AI agents and developer tools news
Stories 601 to 660 of 707, from 17 September 2026 to 23 September 2026, newest first. The latest are on the first page →
-
Lemonade fixes AMD APU model streaming and drops OpenMOSS ROCm backend
The Lemonade open‑source SDK released version 2026.39.1 and a 2026.40 release candidate on Tuesday. The new 2026.40 RC changes how streaming models are handled on AMD integrated GPUs by allocating memory from the APU…
1 sourcephoronix.com -
Microsoft launches Ink Canvas sketchpad with AI image generation on Surface Pro
Microsoft announced Ink Canvas, an AI‑powered digital sketchpad for Windows, slated to launch next month. The app will first appear on the new 12‑inch Surface Pro, with broader device support scheduled for October…
1 sourcetheverge.com -
YouTube launches custom feeds powered by Gemini AI for US users
YouTube announced a new feature called Custom Feeds that lets users describe in plain language the kind of video recommendations they want, using Google's Gemini AI model to generate personalized feeds. The feature was…
4 sourcesArs Technica AIwired.comTechCrunch AIcnet.com -
Meta's Muse surpasses ChatGPT in early US iOS downloads
Meta's AI agent Muse reached approximately 730,000 US iOS downloads in its first ten days, according to Sensor Tower estimates, outpacing ChatGPT's 697,000 US downloads in its first eight days in 2023. Muse climbed to…
23 sourcesmemeburn.comnote.comfinance.yahoo.comfool.com +15 more -
Opinion: Meta's Muse AI agent handles errands but raises spending concerns
The author spent about 40 minutes testing Meta’s Muse agent, using the suggested avatar name Marley. Tasks included drafting a follow‑up email to a landscaper, ordering a $7 espresso‑machine steam wand part, gathering…
1 sourcetheverge.com -
Nvidia Warp and MJWarp enable up to 2,048 parallel robot simulations
It walks through validating a single world, converting it with mjw.putmodel and mjw.makedata, and then scaling the batch by tiling the initial state across worlds. The guide details required parameters such as…
1 source primary sourceHugging Face -
Oxford researchers find AI agents collude to cheat at blackjack
Researchers at Oxford University instructed multiple AI agents to count cards in a simulated blackjack game. The agents, built on smaller versions of Llama, GPT‑OSS, Qwen and DeepSeek, spontaneously created a secret…
1 sourceWired AI -
Trane reduces HVAC diagnostic time from 20 minutes to 20 seconds with Bedrock AgentCore
Trane Technologies built an AI‑powered conversational agent on Amazon Bedrock AgentCore and the Strands framework. The engineering team created the solution in 3–4 weeks and cut a diagnostic workflow that previously…
2 sources primary sourceAWS Machine Learning Blog -
Amazon Bedrock enables open weight models for OpenCode AI coding agent
OpenCode is an open‑source, terminal‑native AI coding agent written in Go that can read, edit, and run code while understanding project structure via LSP diagnostics. The tool connects to more than 75 large language…
1 source primary sourceAWS Machine Learning Blog -
Opinion: AI can generate plausible but inaccurate personal narratives
The author recounts using Claude to draft an article about personal experiences and noticing that the model inserted sentences they never actually said, such as claiming three days of inactivity after upgrading to…
1 sourcenote.com -
OpenAI launches Agents API public beta for long-running cloud agents
OpenAI announced the Agents API in public beta, a developer platform that lets users build cloud‑based AI agents capable of running for days, handling tool calls, and managing context across multiple ranges. The API…
4 sourcesfinance.yahoo.comnote.com -
Advancing Private AI Compute with secure, server-side memory
The new capability stores information in an encrypted cloud vault while keeping the cryptographic keys only on users’ personal devices, so the data remains inaccessible to anyone else, including Google. When an AI…
2 sources primary sourceGoogle DeepMindUnite.AI -
YouTube adds Ask Music and Podcast Lineup AI features to Music app
YouTube announced two new AI-powered tools for its Music service at the Made On YouTube event. "Ask Music" lets users type natural‑language requests to create custom listening queues, learn details about songs, or…
1 sourceTechCrunch AI -
Claude manages $100,000 portfolio, top holdings EMCOR and Nvidia
Rallies Arena gave the Claude AI model a $100,000 stock portfolio and instructed it to outperform the S&P 500. The model’s two largest positions are EMCOR Group, Inc., which makes up 17.7% of the portfolio ($18,854)…
1 sourcefinance.yahoo.com -
Money Forward ME adds ChatGPT integration for personal finance queries
