code, Anthropic notes the advice applies broadly to any AI-assisted coding workflow.

0
ProgrammingDEV Community ·

Encryption Won, So Law Enforcement Turned to Hacking Your Devices Instead

Cryptography engineer Matthew Green argues in a widely circulated essay that law enforcement agencies have quietly abandoned the push for encryption backdoors in favor of directly hacking target devices. Rather than breaking end-to-end encryption, agencies now deploy spyware to read messages before they are encrypted or after they are decrypted on a user's phone or laptop. This shift has fueled a commercial spyware industry, with firms like NSO Group and Cytrox selling tools to governments worldwide — tools that have been used against journalists, activists, and political opponents. Green warns that while public debate remains focused on encryption backdoors, device hacking operates in a largely ungoverned legal grey area with far fewer oversight requirements than traditional wiretapping. The result, he contends, is a new surveillance frontier with significant civil liberties implications and almost no adequate policy framework to address it.

0
ProgrammingDEV Community ·

Murya AI: First MIT-Licensed Hausa Text-to-Speech Tool Hits 10,000 Views

A developer has launched Murya, a free, open-source text-to-speech (TTS) platform designed for the Hausa language, which has over 95 million speakers worldwide. The project, whose name means 'Voice' in Hausa, is believed to be the first publicly available, MIT-licensed TTS tool built by a native Hausa speaker. The creator argues that labeling languages like Hausa as 'low resource' is a socially constructed framing rather than an inherent limitation. Within its first month, the platform at app.murya.ng surpassed 10,000 page views according to Cloudflare analytics. The project is open for contributions and is hosted on Hugging Face, with the developer citing support from AI research communities focused on underrepresented languages.

0
ProgrammingDEV Community ·

Google Cuts Homomorphic Encryption Overhead to 10-50x, Bringing Private AI Closer to Reality

Google has announced a significant advancement in homomorphic encryption (HE) that makes private AI inference more practically viable. The research focuses on optimizing HE for neural network matrix operations, developing TPU-based hardware acceleration, and combining HE with secure multi-party computation. These improvements reduce the computational overhead from roughly 1,000 times slower than standard processing down to approximately 10 to 50 times, crossing a key usability threshold. The breakthrough could enable sensitive sectors such as healthcare and finance to use cloud-based AI services without exposing underlying data. While overhead remains substantial and challenges persist for large models, the advance marks a notable step beyond prior work by Microsoft, IBM, and others in the HE space.

0
ProgrammingDEV Community ·

Dev team corrects landing page that falsely claimed free plan needed no registration

A marketing review uncovered a contradiction on a software product's landing page, where one section asked users to register via email while the FAQ stated no registration was required. Code review confirmed that email registration is mandatory for the free plan, making the FAQ claim factually incorrect. The team traced the error to a gradual shortening of 'no credit card required' into the broader 'no registration required' over time. Rather than removing the email requirement, developers updated the FAQ and Quick Start copy across three files and four locations in both English and Japanese to accurately reflect the sign-up process. The revision preserved key cost-related assurances — no credit card and no time limit — while correcting only the procedural claim about registration.

0
ProgrammingDEV Community ·

Circle Ran 8 AI Agents With Real USDC Wallets to Bet on World Cup Matches

Circle, the company behind the USDC stablecoin, conducted a public experiment called 'Steve' in which eight autonomous AI agents were each given their own USDC wallet and tasked with predicting the final three matches of the 2026 World Cup. Operating without real-time human oversight, the agents independently purchased live match data and social sentiment feeds via the x402 payment protocol, then placed prediction bets on Polymarket. Every transaction and balance change was broadcast live on a public website, and each agent operated under strict per-transaction and total spending caps it could not override. After the experiment concluded, the remaining funds — totalling over $10,000 — were donated to the Apache Software Foundation, with Circle matching the amount dollar-for-dollar for a combined contribution exceeding $20,000. The demo is being cited as an early real-world example of the emerging 'agent economy,' where AI systems autonomously buy, sell, and settle for services using on-chain payments.

0
ProgrammingDEV Community ·

Prompt Injection Remains a Critical Threat to AI-Powered Apps in 2026

Prompt injection attacks occur when untrusted text fed into an AI model contains hidden instructions that the model executes as legitimate commands, bypassing intended boundaries. A common example involves a support chatbot that can take actions — such as issuing refunds — being tricked by a customer ticket containing embedded directives. Unlike traditional exploits, these attacks require no code vulnerabilities; a carefully worded sentence is sufficient to manipulate the model. Security experts warn that adding defensive language to system prompts is ineffective, since user input and system instructions occupy the same token stream and compete for the model's attention. Recommended mitigations include structurally separating untrusted content, restricting the model's available tools to a minimal set, requiring deterministic code checks before any action is executed, and placing human approval gates on irreversible operations.

0
ProgrammingDEV Community ·

Developer builds open-source WhatsApp AI platform after finding all existing frameworks inadequate

A developer reviewed every major open-source WhatsApp bot framework on GitHub and concluded they all share a fundamental architectural flaw: they function as basic API wrappers that add generic AI without domain-specific capabilities. Existing tools route messages and append conversation history to a language model, but lack the industry-specific logic needed for businesses like restaurants or law firms that require distinct workflows and data access. Enterprise platforms from providers such as Twilio and Intercom charge up to $2,000 per month but similarly fail to integrate meaningfully with a client's own systems or data. The developer also highlighted that most frameworks ignore critical production concerns such as GDPR-sensitive PII flowing through third-party APIs and the risk of service outages from a single AI provider. In response, they built SARA, an open-source WhatsApp AI agent platform featuring 20 vertical-specific agent profiles, each equipped with its own tailored toolset and function-calling capabilities.

