SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Five RAG chunking strategies with runnable Python examples

A guide on DEV Community walks through five chunking strategies that cover most production retrieval-augmented generation pipelines in 2026. The article argues that chunking is often glossed over in tutorials, yet it drives many retrieval failures, and that the right approach depends on document structure and query shape. It identifies three failure modes: chunks that are too large and dilute specificity, chunks that are too small and lose context, and chunks that split definitions from their examples. Each strategy is accompanied by runnable Python code and guidance on when it performs best. The piece follows up on an earlier article covering the full four-step RAG loop.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Four offline checks to run before giving a marketing agent live API access

A marketing agent can return valid JSON and still reach a wrong conclusion, such as averaging ratios rather than summing totals or treating an incomplete CRM response as proof of absence. The author presents four small offline exercises using synthetic data and Python 3.9+ standard library tools to expose these errors before any live account is connected. The first tool audits a normalized placement report, where total CPA is computed from summed cost over summed conversions and flags placements for manual review. The second compares local competitor offer snapshots and warns when changed scope prevents a like-for-like price comparison. The third is a read-only Bitrix CRM demonstration that distinguishes scheduled, overdue, undated and incomplete activity reads. The article stresses that these outputs do not establish causation, complete customer journeys, or actual offer availability.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Mobile App, PWA or Website? A 10-Point Scorecard for Small Businesses

Choosing between a mobile app, a progressive web app (PWA), or a plain website depends mainly on how often customers use your service and what device features you need, not on what looks most impressive. A PWA is a website that can be installed on a phone's home screen and can send notifications and work partly offline. Native apps make sense when customers engage weekly or more, or when hardware and background features are essential, but they bring store fees, review processes and ongoing admin. The biggest limitation for PWAs is iPhone, where web push notifications only work after users manually add the site to their Home Screen, while Android offers a smoother install experience.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

7 Common Developer Mistakes That Quietly Hurt Google Rankings

A well-built, fast website can still fail to rank because of small technical decisions made during development rather than weak content. A staging robots.txt rule that blocks all crawling can accidentally be pushed to production and remove a whole site from Google's index. Leftover noindex meta tags from staging are another risk, and the article suggests controlling them with an environment variable. Client-side-only rendering can leave pages blank to crawlers, so server-side rendering or static generation is recommended. The piece also covers duplicate or missing page titles and canonical tags, non-link navigation that crawlers cannot follow, Core Web Vitals targets, and structured data validation, and it closes with a pre-launch SEO checklist.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Devcon 8 in Mumbai: how builders can plan their week

Devcon 8, the Ethereum Foundation's conference for Ethereum builders, takes place 3 to 6 November 2026 at Jio World Centre in Bandra Kurla Complex, Mumbai. The article advises arranging short meetings with grant staff, hiring managers and developer relations leads before arriving, since the main venue is where most builders gather. It argues that hallway conversations at the main venue often beat side events, which can cost hours in Mumbai traffic. Side events should be chosen by their host and intended audience, with a preference for dinners hosted by partner protocols or funds and for Indian builder communities. Builders are also encouraged to make their registration answers and linked work easy to verify, and to join waitlists.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Chrome clipboard managers after the Manifest V2 removal: what to check before switching

Google's phase-out of Manifest V2 extensions in Chrome has now concluded, with the Chrome Web Store removing remaining V2 extensions by August 31, 2026. Many classic clipboard managers were abandoned before the deadline, leaving users without warning when their tools stopped working or disappeared. The article advises checking whether a replacement is Manifest V3, handles text versus rich content, where it captures copied data, and where its history is stored. It also recommends looking for password and Incognito exclusions, search, pins, export options, and clear free-tier limits. The piece is AI-assisted and includes a promotional mention of Clipstay, a text-only clipboard manager with limitations the author notes.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Developer says Anthropic Console charged $504 instead of $5.04 due to decimal input bug

A Costa Rica-based LegalTech attorney reports that loading credits onto an Anthropic API account resulted in a $504.00 charge after entering 5.04, which they attribute to a decimal separator handling error. They say the platform silently removed the decimal and processed the charge without a confirmation screen. The user says they filed a chargeback with their bank, which can take up to 120 days, and that automated support denied a refund. They argue the same flaw could cause larger overcharges for other users and are calling on Anthropic's billing team to review the case.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

Piramidal, a YC W24 company, is hiring

Piramidal, a startup from Y Combinator's Winter 2024 batch, has posted open job listings on its Greenhouse careers page. The post was shared on Hacker News, where it has received no points or comments so far. The source material provides only the job board link and basic Hacker News metadata, so the specific roles, locations, and company details are not available. Interested applicants would need to review the listing directly for position details.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

Most US Work Hours Fall Outside What AI Benchmarks Measure, Study Finds

A report published via Mercor's research site argues that roughly 90% of time US workers spend on the job is devoted to tasks that current AI benchmarks do not adequately evaluate. The analysis maps occupational activities against existing AI evaluation tests to identify where coverage is weak. The findings suggest that benchmark scores may give an incomplete picture of AI's readiness for real workplace use. The piece was shared on Hacker News, where it drew moderate attention and discussion.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Shared Compute Let an AI Agent Leak Bank Data to Slack, Repro Shows

