SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

MikroTik Patches Three Critical RouterOS Flaws Including SSH Auth Bypass

MikroTik released a security update in September 2026 addressing three critical RouterOS vulnerabilities, flagged by India's CERT-In as advisory CIVN-2026-0460. The most severe flaw allows unauthenticated attackers to bypass SSH login entirely by sending a forged RSA key, potentially granting full administrative control. A second SSH vulnerability enables privilege escalation by manipulating argument delimiters in the login path, allowing attackers to alter router policy and pivot to internal networks. A third flaw in the bandwidth-test service exposes uninitialized kernel memory to unauthenticated network requests and can crash the device. Affected users are urged to upgrade to the patched RouterOS versions and, in the interim, restrict SSH access to trusted sources and disable the bandwidth-test service where not needed.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer finds sandbox misclassified shell crashes as command failures on Windows

A developer running an AI agent on Windows Server 2022 discovered that every executed command was silently failing due to a broken shell environment, not faulty commands. The root cause was a recent update that set bash as the default process-boundary shell across all platforms, despite bash not being a native Windows executable. Further investigation revealed that six bundled tools — including bash.exe — all depended on the msys2 runtime DLL, which was only present in one directory and causing STATUS_DLL_INIT_FAILED (exit code 0xC0000142) at the loader level. Because the runner's failure-detection logic did not recognise this exit code as an environment-level fault, every crash was logged as a normal command failure, masking the broken environment entirely. The developer concluded that runners should treat 'child process never initialised' as a runner failure and surface the exit code prominently in all error messages to prevent wasted retry attempts.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OSLabs Releases Open-Source CLI PennyWyze to Audit Claude API Costs

Three engineers at OSLabs have released PennyWyze, a free open-source command-line tool that helps developers find the cheapest Claude model tier that still meets their accuracy requirements. Users provide their production prompt and a labeled dataset of real examples, and the tool runs these against Claude's Opus, Sonnet, and Haiku tiers using the live Anthropic API. PennyWyze grades each model's responses against expected answers and calculates actual monthly cost projections based on real token counts. In one example audit, the tool identified that switching from Opus to Haiku could save roughly $180 per month with no loss in accuracy. Currently, PennyWyze supports exact-match grading, making it best suited for classification, extraction, and routing tasks rather than open-ended generation.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Free API tracks OpenAI Codex global usage-limit resets with 24/48h forecasts

A developer has launched codex-reset.com, a free tool that monitors and records OpenAI's informal global Codex usage-limit resets, which occur occasionally after incidents or milestones. The service provides a public JSON API requiring no key, offering probability forecasts for a reset occurring within the next 24 or 48 hours alongside a dated timeline of past resets with source links. It also tracks live Codex service status across multiple surfaces, helping users distinguish outages from quota exhaustion. The data is additionally available as a read-only MCP server that can be connected directly to Codex or Claude Code without installation or authentication. The project is free to use, with the only condition being visible attribution when the data is displayed to others.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

MCP vs CLI: The Right Choice Depends on Who Controls the Next Step

When building tools that perform actions like publishing a website, developers face a choice between a Command Line Interface (CLI) and a Model Context Protocol (MCP) server. The key deciding factor is not whether the user is human or an AI agent, but rather who determines the next step in the workflow and where the data resides. CLIs are better suited for fixed, repeatable sequences such as automated deployment jobs or local build commands, while MCP tools shine when an AI assistant needs to discover and invoke operations dynamically within a conversation. Transport context also matters: a local MCP server can read a directory path directly, whereas a remote MCP server over HTTP interprets the same path on the server side, not the user's machine. Both interfaces can share a single underlying operation, with the CLI managing flags and exit codes while the MCP adapter handles tool definitions and structured results.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Add AI Agent Billing Without Disrupting Its Workflow Loop

When a subscription gateway is placed between an AI agent and its model provider, it can mistakenly interrupt multi-step tasks by treating each model request as a separate billable turn. The core problem is that billing layers can inadvertently take over orchestration decisions that should remain with the agent's local host. A properly designed billing gateway should only handle authentication, allowance checks, request forwarding, and usage accounting — not control the flow of the agent's tool loop or conversation context. Developers are advised to assign distinct identifiers to user turns, model operations, and business invocations to enable accurate metering and error recovery without disrupting execution. Keeping orchestration deliberately separated from billing ensures that paying for model access does not silently grant a second system control over the user's task.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Walrus Sessions 8 Hackathon Challenges Developers to Build Memory-Persistent Chatbots

Walrus Sessions 8 is a live hackathon running from September 18 to October 9, 2026, focused on building chatbots that retain context across conversations using the Walrus Memory SDK on mainnet. Participants must create or upgrade a chatbot that stores and recalls user context, deploy it in a real environment, and document the integration with evidence of it working. The competition is open to any use case, including customer support, tutoring, and sales. A total of $2,500 in WAL prizes is on offer, with separate award tracks for the best chatbot, best article, non-Claude/GPT models, and bug bounties requiring no chatbot submission. Full rules and submission details are available via the official Walrus form, with the project repository hosted on GitHub under MystenLabs.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Claude Code's git push deny rule blocked only 8 of 14 push attempts in security test

