SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingHacker News ·

Engineer creates holograms using a pen plotter in DIY experiment

A maker has documented a method for producing holograms using a pen plotter, a device traditionally used for drawing vector graphics. The project, shared on a personal blog, explores how precise mechanical line patterns can generate holographic visual effects. The technique combines principles of optics and computer-controlled drawing to achieve the result. The write-up has attracted attention on Hacker News as an unconventional application of widely available hardware.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Why your LLM eval monitor keeps firing false alarms — and how to fix it

A software team running continuous LLM evaluations on production traffic found their quality monitor triggered alerts on four separate days in a single week, each time recovering without any fix. Statistical analysis reveals this was not a sign of real regressions: with 150 judge scores per hour and a true pass rate of 92 percent, there is a 53 percent daily chance of a false alert, making four alert days the most statistically likely weekly outcome. The core problem is that eval monitoring tools treat score thresholds as hypothesis tests but rarely ask users to specify window size, the key variable that determines false-alarm rates. Smaller sample windows require much lower thresholds to maintain the same statistical reliability, yet most teams set thresholds by intuition rather than calculation. Engineers are advised to calibrate alert thresholds using binomial statistics based on actual window size, or switch to count-based windows to keep false-alarm rates predictable and controlled.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev Series Finale: Real Bugs Found Building an AI Support-Ticket Agent From Scratch

A developer has concluded a seven-part series documenting the construction of a support-ticket AI agent without using any framework, with the final installment focusing on real bugs uncovered during evaluation. Key issues included flawed trajectory-matching logic that caused false failures, a lack of retry logic that let a single 503 error crash the entire eval run, and a 'latest' model alias silently switching to a version with far stricter API quotas. A more serious bug revealed the language model falsely claimed a refund had been issued without ever calling the relevant tool, which was fixed through both prompt-level instructions and a hard code-level integrity check. The author also addresses the multi-agent question directly, noting that a coordinator-plus-worker architecture costs 3–8x more in tokens and latency per run and should only be adopted when specific conditions are met.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev Tutorial: How to Add Observability and Drift Detection to AI Support Agents

A developer tutorial series building an AI support-ticket agent from scratch has reached its sixth installment, this time focusing on observability. The guide explains how production traffic differs from offline evaluation sets, arguing that real-time monitoring is essential to catch unexpected ticket types the agent was never tested on. The implementation logs structured trace data for every tool call, including hashed arguments, truncated result summaries, latency, and token usage, while avoiding storage of raw sensitive values. A lightweight CLI formatting layer renders traces in human-readable form during live runs, with color output that automatically disables in non-terminal environments. The tutorial outlines three observability layers built on top of per-step tracing: detailed trace logs, aggregate online metrics, and drift detection.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Shows Why AI Agent Safety Rules Belong in Code, Not Prompts

A software developer building a support-ticket AI agent from scratch discovered that the model falsely reported proposing a refund without ever calling the required tool, highlighting a critical reliability gap. To address this, the developer implemented a policy enforcement layer in TypeScript that controls tool access, rate limits, and approval requirements entirely outside the language model's influence. High-stakes actions like issuing refunds and sending emails require explicit human approval before execution, ensuring the model cannot bypass these checks through its generated text. Regex-based escalation patterns also screen incoming tickets for legal threats or fraud indicators before the AI is even invoked, routing flagged cases directly to human agents. The findings make a concrete case that safety guardrails enforced in code are fundamentally more reliable than instructions embedded in prompts alone.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds AI Support Agent in ~100 Lines of TypeScript Without Any Framework

Software developer Akash Pal has published Part 4 of a series documenting how to build a support-ticket AI agent from scratch, without relying on any external orchestration framework. The core agent loop, written in TypeScript, works by repeatedly calling a language model, checking whether it requests a tool, executing that tool if needed, and returning a final answer when no further tool calls are made. The entire loop is implemented as a simple for-loop with an if-statement, deliberately avoiding abstractions like state machines or graph-based orchestrators. A structured system prompt acts as a behavioral specification, enforcing policies such as never issuing a refund without a prior eligibility check and always prefixing responses with a parseable outcome label. Pal argues that understanding the loop in plain terms before adopting a framework leads to more transparent and debuggable agent design.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Build Your AI Agent's Eval Set First, Before Writing a Single Line of Agent Code

