SShortSingh.
0
IndiaTimes of India ·

Siddaramaiah skips Rahul Gandhi's Diwali Milan, sparking fresh speculation

Karnataka Chief Minister Siddaramaiah did not attend Rahul Gandhi's Diwali Milan event, a absence that has triggered renewed speculation in political circles. Observers have linked the no-show to an ongoing probe, though the reasons for the skipped appearance have not been officially confirmed. The Diwali Milan gathering is a customary occasion for party leaders to come together during the festive season. The development has drawn attention to the relationship between the state leadership and the national party leadership. Further clarity is awaited on both the event absence and the status of the investigation.

0
WorldBBC World ·

Houthis report at least 25 killed in Saudi air strike in Yemen

The Houthi movement says at least 25 people were killed in an air strike attributed to Saudi Arabia. Saudi Arabia leads a military coalition fighting the Houthis in Yemen. The article provided offers no further details on the location, timing, or the victims. Saudi Arabia has not publicly commented on the reported strike. Independent confirmation of the casualty figures was not available in the source material.

0
IndiaTimes of India ·

Allahabad HC orders release of man detained by Bijnor police, criticises UP Police

The Allahabad High Court has directed the release of Aausad Ahmad, who was held in custody by Bijnor police. The court found he had been detained without following proper arrest procedures. Judges rejected the police's explanation for the absence of CCTV footage as unconvincing and pointed to a recurring pattern of similar incidents. The bench also expressed displeasure with the station house officer for not producing the required evidence. The ruling raises broader concerns about how the police treat fundamental rights and civil liberties.

0
ProgrammingHacker News ·

Apple Keyboard Reportedly Blocks Swipe Typing for Certain Offensive Phrases

Reports indicate that Apple's iPhone keyboard contains a list of phrases that cannot be entered using swipe typing. The restriction was highlighted in a Guardian article published in November 2026, which described the so-called grey list. The article notes that users attempting to swipe certain terms may find the iPhone unhelpful in spelling them out. The source material provided is limited to a link and basic discussion metrics from Hacker News, so the full scope and exact contents of the list are not confirmed here.

0
TechnologyThe Verge ·

UV-Tracking Necklace Gem Prompts Reassessment of Sun Exposure and Health

A reviewer wearing The90's Gem, a $149 UV-tracking necklace, tested its warnings during a hot August outdoor party in Los Angeles. Sheltered under a patio umbrella with a fan, she still began feeling exhausted, dizzy, and nauseous, symptoms consistent with heat exhaustion. At the same time, the companion app alerted her that her UV exposure had risen significantly. The experience led her to consider how sunlight affects overall health, not just skin appearance. The full article details the explanation for how UV radiation was involved.

0
ProgrammingDEV Community ·

osFoundry adds Claude Code and Codex as @mention agents in team chat channels

osFoundry's Message feature now lets users add a coding agent such as Claude Code or Codex to a channel, where it appears as a member with its own handle like @claude. When someone mentions it, the agent starts on the user's own machine and works in a disposable git worktree of the channel's repository, reading recent channel messages for context. It then posts its answer back to the channel and can push changes with a pull request link. Users can choose whether the agent starts fresh for each mention or stays open to remember earlier conversation, and it runs only on the desktop app because it executes locally.

0
ProgrammingDEV Community ·

Commitment Lab: a practice room for untangling stakeholder promises across teams

The developer built Commitment Lab for the Flip the Switch hackathon as a personal practice space for handling conflicting expectations between teams. Learners work through a fictional Friday launch dispute involving Sales, engineering, and a customer, using a shared Switch chat room with AI agents. The exercise aims to clarify which statements are agreed commitments, which are only proposals, and which remain open. The recorded demo was run by a single operator playing both learner and mentor, so it shows role-play with two agents rather than verified independent mentor participation or a validated training outcome. Switch connects the agents to a Mattermost channel, with the coach agent reviewing the conversation afterward to produce a debrief and commitment record.

0
ProgrammingDEV Community ·

Sharding: How Splitting a Database Across Machines Works and Its Trade-offs

A single database server eventually hits limits on disk, CPU, and write capacity, and read replicas cannot relieve write load. Sharding splits data across many machines, each owning a slice, and is considered the last and most costly scaling option. The choice of shard key is critical: it must spread load evenly while also appearing in common queries, or the system suffers from scatter-gather queries or concentrated hot spots. The approach introduces costs including cross-shard queries, a routing layer to locate data, rebalancing when data moves, and distributed transactions. Engineers are generally advised to exhaust other optimizations before sharding.

0
ProgrammingDEV Community ·

Code Court: AI agents argue a pull request before human reviewers vote

A developer built Code Court, a project for the Flip the Switch Hackathon that stages a mock trial of GitHub pull requests inside a Discord channel. Three AI agents run on Claude Opus 5.5: a judge who opens the case and rules, a prosecutor who files charges with severity and line numbers, and a defense that concedes, mitigates, or contests each one. Human team members act as the jury, voting to accept or overturn the ruling before anyone merges. In a demo on a sample coupon-discount PR, the prosecutor flagged missing input validation that could yield negative totals, a false claim that existing callers were unaffected, a half-cent rounding error, and absent tests. The defense conceded the validation hole and offered a fix for the rounding issue.

