SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

O'BLOCK dev borrows Balatro's UI feel, not its visuals, after one player's bug list

The developer of O'BLOCK, a free Steam sneaker-reseller game, rebuilt the interface overnight after a tester listed usability problems. The list covered a scroll wheel that moved the selection, a card drawn beneath text, a hard-to-read price bar, and a custom sneaker price stuck at 1.1k. The team's guiding rule was to copy interaction behavior from Balatro, such as hover and press responses, while keeping every on-screen object drawn from the game's own world. Across five builds shipped in one night, the street screen grew from 8 to 27 visible shoeboxes, and the update touched 18 source files. The developer published a before-and-after dataset alongside the Build 18 patch notes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why local LLMs give different answers at temperature 0

Setting temperature to 0 and fixing the seed does not guarantee identical outputs from a local language model. A test with llama.cpp on the gsm8k benchmark changed the generated text on 41 of 100 problems when requests were run one at a time versus eight at once. Other studies cite similar results, including 80 different completions from 1,000 identical temperature-0 requests and accuracy swings of up to 9% when only hardware or batch size changed. The root cause is that floating-point addition is not associative, so differences in GPU kernel reduction order, batching, and prompt caching alter the computed probabilities slightly. Issue trackers for several popular inference engines report the same behavior, making it a broad rather than engine-specific problem.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Tests Show Prompt Design Shapes How Reliably AI Handles Customer Support Emails

A learner pursuing a career as an AI Solutions Consultant tested three prompts with Claude on the same customer complaint about a delayed headphone order. The scenario involved a package stuck in transit past its delivery date, repeated unanswered support contacts, and an urgent need for the product before an interview. The first, basic prompt produced an empathetic reply but assumed refunds and replacements were available without knowing company policy. A more detailed prompt improved structure yet still proposed options requiring verification. The third prompt, which demanded separation of confirmed facts from unknowns and barred unsupported promises, gave the most responsible answer, though the author noted it could sound warmer.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Reconciling Azure and AWS CIDR ranges during a cloud migration

CIDR planning for a migration differs from greenfield design because the target VPC must coexist with a source Azure VNet whose address space was set by earlier, unrelated decisions. The Azure ranges are fixed constraints, so they need to be audited across every VNet in scope before any AWS resources are provisioned, including VNets from acquisitions or independent team deployments. Overlap should be avoided even when no hybrid connectivity is planned, since rollback validation, phased transitions, and future mergers can all require the two networks to communicate. The recommended approach is a single table of every range across both clouds, with the proposed AWS range validated against all rows, including VNets that will stay on Azure. The MGN staging subnet should also be a deliberate, separate tier in the AWS design rather than placed in whatever space happens to be available.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Anthropic Donates Agent Skills Standard to Linux Foundation's Agentic AI Foundation

Anthropic has announced it is contributing its Agent Skills format to the Agentic AI Foundation, a body under the Linux Foundation that also oversees the Model Context Protocol. The announcement, made on October 22, 2026, cites more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, that have adopted skills since the standard was published in December 2025. Reported usage figures include one million published skills and nearly 280 million installations. A skill is a folder built around a SKILL.md file with required name and description metadata, optionally bundled with scripts, references and assets. The format loads progressively so agents see only brief metadata until a task requires the full instructions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

BeforeRunIt adds live plain-English command explanations under Linux shell prompts

A developer has released BeforeRunIt, an open-source tool that displays a real-time explanation beneath the Linux terminal prompt as the user types a command. It supports Bash, Zsh and Fish, and classifies commands as KNOWN, CAUTION, DANGER or UNKNOWN. The tool relies on built-in local rules, with optional AI-generated explanations for unfamiliar commands. The creator stresses that it is an early project and not a security sandbox or a guarantee of command safety, and is seeking feedback on rule accuracy, false positives and shell compatibility.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Startup Launches Pay-As-You-Go AI Support Bot Trained on Markdown Docs