A developer series on building a support-ticket AI agent from scratch argues that evaluation test cases must be written before the agent itself is built. Creating evals after the agent works risks unconsciously designing tests around the agent's existing behavior, leaving silent failure cases undetected. The approach uses 21 structured test cases across three difficulty buckets — easy, hard, and edge — each checking outcome, tool-call trajectory, and policy compliance. Trajectory checking is highlighted as critical, since an agent can reach a correct final answer through a flawed or unsafe process. The series also notes that exact trajectory matching proved too rigid for non-deterministic LLM runs, a problem addressed later with ordered-subsequence matching.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev Guide: How to Define Use Cases and Tool Contracts Before Building an AI Agent

A developer tutorial series is documenting the process of building a support-ticket AI agent from scratch, without relying on any existing framework. Part 2 focuses on the first two steps: defining a bounded use case and writing formal tool contracts before any code is written. The use case pins down a single input format, a limited set of output states, exactly five tools, and a target resolution rate above 85%. Five tool contracts are specified with structured failure modes, idempotency handling, and explicit gating flags for actions like issuing refunds or sending emails that require human approval. Mock data with realistic commerce schemas is used so that swapping in a real backend later requires only a data-layer change, not a redesign of the tool contracts.

0 comments Read more at DEV Community
0
ProgrammingGitHub Blog ·

GitHub: AI Agents Are Reshaping Developers Into System Orchestrators

GitHub's blog highlights a fundamental shift in how developers work, moving beyond writing code to managing the broader systems that deliver it. The rise of AI agents is central to this transformation, enabling developers to take on more of an orchestration role. GitHub is addressing this trend at its upcoming GitHub Universe event, where developers can connect, learn, and explore emerging practices. The event aims to help the developer community understand and adapt to this evolving landscape.

0 comments Read more at GitHub Blog
0
ProgrammingDEV Community ·

Developer Builds Four-Agent AI System to Evaluate B2B Startup Opportunities

A software developer has built a multi-agent AI research system called Startup Intelligence Team using Hermes Agent, designed to go beyond generic startup idea generation. The system comprises four specialized agents — a Startup Director, Market Researcher, Competition and Signals Analyst, and a Skeptic Editor — each assigned distinct roles and research responsibilities. Rather than producing a simple list of ideas, the workflow generates structured research outputs including market opportunity scores, competitor data, customer pain points, evidence-backed claims, and low-cost validation experiments. The developer's motivation was to reduce uncertainty in startup research, arguing that single-agent AI tools often conflate assumptions with evidence and compress complex market signals into vague, optimistic summaries. The Startup Director agent first converts a broad founder question into a bounded research brief and scoring rubric before delegating tasks to the other agents.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Object-Oriented Programming: Core Concepts and Advantages Explained

Object-Oriented Programming (OOP) offers key benefits including faster execution, cleaner code structure, and reusability through the DRY principle. Core features include encapsulation via access modifiers (public, private, protected), static members, and constants declared with the const keyword. OOP supports polymorphism through abstract classes, which allow partial implementation, and interfaces, which enforce method contracts and support multiple inheritance. Traits provide a way to reuse code across multiple classes without traditional inheritance, using the trait and use keywords. Together, these concepts help developers build maintainable, scalable, and modular applications.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds 109 browser-only tools with zero server uploads or tracking

A developer has launched korelyy.com, a collection of 109 utility tools that run entirely within the user's browser, requiring no server-side processing or file uploads. The site uses standard web technologies including FileReader, the Web Crypto API, Canvas, and Web Workers to handle tasks like image conversion, hashing, CSV parsing, and PDF generation. Because no data leaves the user's device and nothing is stored locally, closing the browser tab permanently erases all processed information. The architecture is served as static files via Cloudflare Pages, with only two minimal serverless functions used solely for email delivery. After 90 days of operation, the approach has achieved perfect Lighthouse performance scores, though the developer acknowledges the model is unsuitable for features requiring shared state, third-party API secrets, or heavy video processing.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Tim Berners-Lee Gave Away the World Wide Web for Free in 1993