A Grok Bot agent given read-only access to a CEO's bank account reportedly posted his balances to his company's Slack, after sharing a cloud compute instance with a separate Slack-posting agent. The author reproduced the mechanism with two toy Docker agents that mount the same volume, with the Slack agent posting whatever file it finds at a shared path. The leak comes from the absence of an isolation boundary, not from either agent's code, since the Slack agent has no finance-specific logic. The demo repo includes a vulnerable compose file and the author discloses working at Solo.io, which created kagent.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Docker Stop Signals Go to PID 1 Only, Not Every Container Process

A widespread belief holds that docker stop delivers SIGTERM to all processes in a container, but the signal actually goes only to the container's main process, PID 1. When a Dockerfile uses shell form, that shell becomes PID 1 and may not pass the signal along to the actual application. If the grace period lapses, Docker sends SIGKILL, which cannot be intercepted and forces an immediate termination. Using exec form, or a wrapper script that calls exec, lets the application itself receive the stop signal directly. Because PID 1 also reaps orphaned child processes, a lightweight init such as Docker's --init option can help applications that spawn subprocesses.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Realtime Room Admission Control Using Scoped Tokens and Client Trust Signals

A server-side admission gate issues short-lived, room-scoped capabilities only after verifying the operator, requested IoT devices, and allowed actions. Observability should track the admission decision, its reason, token scope, policy version, and decision latency, separate from WebRTC connection state. Transport health alone cannot prove a client was authorized to join, so the two must be kept distinct. Connection state values such as connected or failed do not indicate whether policy granted or denied access. Teams are advised to log one structured event per decision while keeping bearer tokens and command payloads out of logs, and to keep metric labels bounded to avoid unmanageable cardinality.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Autonomous AI Coding Agent That Refines Code Using Test Feedback

A developer has built an autonomous coding agent designed to tackle challenging programming tasks in competitive settings. Instead of producing code in a single LLM pass, the system follows a full workflow of repository analysis, planning, implementation, execution, testing, and refinement. Its key feature is an iterative feedback loop in which failed tests are diagnosed, the code is updated, and the relevant tests are rerun before submission. The author plans to compare performance with and without this loop using success rate, tests passed, completion time, and compute cost, though the article presents these metrics only as placeholders. Further optimizations include model selection, prompt design, tool usage, and failure recovery.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

GalaxyAtlas: A Browser-Based 3D Galaxy Built With AI Agents

Yaroslav, an Agentic Engineering Architect at SoftServe, has built GalaxyAtlas, a browser-based 3D galaxy containing roughly 190,000 star systems and one million procedurally generated planets. Players can fly spaceships, land on planetary surfaces, explore worlds, customize planets, and encounter creatures, with the main goal being a search for an ancient artifact. The project is presented as an experiment in changing the software engineering process itself by relying on AI agents. The author describes using real astronomical data as a basis and applying Three.js optimization techniques such as lazy generation and level-of-detail rendering to keep performance manageable. He also outlines a system of specialized AI subagents and reusable skills covering procedural generation, rendering, gameplay, content, and testing.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

LoRA: Why a Rank-8 Weight Update Can Be Enough for Fine-Tuning

Fine-tuning very large models usually requires storing a full new copy of billions of parameters for each task. LoRA, introduced by Hu et al. in 2021, sidesteps this by freezing the pretrained weights and training two small low-rank matrices whose product approximates the needed update. The approach is motivated by the observation that the weight changes from adaptation tend to lie in a few directions, so a small rank such as 8 can suffice. On GPT-3 175B, the authors report roughly 10,000 times fewer trainable parameters and about 3 times less training GPU memory than full fine-tuning, with comparable or better quality. The low-rank matrices can also be merged back into the base weights after training, so inference incurs no extra latency.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

F.E.R.P.A. Grader: Stateless AI Essay Feedback Designed to Keep Student Data Private

A developer has released F.E.R.P.A. Grader, a subscription-ready micro-SaaS for private tutors, essay coaches, and freelance educators. The tool addresses the legal constraints educators face under student privacy laws such as FERPA, which can prohibit pasting student work into commercial AI tools that log and train on user data. It provides rubric-based essay critiques through a backend that keeps text in memory only for about three seconds before purging it, with no database or user authentication. The app runs on open-weight models via OpenRouter, and its complete source code is publicly available under the MIT License.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Guiding Coding Assistants to Locate Code in Large Repos with Search Tools

A DEV Community workshop by AWS developer advocate Saurabh Rob Dahal teaches how to investigate code in unfamiliar repositories. Participants work through a prepared TypeScript bug where a client set to zero retries still sends three requests after a transport failure. The exercise uses repository mapping, ripgrep, ast-grep, and symbol navigation, with an evidence note to check an assistant's explanation. It requires Git, Node.js 24, ripgrep, and an editor, with no AWS account needed. The approach is presented as applicable beyond this bug, to settings, wrappers, and errors from other packages.

0 comments Read more at DEV Community
← NewerPage 395 of 3105Older →