0
ProgrammingHacker News ·

Ploopy Launches A+ Trackball, Its Latest Open-Source Peripheral

Ploopy has announced the release of its new A+ trackball device, adding to its lineup of open-source input peripherals. The product was shared via the company's official blog, generating early attention in the tech community. Ploopy is known for producing customizable, open-hardware pointing devices aimed at enthusiasts. The A+ model appears to be an updated or upgraded variant within the Ploopy trackball range. Further technical details and pricing are available through the official Ploopy blog post.

0
ProgrammingDEV Community ·

Lean Founder on Formal Verification, AI, and the Future of Software Proofs

Leonardo de Moura, creator of the Lean proof assistant, was interviewed about how formal verification can guarantee the absence of software bugs — going beyond what traditional testing allows. Lean functions both as a programming language and a proof system built on dependent type theory, enabling mathematical and software verification within a single platform. The Lean nonprofit foundation, established in 2023 with major support from AWS, is now accelerating Lean's role as a programmable software verification system after a decade as an academic project. De Moura highlighted a key future use case: AI-generated code optimizations paired with machine-checked proofs of behavioral equivalence, removing the risk from automated refactoring. He also recommended learning Lean today by pairing it with AI agents, using official resources such as 'Theorem Proving in Lean' and 'Functional Programming in Lean'.

0
ProgrammingHacker News ·

Disney Research uses 2D Gaussian Splatting to convert line art into vector curves

Disney Research Studios has published a new technique that applies 2D Gaussian Splatting to automatically vectorize line art into Bézier spline representations. The method was presented in July 2026 and is detailed on the Disney Research website. The approach aims to improve the accuracy and efficiency of converting raster line drawings into scalable vector graphics. This kind of vectorization is valuable in animation and digital art pipelines where clean, editable curve data is essential.

0
ProgrammingDEV Community ·

Developer Shares Token Budget Guard to Prevent Free AI API Tier Overuse

A developer has published a lightweight Python script designed to track and limit token consumption when using free AI model endpoints. The tool maintains a local JSON ledger that estimates token costs before each API call and records actual usage afterward, refusing to proceed if the projected total would exceed a set budget. The approach addresses common pitfalls such as runaway retry loops and oversized context buffers that can silently drain a free tier's token allowance before dashboard alerts trigger. The script is built around the OpenAI-style chat completion format and can be adapted to other compatible endpoints by swapping a single function. The article was written as part of promotional outreach for MonkeyCode, an open-source project offering a free model route and hosted server.

0
ProgrammingDEV Community ·

How a Simple Token Ledger Can Prevent Costly Free-Tier API Overruns

Developers using free AI model tiers often discover quota overruns only after a batch job fails, wasting both time and retry budget. A lightweight Python token ledger can predict whether a batch job fits within a given allowance before any live API call is made. The approach uses a rough heuristic of one token per four characters to estimate prompt and completion costs across thousands of calls. For example, a job requiring 6,000 summaries with 21,000-character prompts would need an estimated 32.25 million tokens, exceeding a 30-million-token free allowance. The ledger does not replace a real tokenizer but serves as a fast, deterministic pass/fail filter that separates budget planning from actual API spending.

0
ProgrammingDEV Community ·

Dart Devs Achieve 100K Ops/Sec on Web Using BlocSignal and Jaspr Framework

The team behind BlocSignal rebuilt their official documentation site at blocsignal.dev using Jaspr, a lightweight Dart-to-HTML framework, to avoid the performance overhead of Flutter Web. Their initial implementation relied on manual .subscribe() callbacks inside StatefulComponent lifecycles, which caused double re-renders, UI thrashing during high-frequency events, and unnecessarily broad component rebuilds. To fix these issues, they migrated to declarative consumer components provided by the bloc_signals_jaspr package, mirroring patterns already established in bloc_signals_flutter. The refactored architecture achieved 100,000 operations per second in compiled JavaScript while also improving developer ergonomics through Dart 3.13 primary constructors. The project serves as a practical case study in dogfooding their own library to validate its real-world performance and idiomatic usage.

0
ProgrammingDEV Community ·

Restaurant Manager Reverse-Engineers Undocumented ERP, Maps 390 Tables Without Vendor Help

A self-taught developer working as an operations manager at a restaurant in Itaúna, Minas Gerais, Brazil, spent months reverse-engineering TronSoft, a proprietary ERP system built on a Firebird database, entirely without vendor documentation. Needing to automate payment reconciliation, comanda closures, and fiscal document emission, he had no API references or schema diagrams to rely on. By systematically observing the live production database, he documented 390 tables, 514 foreign keys, and roughly 40 functional modules through trial, error, and inference alone. He overcame challenges including Firebird's non-standard SQL dialect, generator-based primary keys, and unpredictable transaction visibility before eventually moving from fragile UI automation to direct, reliable database writes. The experience highlighted how constraint and lack of resources can drive deeper technical understanding than conventional, well-documented development environments.

0
ProgrammingHacker News ·

Researchers Propose Simplified Approach to Teaching Introductory Calculus

A research paper published on arXiv argues for simplifying and refactoring how introductory calculus is taught. The work suggests that the traditional structure of calculus education may benefit from reorganization to improve clarity and accessibility. The paper proposes a refined pedagogical framework aimed at making foundational calculus concepts easier to learn. It was shared on Hacker News, where it received modest attention with 8 points and no comments at the time of posting.

← NewerPage 90 of 1301Older →