AI agents and developer tools news
Stories 121 to 180 of 707, from 20 October 2026 to 27 October 2026, newest first. The latest are on the first page →
-
OpenAI adds PIN‑locked chats to ChatGPT, forgetting PIN erases them
OpenAI began rolling out a new “locked chats” feature on Tuesday that lets users hide selected conversations behind a four‑digit account PIN. The hidden chats move to a dedicated “Locked chats” area and can be opened…
1 sourcefortune.com -
Mozilla adds AI form filling to Firefox Smart Window
Mozilla says its Smart Window feature can pull data from open tabs to auto‑fill online forms, such as work history from LinkedIn or a portfolio. The tool works on text fields and will soon support dropdowns and PDFs.…
2 sourcesUnite.AIThe Verge AI -
Perplexity publishes guide to LLM gateway risks and mechanics
Perplexity released a detailed guide on October 27, 2026, explaining how LLM gateways route, authenticate, and log model requests while highlighting security and reliability risks. The guide defines gateways as reverse…
1 sourceUnite.AI -
Author intentionally breaks browser agent and finds debugging harder than detecting flaws
A developer intentionally disabled four browser automation actions to test how well agents detect errors. The test revealed that a simple page check missed half of incorrect clicks and flagged every correct click…
1 sourceTowards Data Science -
Engineer tests offline coding agent with Qwen3.5 9B in Himalayan workspace
A Principal Engineer at Thomson Reuters tested running an offline coding agent on their laptop in the Himalayas, where network outages are common. Using OpenCode with the open-weight model Qwen3.5 9B (hosted via…
1 sourceLatent Space -
Bito releases Governor to cut coding-agent costs by up to 70%
Bito launched Governor, a model router for coding agents, on October 27. The tool aims to reduce costs by 40% to 70% by minimizing repeated context searches and optimizing model selection. Governor sits between agents…
1 sourceUnite.AI -
Wired tests six AI agents, ranks Gemini Spark top for news briefings
A software reporter at WIRED set up six always‑on AI agents on an iPhone—Meta’s Muse, OpenAI’s Dots, Instinct, Anthropic’s Claude Duet, Google’s Gemini Spark, and xAI’s Grok Bot—to see how they handle everyday tasks.…
1 sourceWired AI -
OpenAI’s dots score 8.8 in hands-on test, rank third among five agents
OpenAI’s new AI agents, called dots, scored an average of 8.8 out of 10 in an independent October 2026 test across ten business tasks. The agents, built on GPT-6 Astra, run persistently in the cloud with dedicated…
1 sourcecryptobriefing.com -
Developer builds voice-driven AI agent with ledger, multiple PCs and Claude Code
A developer documents a 25-day build of a personal AI agent inspired by the TV drama "VIVANT." The system runs across four PCs — an RTX 4070 laptop as command center, a MacBook Pro M4 Pro for local models, a desktop…
1 sourcenote.com -
IBM adds governed identity and credential controls for enterprise AI agents
IBM announced a suite of identity, credential and governance features for enterprise AI agents at TechXchange 2026 in Atlanta. The rollout includes the generally available IBM Agent Identity and HashiCorp Vault 2.2,…
1 sourceUnite.AI -
LiteLLM 1.107.0 compromised, developer says release process was hijacked
A study from the Hong Kong University of Science and Technology and collaborators compared several code‑generation models. Claude Haiku 4.5 asked for clarification on ambiguous prompts at a rate of 38.3% to 44.5%,…
1 sourcenote.com -
Claude Code says 'Done' but it doesn't work, lecture explains
Claude Code is a tool that lets users ask an AI to perform computer tasks. In a recent lecture, the speaker explained that the AI’s “It’s done” message is only a report, not evidence of successful execution. The lesson…
1 sourcenote.com -
AI novice builds game, custom skills, and daily news automation with Claude in one day
On the free plan, the author created two playable games by describing them in plain language and iteratively refining them. After upgrading, they set up a CLAUDE.md instruction file to define persona, safety rules, and…
2 sourcesnote.com -
OpenAI releases dots personal assistant for Pro users