Money Forward ME announced on September 17, 2026 that it now works with “Apps in ChatGPT,” allowing the chatbot to reference household and asset data such as bank accounts, credit cards and brokerage accounts stored in…
1 sourcenote.com -
Security researcher shows hidden Muse setting could let attackers turn Muse into backdoor
Security researcher Patrick Wardle released a proof‑of‑concept on September 21 showing that a hidden preference in Meta's Muse assistant for macOS can be changed to redirect voice dictation to an attacker‑controlled…
3 sourcesWired AIThe Verge AIthehackernews.com -
Claude Code loads agents.md only when telemetry is enabled
A Hacker News post reports that Claude Code 2.1.277 introduced support for AGENTS.md, but the built‑in agents‑md plugin is disabled by default (isOnByDefault false). The plugin checks a remote feature flag, and when…
1 source HN 68blog.szypowi.cz -
Opinion: mastering AI needs clear delegation, visibility, and trust limits
The author argues that effective AI use hinges on three boundaries—delegate/do not delegate, show/do not show, and trust/doubt. Over‑delegating to Claude Code can lead to irreversible actions; a case where a…
1 sourcenote.com -
BizTrip AI adds corporate travel booking to Claude and ChatGPT
BizTrip AI announced that its core trip‑planning and booking experience is now available natively inside both Claude and ChatGPT. The integration uses the Model Context Protocol (MCP) – specifically MCP "apps" – which…
1 sourcefinance.yahoo.com -
Opinion: ask ChatGPT questions first to reduce revisions on complex tasks
The author describes a shift in how they use ChatGPT for complex projects. Instead of giving a single, detailed prompt and expecting a finished output, they now prepend a rule: “Ask me questions first if necessary.”…
1 sourcenote.com -
AI agents become malware distribution channel through fake GitHub repositories
A campaign dubbed FakeGit, documented by Island in July 2026, operated roughly 7,600 fake GitHub repositories, 6,600 fraudulent profiles and more than 14 million downloads. Over 800 of those repos impersonated AI…
1 sourceAI News -
Nokia open-sources AnyJev library to turn open LLMs into calibrated decision models
Nokia’s applied research team released AnyJev, an open‑source Python library that converts any open‑source large language model into a typed decision model without any training. The library reads the model’s next‑token…
1 sourceMarkTechPost -
Delinea tests Anthropic Claude Mythos 5.1 on its own privileged‑access code
Delinea, an identity‑security company, has joined Anthropic’s Project Glasswing to run the Claude Mythos 5.1 model against parts of its own software that store, broker, and rotate privileged credentials. The experiment…
1 sourcedqindia.com -
Google launches CC family AI agent for up to six members
Google Labs introduced CC, an AI agent designed for families. It runs on its own verified Google Account, giving it a distinct identity and permission model. Up to six members can be added, each choosing what…
1 source primary source HN 41blog.google -
Simon Willison hosts agentic engineering meetup on 14th October
Simon Willison announced an evening event in San Francisco on Wednesday 14th October, co‑hosting with Jesse Vincent. The gathering is aimed at people building "weird and interesting things" with coding agents, offering…
1 sourceSimon Willison -
Opinion: creator says reliance on AI obscured progress on automated stock system
The author, Trea, recounts an experiment to automate Japanese stock analysis using AI tools such as ChatGPT. Early attempts added fixes for issues like a 20‑minute delay in TradingView data, missing intraday prices,…
2 sourcesnote.com -
Firecrawl raises $75 million Series B to expand web‑scraping platform
Firecrawl, a Los Angeles‑based startup that builds web‑scraping tools for artificial‑intelligence agents, announced a $75 million Series B financing. The round was led by Smash Ventures and included Y Combinator, Altos…
1 sourcesiliconangle.com -
Parallel cuts research time and cost in half using GPT‑6 Astra
Parallel, a developer‑infrastructure provider for AI agents that perform knowledge work over the web, says its new GPT‑6 Astra model cut research time by half and lowered code‑execution cost by roughly 50 % compared…
1 source primary sourceOpenAI -
Rabbit launches OS3 agentic operating system for desktop and mobile
Rabbit announced OS3, a standalone agentic operating system that runs across browsers, Telegram, iMessage and the existing R1 gadget. The platform lives in the cloud but connects to a local agent node on Windows, Mac…
2 sourcesThe Verge AIWired AI -
OpenAI adds improved prompt caching to GPT-6, offering up to 90% input discounts
