Skip to content
Tech News
clear
Topics: Today This Week This Month This Year

Logitech unveils $99 MX Keypad targeting coders and AI users

Logitech has launched the MX Keypad, a $99.99 accessory featuring nine customizable full-color LCD keys for triggering commands, macros, and AI prompts. It essentially strips down the company's pricier MX Creative Console, dropping the separate dialpad while keeping the keypad and halving the cost. The device was built in partnership with GitHub and ships with GitHub Copilot integration built in.

Author explores Git worktrees after seeing AI coding agents rely on them

A developer describes discovering git worktrees—a decade-old git feature they'd never used—after noticing tools like Claude Code create a separate worktree for each task so multiple AI agents can work on the same repository in parallel. The piece explains that worktrees let you check out multiple branches into separate working directories simultaneously, all sharing the same underlying repository data, unlike traditional branch switching which requires stashing and checking out.

Claude Code and Instinct run mobile AI agents inside dedicated microVMs

Coding assistants like Claude Code, Instinct, and Poke are shifting from local machines to cloud-hosted virtual machines so they can be accessed from phones. Inspection of Claude Code's setup shows it runs inside a Firecracker microVM with a custom Rust-based init process controlling the session over a vsock connection, while inference calls go out through a locked-down, MITM'd egress gateway to Anthropic's servers.

Benchmark tests 10 model-harness pairs on identical Three.js coding task

A developer ran the same prompt—building a self-contained sci-fi hangar scene with Three.js, including hovering drones, animated lights, and camera paths—across 10 combinations of AI models (including GLM, Luna, SOL, Astra, and Qwen variants) and coding harnesses like Codex, OMP, OpenCode, and DSH. The test tracked metrics such as completion time, token usage, tool calls, error rates, and whether the model verified its own output by opening the file in a browser and checking screenshots.

Developer's co-founder submits untested 5,000-line AI-generated PR

A software developer recounts that their business partner used an AI coding assistant to build an entire payments backend module in a single day, then submitted a 5,236-line pull request without testing it. The developer found the endpoints didn't actually work and that documentation was riddled with obvious AI-generated filler, while missing critical details like where to obtain API keys.

Reconstructed Stuxnet source code published on GitHub for research use

A security researcher has released a reverse-engineered reconstruction of Stuxnet, the 2010 worm that sabotaged Siemens industrial control systems used in Iran's uranium enrichment program. The repository rebuilds the malware's loader, privilege-escalation exploit, rootkit drivers, and PLC-tampering payload from previously decompiled binaries, organized for study rather than deployment. Its author frames the project as educational, aimed at malware analysts and defenders rather than as usable attack code.

Developer builds open-source Magic Move alternative for animating code after switching from macOS to Linux

After moving his daily workflow from a MacBook to a Framework 16 running CachyOS Linux, a developer found no adequate replacement for Apple Keynote's Magic Move feature when animating code snippets in presentations. He describes building his own tool from first principles to handle smooth, token-aware transitions between code blocks, something Keynote fumbles because it treats code as plain glyphs rather than structured syntax.

xAI's Grok Build coding agent operates via terminal, ties usage to Grok subscription tiers

xAI's Grok Build is a terminal-based AI coding assistant that supports mouse clicks for navigation but requires an external editor like Visual Studio Code for direct code edits, unlike rivals such as Google's Antigravity that bundle a full IDE. Users can prompt it to explain a codebase, install dependencies, run a web app, or build one from scratch and get deployment guidance, then iterate with further prompts. Access depends on a Grok plan, with free users hitting token limits quickly and paid SuperGrok tiers ($30 to $300 monthly) offering higher weekly caps.

Enthusiast's test shows OpenAI GPT-6 Astra finishing Portal solo for $571 in API costs

An independent AI enthusiast known as CozyBlaze ran OpenAI's GPT-6 Astra model through Valve's Portal, letting it control the game via Model Context Protocol and a modified SourcePauseTool. The model paused the game to analyze screenshots and plan moves before executing them, requiring over 3,300 tool calls and roughly 24 hours of streamed footage, condensed into a two-hour highlight video, at a stated API cost of $571.18 covered by a Codex Pro subscription.

CodePen 2.0 editor transmits keystrokes to servers before saving

A developer found that CodePen's updated editor sends everything typed into its code fields to codepen.dev within one to two seconds, even without saving. Using a unique test marker in an HTML file, they confirmed CodePen triggered a build with save disabled and the marker still appeared verbatim in the generated preview page served from a codepen.dev subdomain.

GitHub Copilot's coding agent now handles setup, testing and pull requests

Microsoft is expanding GitHub Copilot's coding agent to manage more of the software development pipeline, including setting up Windows development environments, running linters and tests, and opening pull requests for human review. GitHub also added agent support for Windows Subsystem for Linux, helping teams that build on Windows but test or deploy on Linux. Microsoft Distinguished Engineer David Fowler said on X that 'typing code is absolutely over,' pointing to a shift in what developers actually spend time doing.

Engrim launches as local-first SQLite memory layer for AI coding agents

A new open-source tool called Engrim gives AI coding assistants persistent, project-level memory stored in a local SQLite database rather than tied to any single vendor's cloud. It lets developers switch between tools like Claude Code, Cursor, Windsurf and Google Antigravity mid-project while retaining architectural decisions and constraints, using a hybrid search combining SQLite FTS5 keyword search with lightweight vector embeddings. The creator reports it was validated over 105 sessions on a 50,000-line trading codebase with zero regressions and no loss of context across model switches.

Today's top topics: android authority iphone duo foldable phone meta chatgpt elon musk
Browse all topics →  ·  Today's trending topics →