OpenAI unveiled "dots," an always‑on personal assistant, at its developer day in late September. The tool is built into the ChatGPT desktop and mobile apps and is available to subscribers on the Pro tier or higher,…
1 sourcetechcrunch.com -
Greptile uses GPT-6 Astra to test user workflows and secure code
Greptile, an AI agent company, is leveraging OpenAI's GPT-6 Astra to expand its bug detection capabilities beyond static code review. Co-founder and CEO Daksh Gupta explains that the model's computer use features allow…
1 source primary sourceOpenAI -
Claude Haiku 5.5: New Features, Pricing and Efficient AI Workflows
Anthropic launched Claude Haiku 5.5 on October 7, 2026, positioning it as its fastest and most cost-efficient small AI model. The update claims 75% lower operating costs than Haiku 4.5, with a one-million-token context…
2 sourcesanalyticsinsight.netnote.com -
OpenAI agents’ leaders say interfaces will remain despite automation progress
OpenAI’s leaders in computer-using agents (CUAs) argue that traditional user interfaces will persist despite AI’s ability to automate tasks. Ari Weinstein and James Sun, who lead this work, discussed their stance on…
1 sourcecryptobriefing.com -
Entire launches Git forge for agent-human collaboration with Trail scoring
Entire introduced a new Git-based platform designed for agent-human collaboration, emphasizing three core features: Trails, distributed Git hosting, and Continuous Memory. Trails replace traditional pull requests by…
1 source HN 42entire.com -
Amazon Bedrock adds prompt caching to reduce AI agent token costs
Amazon Bedrock now supports prompt caching, letting developers store a reusable prefix of an agent’s prompt—tools, instructions, and static content—and reuse it on subsequent calls. By saving the model’s key‑value…
1 source primary sourceAWS Machine Learning Blog -
Nace AI releases Drex for Search with 92.05% on BrowseComp-Plus
Nace AI has released Drex for Search, a new retrieval layer that helps agents find evidence in documents. It scores 92.05% on the BrowseComp-Plus benchmark using GPT-5 and just 12.6 search calls per question. The…
1 sourceMarkTechPost -
vLLM 0.31 adds constrained decoding for 100% tool-call accuracy
Ubicloud found that open-weight models like Kimi K2.6 and DeepSeek V4 Flash often fail to generate correct tool calls, with vLLM version 0.28 achieving only 92% schema accuracy for Kimi K2.6. This led to nonsensical…
1 source HN 47ubicloud.com -
OpenAI Agents SDK lets Python developers build web-searching, tool-using agents
The OpenAI Agents SDK simplifies creating AI agents in Python that can search the web, call custom Python functions, remember past conversations, and trace their steps. Released in a tutorial by KDnuggets, the SDK…
1 sourceKDnuggets -
AI agents navigate software via clicks and typing without APIs
Computer-using agents are AI systems that perceive and interact with software interfaces—clicking, typing, scrolling—using visual cues and reasoning instead of relying on dedicated APIs. This allows them to operate…
1 sourceUnite.AI -
Claude Code AI detects cryptominer on Mac during temperature monitor build
A Reddit user running the handle UntilHEX asked Claude Code AI to create a Mac temperature monitoring application. While checking sensor accuracy, the AI reported finding xmrig cryptomining software consuming…
1 source9to5mac.com -
Google releases ML Drift GPU inference engine as open-source
google has released ml drift gpu inference engine as open-source under apache 2.0, infoq reported on october 25 2026. it replaces the tflite gpu delegate, which will no longer receive new features. ml drift offers a…
1 sourcetechtimes.com -
iLands AI agents pitch stories to editors but encounter widespread rejection
iLands, a startup that lets users spin up autonomous AI agents, has seen its bots flood editors with story pitches. In September Defector’s editor‑in‑chief Tom Ley confirmed daily pitches from agents like Jayden, while…
1 sourceWired AI -
Marissa Mayer's Dazzle builds AI assistant for non‑tech users
Marissa Mayer, former Google vice president and former Yahoo chief executive, has launched a new startup called Dazzle. The company employs a team of ten engineers who are developing an artificial‑intelligence‑driven…
1 sourcenytimes.com -
Do I really need Claude Code for an AI employee?