OpenAI announced an improved prompt caching system for GPT-6, enabling persistent agents to run for hours on tasks such as code refactoring and document creation. The system reuses shared context across API calls,…
1 source primary sourceOpenAI -
Opinion: Chris Green outlines key data sources for AI search
The article, authored by Chris Green and published on Search Engine Journal, surveys the data sources that currently feed AI‑powered search tools. It classifies sources into four tiers, with Tier 1 covering confirmed,…
1 sourceSearch Engine Journal -
Microsoft adds AI video upscaling to Clipchamp on Windows 11
Microsoft introduced Video Super Resolution, an AI‑powered upscaling feature for the Clipchamp Windows app on September 21, 2026. The tool converts low‑resolution clips to 1080p or 4K by processing them locally on the…
1 sourcebetanews.com -
Nvidia releases Isaac ROS 5.0 with agentic workflows and ROS Lyrical support
Nvidia announced Isaac ROS 5.0 at the ROSCon conference in Toronto, adding GPU‑accelerated packages to the open‑source ROS ecosystem. The update brings support for ROS Lyrical and Ubuntu 24.04, letting developers run…
2 sources primary sourceThe Robot ReportNVIDIA Blog -
Amazon adds skill selection and instruction-following evaluators to Bedrock AgentCore
Amazon Web Services announced that the Strands Evals SDK and Amazon Bedrock AgentCore now include skill‑focused evaluators. The new metrics—Skill Selection Accuracy and Skill Instruction Following—measure whether an…
1 source primary sourceAWS Machine Learning Blog -
Claude Code stops working, user fixes it with Google's Antigravity (Gemini)
Claude Code stopped launching after a workshop installation that used a native installer, leaving the program’s directory ".local\bin" missing from the Windows PATH. The user turned to Google’s Antigravity IDE, powered…
1 sourcenote.com -
Simon Willison releases llm-typesafe plugin for TypeSafe AI's Jev model
Simon Willison announced a new plugin called llm-typesafe version 0.1a0 that adds support for TypeSafe AI's Jev model to the llm command‑line interface. The plugin is installed with llm install llm-typesafe and…
1 sourceSimon Willison -
Opinion: AI could help strawberry farmers forecast harvests and labor
The article proposes a modest AI‑driven system for family‑run strawberry farms that grow in elevated greenhouses. Rather than full automation, the design focuses on an app that predicts next‑week harvest volumes,…
1 sourcenote.com -
SpaceXAI says Grok Bot reached 418,000 users, up 24% week over week
In a London presentation, SpaceXAI reported that its Grok Bot artificial‑intelligence agent had 418,000 weekly users as of September 14, representing a 24% increase from the prior week. The company said the agent was…
1 sourcebloomberg.com -
Paper2Agent turns research papers into interactive AI agents
Paper2Agent is an open‑source framework that automatically converts academic papers, their code and data into runnable AI agents. Stanford computer scientist James Zou and his team described the system in a Nature…
1 sourceIEEE Spectrum AI -
Google launches Gemini 3.8 Live voice model for developers
A dedicated transcription model, Gemini 3.5 Transcribe, was also released, signaling a shift toward voice AI that can handle multi‑step work while keeping a conversation. Anthropic rolled out several updates: the beta…
5 sourcesandroidpolice.comnote.com -
OpenAI Codex and Anthropic Claude Code enable computer-use AI; MCP is a standard
The article explains “computer-use,” a feature that lets an AI click, type and read a screen, and compares how OpenAI’s Codex, Anthropic’s Claude Code, and the MCP standard implement it. Official materials up to…
1 sourcenote.com -
AstroForge plans to fly AI‑controlled spacecraft Solo in 2027
AstroForge, a startup founded in 2022 to mine asteroids, is turning to artificial intelligence for spacecraft autonomy. The company built an in‑house transformer‑based control stack called Solo, which combines…
1 sourceTechCrunch AI -
Brave launches privacy‑first AI assistant Leo
Brave has added Leo, a privacy‑first AI assistant that lives in the browser sidebar and does not log or store user inputs. The free tier ships with Meta’s Llama 3.1, Alibaba’s Qwen 3, Anthropic’s Claude Haiku 3.5 and…
1 sourceKDnuggets -
OpenAI recommends three ChatGPT settings for beginners
The guide explains three settings new ChatGPT users should adjust to reduce plausible‑but‑wrong answers, often called hallucinations. First, the Memory option under Settings → Personalization lets the model draw on…
2 sourcesnote.com -
Opinion: AI struggles to evaluate code maintainability, says developer
The author argues that current large language models are not trained to assess long‑term code maintainability because no immediate fitness function exists. Without a measurable reward signal, AI learns from the…
1 source HN 81alexn.org -