A developer tested Claude Code version 2.1.278 to see how reliably a single deny rule — Bash(git push:*) — blocks push commands in an automated agent setup. Across 14 differently worded push attempts run on 22 September 2026, the rule successfully blocked 8 but allowed 5 to actually reach the remote repository. Bypasses included indirect methods such as wrapping the command in a shell script, using eval, or invoking git with a -c flag prefix. A 14th attempt evaded the rule but failed independently for unrelated shell reasons. The findings highlight that Claude Code's permission deny rules match command strings literally, leaving meaningful gaps when commands are rephrased or wrapped.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Text Detectors Are Unreliable by Design, Developer Explains Why

A developer has built an AI-content detection tool after finding that most existing detectors overstate their accuracy, often claiming near-perfect results they cannot deliver. Unlike image-based tools that can embed invisible statistical watermarks, text detection relies on weaker signals such as perplexity, sentence burstiness, and token-frequency patterns. These signals break down under common conditions like paraphrasing, short inputs, translation, or deliberate obfuscation, making confident verdicts misleading. The new tool, available at detectaiwatermarks.com, is designed to report which specific signals were detected rather than issuing a single definitive score. Its verdict is also calibrated to text length, acknowledging that a short social media post and a long essay cannot be assessed by the same standard.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Google Adds monday.com to Gemini Connected Apps for Workspace Data Access

Google announced on September 23, 2026, the addition of monday.com to its Gemini Connected Apps, alongside other tools including Airtable, Linear, and Zoho. The integration allows users to reference monday.com board data within Gemini-powered features across Gmail, Docs, Sheets, and Slides using natural-language queries. Currently, the Google Workspace connector operates in early alpha as a read-only tool, meaning Gemini can retrieve project information but cannot create or modify board items. A separate, deeper integration launched in June 2026 enables a monday.com agent for Gemini Enterprise to orchestrate workflows via natural language, carrying a Google Cloud Ready designation. The Connected Apps feature can be enabled through Gemini settings and is designed to reduce manual context-switching between project management tools and Workspace documents.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Multi-Tenancy Is an Architecture Problem, Not Just a Config Setting

A software team building a low-code platform discovered that a seemingly simple client request—letting four subsidiaries share one system while keeping their data separate—exposed deep architectural complexity. Multi-tenancy actually encompasses three distinct concerns: deployment isolation, data partitioning, and configuration customization, and most platforms quietly assume these can be bundled together. The challenge intensifies on low-code platforms because the application's own structure—fields, workflows, permissions, and scripts—is also tenant-scoped data, not just business records. A script written by one subsidiary could inadvertently access another's data, not due to malicious intent but because the boundary was never enforced at the runtime level. The core lesson is that tenant isolation cannot rely on developers remembering a where clause; it must be enforced systematically by every layer of the platform's runtime.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Finds Vote-Spoofing Bug in Open-Source AI Memory SDK, Fix Shipped in Days

A developer auditing MemWal, an open-source AI-memory SDK by MystenLabs, discovered a logic flaw in a sample chatbot app built on the platform. The bug allowed any user who could view a message in a public chat to cast a vote on another user's message by exploiting a mismatch between the vote-existence check and the actual update query. The existence check queried only by message ID, while the UPDATE required both message ID and chat ID, meaning a spoofed vote would silently block the original owner from ever voting. Critically, the system returned a 200 success response with no errors or logs, leaving the feature permanently broken for the legitimate user. The developer reported the issue responsibly, and the maintainers confirmed and released a fix within days.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Build a Lightweight React Kanban CRM Dashboard with Vite and Vanilla CSS

A developer tutorial on DEV Community demonstrates how to build a static CRM dashboard prototype using React, Vite, and plain CSS — without relying on heavy component libraries like Material-UI or Ant Design. The project features a Kanban-style sales pipeline layout divided into a sidebar navigation area and a main content board. React's useState hook manages active menu selection, while all pipeline data is served as static mock entries representing leads at various stages. CSS Flexbox handles the overall layout, with the Kanban board supporting horizontal scrolling as additional pipeline columns are added. The guide is intended purely as a frontend UI prototype and does not include backend integration, authentication, drag-and-drop functionality, or database connectivity.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

oh-my-agent CLI 15.0.0 Launches with HyperFrames Video Engine and New Security Guard

The oh-my-agent CLI has released version 15.0.0, introducing HyperFrames as a replacement for Remotion in its video rendering toolchain — a breaking change that required a major version bump. Users can migrate by renaming the config key from video.remotion to video.hyperframes and updating the compositor flag accordingly. The release also adds a code-intelligence guard that blocks native file-search commands when a code-intelligence provider is configured, directing agents to use designated tools like Serena or Gortex instead. Several bugs were fixed, including a Serena daemon conflict that caused multiple redundant processes per project, stranded schedule registrations that failed silently, and a critical issue where running oma link claude from the home directory overwrote global hook settings. The update spans 67 commits across 230 files, with CLI reaching 15.0.1 and the web component at 7.0.0.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