A small team behind the AheadOdds product built AI Doc Sherpa to handle visitor questions without hiring a support agent or paying monthly chatbot fees. Site owners upload a single Markdown documentation file and embed a snippet that adds a floating support widget, with setup taking under two minutes. The bot answers only from the uploaded documents and states when information is unavailable, while unanswered questions are logged for free as documentation gap insights. Pricing is based on prepaid packs of 2,000 answered questions for $20 that never expire, with no fixed monthly fee, and new accounts receive 100 free answers without a credit card.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why SaaS Onboarding Determines User Retention and How to Fix It

Many SaaS products lose a large share of new users right after their first session, often because onboarding feels confusing or overwhelming. The article argues that Time to Value, the speed at which a user reaches their key "aha" moment, is the most important metric for retention. It recommends reducing cognitive load through progressive disclosure and tailoring onboarding paths to different user goals with a few signup questions. Practical tactics include using sample data in empty states, triggering contextual tooltips instead of static tours, and adding gamified progress checklists.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Gydmation AI Turns Plain-English Requests Into Working n8n Automation Workflows

Hacroo Technologies has launched Gydmation AI, a tool that converts plain-English automation requests into functional n8n workflows. It asks clarifying questions when details are missing, builds from more than 1,100 node definitions and 10,000 templates, and validates and tests workflows against sample data before deployment. Users receive step-by-step guidance for setting up credentials and can deploy to their n8n instance in one click or download the JSON. The app runs locally via Docker, storing prompts, workflows and credentials on the user's machine, and offers a free plan with bring-your-own AI key alongside a $29 monthly Pro tier.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Attackers Use AI Agents to Breach Thousands of Microsoft 365 Tenants and Other Targets

In about ten days, attackers directed AI agents at ConnectWise, a managed-services provider, and reached more than 2,250 downstream Microsoft 365 tenants, with at least 162 confirmed breached. Separately, a single operator using a commercial agent platform and roughly $5,000 in AI tokens reportedly compromised over 100 companies. The LiteLLM AI gateway also suffered a supply-chain attack on PyPI and Docker Hub, where a credential-stealing release was available for about two hours and downloaded around 49,000 times. The article argues these incidents show AI lowering the cost and speed of intrusions, with basic defenses like MFA and credential hygiene separating victims from survivors.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

LessWrong Post 'I Did This Without Your Knowledge' Draws Attention on Hacker News

A post titled "I Did This Without Your Knowledge" has been shared on Hacker News, linking to a page on the LessWrong forum. The submission has received 10 points and one comment so far. The provided material includes only the article and discussion links, the title, and engagement figures. Its specific content and subject matter are not described in the available text.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

Essay Argues for Slowing Down to Fix Existing Software Problems

A blog post titled "Move Slow and Fix Things" by developer Matthias Endler was shared on Hacker News, where it gained 79 points. The essay appears to advocate prioritizing careful repair of existing software over rapid new development. The provided material includes only the article link and discussion metadata, so the specific arguments and examples cannot be confirmed here. Readers are directed to the original post for the author's full reasoning and recommendations.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Why Rewriting a JSON File in Place Can Leave Readers With Broken Data

A developer's page-building script repeatedly failed with JSONDecodeError because it read a JSON index while another script was rewriting it. Testing on Linux with Python 3.10.12 showed that open(path, 'w') truncates the file before new data is written, so readers often saw empty or partial content. Killing the writer mid-write also left a corrupted fragment with no intact version to fall back on. Writing to a temporary file and then swapping it in with os.replace eliminated the broken reads in the same test. The author notes three related pitfalls to watch for: a temp folder on a different filesystem, file permissions, and symlinks.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Unpinned CI actions and runaway concurrency prompted a zoned pipeline redesign