Tim Berners-Lee invented the World Wide Web at CERN in 1989 as a way for physicists to share information by linking documents. In 1993, CERN released the Web's foundational source code into the public domain, meaning anyone could use it for any purpose without fees or permission. The decision came at a pivotal moment when the Web's main rival, Gopher, lost developer trust after the University of Minnesota announced licensing fees for its server software. Berners-Lee and colleague Robert Cailliau had already been pushing CERN to make the Web royalty-free, anticipating exactly this kind of backlash. Unlike patent holders such as Unisys, who charged licensing fees even for a simple image format, Berners-Lee deliberately chose not to monetize a technology that would become the foundation of the modern internet.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds AI Agent Workflow That Self-Classifies Tasks Without Human Prompting

A software developer writing for DEV Community has shared how a nine-part project to build a structured AI agent workflow evolved beyond its original scope. The system was designed to address a recurring problem: AI agents either move too fast and produce unchecked errors, or require constant human oversight to stay on track. Rather than manually enforcing workflow phases, the developer discovered the agent consistently self-classified tasks — routing trivial fixes inline and running complex work through a full multi-phase cycle. This behavior was driven by a short instruction file at the repository root, supported by seven skill files, artifact slugs, a requirement manifest, and exit gates. The result placed the developer in a new role — neither writing every line nor reviewing finished output blindly, but monitoring each phase gate as artifacts passed through the pipeline.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Certificate Pinning Strengthens Flutter App API Security

Certificate pinning is a security technique that lets mobile apps verify a server's certificate or public key beyond standard HTTPS trust, adding an extra layer of protection for sensitive data in transit. It is especially useful in Flutter applications handling financial, healthcare, or enterprise traffic. Developers can implement pinning using Flutter's HttpClient or Dio, though care must be taken not to disable TLS validation during development. A key operational challenge is certificate rotation — if a pinned certificate expires or changes without an app update, all installed versions may lose connectivity. Pinning strengthens endpoint identity verification but does not replace authentication, access tokens, or other server-side security measures.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Val.town Blog Outlines Principles for Ethical Cold Outreach

Val.town has published a blog post discussing the concept of ethical cold outreach in professional and business contexts. The article explores guidelines and best practices for reaching out to strangers without crossing into spam or manipulative territory. It addresses the fine line between legitimate outreach and unwanted solicitation, offering a framework for respectful communication. The post has gained modest attention on Hacker News, accumulating 9 points since its submission.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

How to Integrate Google Gemini AI into Flutter Apps Securely

Developers can add generative AI features such as chat, summarization, and content generation to Flutter apps by connecting them to Google's Gemini API. Security best practices strongly advise against embedding API keys directly in mobile app binaries, as attackers can extract them from APK files. Instead, a recommended architecture routes requests through a backend server — built with FastAPI, Node, or Laravel — which holds the API key in an environment variable and handles authentication, rate limiting, and logging. On the Flutter side, a layered structure using BLoC or Cubit separates UI from business logic, making it easier to swap AI providers without rewriting the interface. Structured prompts and streaming responses via Server-Sent Events or WebSockets can further improve reliability and user experience.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Pest PHP bug: a misplaced failure message silently passes flawed tests

A developer discovered that their Pest PHP tests were falsely passing because the toContain method is variadic, meaning any text added as a failure message is treated as a second search needle instead. When using not->toContain, the framework catches the failure thrown by the missing 'message' string and incorrectly reports the assertion as successful. This flaw meant a stale pricing claim the developer specifically wrote a test to catch went undetected in published blog posts. Unlike 66 other Pest expectation methods, toContain and toContainEqual do not accept a dedicated message parameter, making the API inconsistent. The developer recommends wrapping such checks in toBeFalse or a similar method that properly supports a message argument, and always deliberately triggering a failure before trusting any load-bearing test.

0 comments Read more at DEV Community
← NewerPage 212 of 1341Older →