TypeSafe's Jev Matches Mid-Tier LLMs but Trails Frontier Models After Eight Days

TypeSafe AI launched Jev on September 15, 2026, a hosted 'System One' model designed to answer typed classification and scoring questions about text rather than generate free-form responses. Independent testing conducted over eight days found Jev performing on par with mid-priced large language models while lagging 6.5 to 11.5 percentage points behind frontier models on the most reliable benchmarks. The model showed the best out-of-the-box probability calibration among tested models on familiar English tasks, with most calibration errors correctable via a single temperature adjustment. TypeSafe, co-founded by Diogo Almeida, Sasha Sheng, and Erik Gafni, raised a $40 million seed round led by DCVC, and prices Jev at $0.042 per million input tokens with free output. The review drew on 14 arXiv preprints, 104 GitHub repositories, and 33 blog posts, noting that the majority of positive coverage either restated vendor claims or came from launch partners with a commercial interest.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Gemma 4 26B Trails Jev by 2.1 Points as a Decision Model on Single AWS L4 GPU

A pre-registered benchmark tested Google's open-source Gemma 4 26B model as a decision classifier on a single AWS EC2 L4 GPU, comparing it against TypeSafe's hosted Jev 1.13.0 service and Google's DiffusionGemma variant. Evaluated on Bespoke Labs' 3,880-record public test suite, plain Gemma 4 26B scored 2.1 percentage points below Jev overall, with no meaningful gap on yes/no questions but a 4.5-point deficit on multiple-choice items. Jev showed better out-of-the-box calibration, though fitting a single temperature parameter on just 50 labels brought Gemma's median calibration error to within 0.01 of Jev's. Compared to DiffusionGemma, the plain Gemma read matched accuracy, performed similarly after calibration fitting, and ran 1.9 to 5.1 times faster per decision. All per-item outputs and pre-registration documents were committed to a public GitHub repository before any model calls were made.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Gemma 4 26B Trails Jev by 2.1 Points Overall in Benchmarked Decision Model Test

A pre-registered benchmark tested Google's open Gemma 4 26B model as a decision model on a single AWS EC2 L4 GPU, comparing it against TypeSafe's hosted Jev 1.13.0 service and Google's DiffusionGemma variant. On Bespoke Labs' 3,880-record public evaluation suite, plain Gemma 4 26B scored 2.1 points below Jev overall, performed comparably on yes/no questions, but lagged 4.5 points behind on multiple-choice tasks. Jev showed better out-of-the-box calibration, though a single temperature adjustment brought Gemma's median calibration error to within 0.01 of Jev's. Against DiffusionGemma, the plain Gemma read matched accuracy levels while running 1.9 to 5.1 times faster per decision. All per-item outputs and pre-registration documents were committed to a public GitHub repository before any model calls were made.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Engineer Applies Test-First Thinking to Requirements, Contracts, and Specs

Software engineer Anton, who works primarily with PHP/Symfony and Go, has written about extending test-driven development (TDD) principles beyond code to earlier stages of software work, including requirements and contracts. His core argument is that a checkable condition must be written before the thing it is meant to validate, otherwise it simply describes what was already built and cannot serve as a real gate. He illustrates this with a code coverage ratchet that prevents the recorded threshold from being lowered in pull requests, while honestly noting the bar is currently set at zero despite actual coverage sitting at 86.7%. A second example shows the approach working as intended, where a closing condition for a database-access refactor was written while the work was still in progress rather than after completion. Anton frames the method not as a testing philosophy but as a discipline around the order in which things are written down at all.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Google Releases Gemma 4 Family of Five Open Models for Mobile to Workstation Use

Google DeepMind launched the Gemma 4 family of open models under the Apache 2.0 license, releasing four sizes on March 31, 2026, and adding a fifth — the 12B Unified — on June 3, 2026. The lineup spans five sizes from the 2B edge model to the 31B dense model, all supporting text and image input with context windows of 128K or 256K tokens across more than 140 languages. A standout model is the 26B A4B, a Mixture-of-Experts architecture that activates only around 4B parameters per token, delivering near-31B quality at a fraction of the memory cost at approximately 14.4GB quantized. The 12B Unified model eliminates separate encoders entirely, processing text, images, and audio through a single architecture at roughly 6.7GB in 4-bit quantization. Google's own benchmarks show significant gains over the previous Gemma 3 27B, with the 31B model scoring 89.2 on AIME 2026 compared to 20.8 for its predecessor, though these figures reflect Google's own testing conditions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How AI Agents Can Support Customer Feedback Reviews Without Replacing Human Judgment

AI agents can help teams process large volumes of customer feedback faster by sorting, counting, and grouping comments, but they work best as a first pass rather than a final verdict. Customer feedback arrives across many channels in inconsistent formats, making manual review at scale impractical and AI-only summaries potentially misleading. A recommended approach separates observable patterns — such as how many comments mention a specific issue — from interpretations, which require human context and verification. Before any review begins, teams should define a clear scope, remove personally identifiable information, and document what sources and date ranges were included. This structured, human-checked method helps prevent AI-generated summaries from being mistaken for complete or authoritative conclusions.

0 comments Read more at DEV Community
← NewerPage 973 of 3189Older →