0
ProgrammingDEV Community ·

FlakeTrail Uses Investigation History to Improve CI Failure Diagnosis

Identical CI error messages can stem from different root causes, so an AI assistant that suggests the same fix every time may mislead developers. FlakeTrail instead analyzes each current failure alongside historical investigations retrieved through Hindsight, a memory layer. Its reasoning module separates verified fixes, failed attempts, merged-pending changes, and unconfirmed hypotheses, so recommendations reflect what actually happened before. Recommendations change depending on which kinds of evidence are available, giving developers context on previous attempts and remaining open questions.

0
ProgrammingDEV Community ·

Offboard tool aims to close access gaps left when team members depart

After a developer left a team, their GitHub, Apple, and Vercel accounts remained active, showing how departures can be incomplete even when the calendar says otherwise. The author built Offboard to discover leftover access, have a human approve each exact change, remove it, and independently verify the result. In a controlled test, a separate GitHub account was investigated within a shared Mattermost room, with a Codex agent using limited tools that cannot approve or remove access. The agent's read-only scan identified the account's organization membership, and the dashboard displayed the same findings and decisions. Approval is handled through a private, single-use email link that expires after 15 minutes and requires explicit confirmation from the person affected.

0
ProgrammingDEV Community ·

A terminal workflow for reviewing code written by AI agents like Claude Code and Codex

As tools such as Claude Code, Codex CLI and Aider generate more code, developers increasingly spend their time reading diffs rather than writing them. The author argues that line-by-line diffs in the order an agent touched files are a poor way to review such changes. The suggested approach is to review whole functions, start with the riskiest changes, and have the agent explain the decisions behind its edits. The author presents Klaub, an open-source tool that ranks changed symbols by priority using a deterministic scoring system and lets reviewers query the agent about what is on screen.

0
ProgrammingDEV Community ·

CI gate passed without checking anything until ripgrep was installed on the runner

A developer found that a red badge on their repository was hiding a deeper problem: a fixture test needed ripgrep, which GitHub's Ubuntu runner lacks, so it failed only in CI. Because GitHub Actions stops a job at its first failed step, later checks had been silently skipped for weeks. A documentation gate that verifies search commands also skipped itself with a passing exit code whenever ripgrep was missing, and once it ran, it exposed three commands that broke on the ripgrep 14 version Ubuntu installs. Another check then found a step-count mismatch in the contributing guide, which the author corrected. A final fix addressed a script that ran commands in a supposedly empty temp directory but left variables unset, so searches expanded to the whole filesystem and passed or failed depending on file traversal order. The author concludes that skipped checks should not appear as passing in CI, and that running all steps locally at once surfaces every failure faster than repeated pushes.

0
ProgrammingDEV Community ·

How a Browser-Only Search Handles Arabic Spelling, Hamza and Prefixes

A developer describing a library of 200 free browser tools explains how they built search covering both English and Arabic pages. Arabic presents problems that simple substring matching cannot handle, such as variant spellings of taa marbuta, attached article and preposition prefixes, and vowel marks. The solution folds all text into a canonical form, handling hamza, digit variants, and sound-alike letters. It then applies light prefix and suffix stemming, which is sufficient for a short list of tool names. The entire approach is roughly 200 lines of plain JavaScript running in the browser, without a library, server, or index file.

0
ProgrammingDEV Community ·

Developer Explains Redacting Secrets From CI Logs Before External Processing

A FlakeTrail developer describes a preprocessing step that removes credentials from CI logs before they are sent to external AI or memory services. CI logs often mix useful error details with sensitive data such as tokens, API keys, passwords, and connection strings. Rather than discarding entire logs, the tool applies pattern-based redaction inside its parser to keep relevant debugging context like stack traces and error messages. The parser also extracts structured fields and limits excerpt sizes to keep processing focused. The author stresses that regex-based redaction is only a first layer and cannot guarantee detection of every secret.

0
ProgrammingDEV Community ·

Claude dapat mengusulkan perubahan kode GitNest yang harus disetujui di osFoundry

Panduan ini menjelaskan cara menghubungkan percakapan Claude ke osFoundry agar Claude bisa membaca repository GitNest dan menyiapkan perubahan seperti task, commit, change set, atau pull request. Claude tidak dapat menulis langsung ke repository, sebab setiap tindakan masuk sebagai usulan di bagian Needs your approval dan baru dijalankan setelah pemilik koneksi menyetujuinya. Repository harus dimasukkan ke project tempat koneksi berada, dan hanya pemilik repository yang bisa memilih project tersebut. Di pengaturan Tools, setiap tool bisa diaktifkan atau dinonaktifkan, namun menghapus centang hanya mencegah Claude memintanya tanpa menghilangkan kewajiban persetujuan. Token penyedia tidak pernah dikirim ke Claude, dan akun GitHub atau GitLab hanya dipakai ketika pull request disetujui.

0
ProgrammingDEV Community ·

Developers Build AI Agents with Long-Term Memory Using LangGraph

Many AI assistants forget user details once a conversation ends, which limits how useful they can be over time. A developer team working on an agent project set out to create one that could recall past interactions and draw on relevant information when making decisions. They used LangGraph to split the agent's work into connected workflow steps, such as retrieving memories, processing input, and generating responses. The approach separates short-term memory, which holds context within a conversation, from long-term memory, which keeps useful details after it ends. The team's design aims to store only information that matters, rather than every exchange.

← NewerPage 479 of 7352Older →