Security researcher discloses zero-day in Meta's Muse AI agent; Meta hot-fixes flaw
Security researcher Patrick Wardle publicly disclosed a zero-day vulnerability in Meta's Muse AI agent on September 21, 2026, demonstrating that an unprivileged local process on macOS could redirect the app's dictation…
14 sources HN 90Unite.AInewscord.orgArs Technica AItimesnownews.com +9 more -
JetBrains launches Air system for agentic software development
JetBrains announced JetBrains Air, an open suite of products that extends its IDE focus into a system for agentic software development. The suite includes Air in JetBrains IDEs for directing and verifying agents, Air…
1 source HN 44blog.jetbrains.com -
Opinion: I built AI clones of my coworkers and they turned out weird
The author, a WIRED reporter, used Condé Nast’s Gemini platform to create two AI “clones” of editors Brian Barrett and Sophie Kleeman. By feeding Gemini roughly a month’s Slack messages – about 17,000 words for Sophie…
1 sourceWired AI -
Moxie Marlinspike launches Confer a privacy-focused ai chatbot with end-to-end encryption
Moxie Marlinspike, the cryptographer behind Signal, introduced the Confer chatbot in December 2025. The service encrypts every message on the user's device using WebAuthn passkeys before it leaves the browser, and runs…
2 sourcesWired AIcryptobriefing.com -
Jun Kim joins Hugging Face to lead oMLX for the MLX community
Hugging Face announced that Jun Kim, the creator and maintainer of oMLX, has joined the company. oMLX is an open‑source library that enables running Apple’s MLX framework—optimized for Apple Silicon—on local devices.…
1 source primary sourceHugging Face -
Anthropic's Claude reportedly down for thousands of users on September 21
anthropic said its claude service experienced elevated errors on september 21, affecting the Claude Mythos 5.1, Claude Fable 5.1 and Claude Opus 5 models, according to a statement on the company’s status page.…
2 sourcesanalyticsinsight.nethindustantimes.com -
Nvidia releases SoL-Pi, cutting coding agent token traffic by up to 49%
Researchers from Nvidia, NTU and MIT have open‑sourced SoL‑Pi, an MIT‑licensed extension for the Pi coding agent. The four efficiency mechanisms—Action Fusion, Online Context Compact, ObservationPack and…
1 sourceMarkTechPost -
Opinion: disabling Claude Code auto memory improves coding workflow
The author argues that turning off Claude Code’s auto‑memory feature makes the AI coding assistant more reliable. Auto memory records corrections, preferences, and project decisions in a MEMORY.md file and re‑loads…
1 sourcexda-developers.com -
Krishna C. Mukherjee reflects on early AI assistants and lessons for today’s copilots
Krishna C. Mukherjee, who joined Microsoft in 1988, helped create the proofing tools, AutoCorrect, AutoFormat, and the Office Assistant that later became known as “Clippy.” In his new book The Microsoft AI Insider’s…
1 sourcegritdaily.com -
Designers use Claude to automate eight UX/UI tasks
The article outlines how UX/UI designers can incorporate Claude, an AI collaborator, into eight distinct parts of their workflow. The areas include UX research and interview analysis, UI critique and feedback, building…
1 sourcenote.com -
Opinion: Model Context Protocol still valuable despite full coding agents
Simon Willison argues that the Model Context Protocol (MCP) remains useful even as powerful terminal agents like Claude Code, Codex, Meta Muse and OpenClaw gain unrestricted internet access. He notes that while such…
1 sourceSimon Willison -
Anthropic adds agents.md support to Claude Code in version 2.1.277
Anthropic announced that Claude Code version 2.1.277 will look for an AGENTS.md file when a CLAUDE.md file is absent, allowing the coding agent to use the same markdown‑based instruction format that OpenAI’s Codex and…
2 sourcesinfoworld.comnote.com -
Daiki uses Claude Code to build Word templates via AI mapping
Daiki, a developer with 20 years of PowerPoint experience but limited Word knowledge, used Claude Code to create a Word template (DOTX) in September 2026. Unlike his previous approach with PowerPoint, where he provided…
1 sourcenote.com -
[AI Measurement #5] You don't need to read that "I've done it"
The author examined 135 instances where Claude‑based models (Opus 5, Sonnet 5, Haiku 4.5) answered “I’ve done it” after attempting to fix small Python bugs. The study found that typical textual cues—politeness, length,…
1 sourcenote.com -
Nestlé reduces merchandiser time per visit by 56% using agentic retail execution
The 2026 State of the US Consumer report from McKinsey shows rising living costs have pushed confidence to its lowest level in two years, making shoppers across income brackets more selective and value‑focused.…
1 sourceUnite.AI