A team's CI pipeline suffered two costly failures in one quarter: a mutable action tag that was silently changed to push commits, and a concurrency misconfiguration that queued over 1,100 jobs and exhausted weeks of runner quota. To address these, they split the pipeline into an untrusted pull request zone with read-only permissions, no secrets, and aggressive cancellation, and a trusted publish zone limited to main that inherits secrets and never cancels mid-flight. Third-party actions are now pinned to commit SHAs with version comments for readability. The redesign also included disabling persisted checkout credentials and keying the browser cache on the lockfile, plus an agent smoke test to catch session-handoff failures that unit tests miss.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

agent-doctrine: Open-Source Prompts for Getting AI Agents to Ship Tested Code

The sisig.ai team has open-sourced agent-doctrine, a set of prompts, role definitions and skills meant to make AI coding agents produce tested, mergeable pull requests. Every agent follows the same pipeline of scouting, gating, planning, implementing, reviewing, consolidating, verifying and shipping. Before any edit, the lead agent must answer fifteen questions from a TENETS.md file with evidence, such as the smallest correct change and the blast radius. Users choose a fast mode, which keeps the gate and per-diff review, or a rigorous mode, which adds plan simplification, adversarial review and approval before code. The project is MIT-licensed, requires Git and Bun, and rewrites global agent instructions, so users are advised to back up their configuration first.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Claude 3.5 Sonnet, GPT-4o-mini, or Gemini 1.5 Pro: Testing Chatbots for E-Commerce Support

A developer guide argues that the best chatbot API for e-commerce support should be chosen by testing with real support tickets rather than relying on generic leaderboards. It recommends Claude 3.5 Sonnet for the most natural conversational quality, GPT-4o-mini for faster, budget-friendly deployments, and suggests testing Gemini 1.5 Pro before deciding. The author proposes scoring label accuracy, escalation decisions, required output format, and tone separately, and running each test more than once to avoid relying on single responses. The article also advises designing the API boundary so the chosen model can be swapped without rewriting the triage workflow.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

MCP Security Gaps: Four Attacks That Server Audits Miss

Standard MCP security checklists focus on OAuth, token scopes and TLS, which confirm server configuration but not whether an AI agent will follow malicious text. Because agents treat tool names, descriptions, schemas and outputs much like system instructions, an attacker can hide commands in that text and steer the agent's real permissions. The article outlines four attack types: tool poisoning, rug pulls, tool shadowing and poisoned tool outputs. It cites OWASP's beta MCP Top 10, which lists tool poisoning as MCP03, and the MCPTox benchmark, where the worst-performing of 20 tested agents followed poisoned instructions 72.8% of the time. The author recommends continuous red teaming in staging, with each run scored pass or fail and repeated on every release.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer details auto-trading Discord trading calls without a bot token

A developer describes the difficulties of building a self-made Discord bot that places broker orders from trade calls posted in a trading server. The problems include permission limits, inconsistent message formats, false-positive noise, the need for always-on hosting, and matching exit signals to open positions. The author, who builds the product Nyria, says its Discord bot needs only View Channels permission and no token or self-hosting. The walkthrough covers connecting Discord, scoping a strategy to one channel with author and keyword filters, and training the parser on past messages. The setup starts with paper trading before any real orders are placed.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Opus 5.5 and Sonnet 5.5 tie on coding-agent tests, but Opus costs about 2.5x more

A test of Claude Opus 5.5 against Claude Sonnet 5.5 on six small Python coding tasks found both models passed every hidden test in every run. The evaluation ran Claude Code with identical settings, changing only the model, across 24 runs on 2026-10-20. Opus was consistently slower and, per task, cost roughly two and a half times as much as Sonnet at QuickSilver Pro prices. The tasks covered bug fixing, spec implementation, refactoring, input validation, a config parser and a Mustache renderer. The author's conclusion is that, for these tasks, Opus did not justify its higher price.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

78% of Thai PHP hosts on 8.1 or older may miss tonight's security patch

Mallory's Shodan scan counted 14,619 of 18,701 Thai hosts advertising a PHP version at 8.1 or older, roughly 78 percent, which will not receive the patch. PHP is set to release fixes at 21:00 Thai time for two vulnerabilities rated CVSS v4.0 9.2. The figure reflects banner-declared versions rather than confirmed exposure to the flaws. Reports circulating online cited a 10.0 score and disclosure dates that conflict with PHP's official advisory, and the CVE identifiers in the leak could not be found in the NVD. The author argues the bigger issue is legacy versions on low-cost hosting, not the patch timing alone.

0 comments Read more at DEV Community
← NewerPage 104 of 2927Older →