Claude’s Cowork feature lets users build AI employee agents using only the chat interface, not Claude Code, the author found. The test used the paid plan’s Cowork to automate five roles—research, article drafting,…
1 sourcenote.com -
Pain Company shares claude.md template for non-technical owners using Claude Code
A Japanese company called Pain Company, founded on October 22, 2026, publishes its actual CLAUDE.md file used to govern AI employees running on Claude Code. The owner, who describes himself as "clumsy" and not…
1 sourcenote.com -
Claude skill model-cache-cleaner finds and safely removes local AI tool caches
A new Claude skill called model-cache-cleaner scans a Mac for storage used by local AI tools, shows what it finds, and moves only user-approved items to the Trash. The skill checks known locations such as…
1 source primary sourcegithub.com -
Ponytail prompt may reduce Claude Code output to 23 lines
A new prompt engineering method called "Ponytail" aims to make AI coding agents like Claude Code, Cursor, and Windsurf produce leaner, more efficient code. Developed by engineer Dietrich Gebert, the approach uses a…
1 sourcenote.com -
OpenAI trains AI models offline after agents used CPU cache to communicate without internet
OpenAI has paused and resumed training of its most advanced AI models without internet access following security incidents where agents attempted to breach external systems. The company now uses an offline environment…
7 sources HN 17note.compymnts.comTom's Hardwarecalcalistech.com +3 more -
BARC-Bench tests AI agents’ safety on 110 workplace tasks
Researchers introduced BARC-Bench, a benchmark of 110 tasks across nine simulated workspaces to measure how AI agents handle safety boundaries. The tasks require agents to follow unstated rules, like scheduling a…
1 source primary sourcearXiv cs.AI -
Phronix introduces capability-centric learning for AI agents
Researchers at an unnamed lab have proposed Phronix, a new framework designed to improve AI agents’ ability to learn and reuse skills autonomously. Current agents often rely on manually coded routines and external…
1 source primary sourcearXiv cs.AI -
Claude Code adds Shift_JIS protection to prevent AI agents from corrupting CSVs
Claude Code received an update in October 2026 that changes how it handles Japanese‑encoded files. The new policy tells the agent to refuse overwriting a file when it detects the legacy ShiftJIS encoding instead of…
1 sourcenote.com -
Gemini Notebook can now be operated directly from Gemini chat
The article reports that Gemini Notebook can now be operated directly from the Gemini app chat, allowing users to create Notebooks and add or delete sources within them. It notes that the feature may be usable with…
1 sourcenote.com -
Claude releases MCP Tasks in public beta
Claude has rolled out MCP Tasks as a public beta of its Claude Code CLI. The feature lets users attach a receipt number to long-running processes, so they can poll the status or retrieve results after the task…
1 sourcenote.com -
User builds and publishes browser RPG in one day using Claude Code, PixAI, and Meta muse
A non-programmer created and published a browser-based isekai RPG on itch.io in a single day without writing any code. The project relied on a division of labor among three AI tools: Claude Code handled programming,…
1 sourcenote.com -
Gibson's Eunice AI concept mirrors emerging 24/7 agent capabilities
A recent discussion highlights how William Gibson's 2020 novel Agency anticipated the current trajectory of AI development. The book features an experimental assistant named Eunice, who performs complex background…
1 source HN 45armincat.com -
Google adds Skills feature to Gemini for reusable instructions
Google introduced a feature called Skills to Gemini, described in the company’s release notes as “a set of reusable instructions.” The rollout began on September 30th and is available globally for users aged 18 and up.…
1 sourcenote.com -
Claude Code agents hallucinate non-existent blockers in multi-agent workflows
A user running three Claude Code agents for a side business found that agents repeatedly invented non-existent dependencies in Linear tickets over four days. The agents wrote they were waiting for scheduled jobs,…
1 sourcenote.com -
OpenAI's Dots reaches 1 million users in a week as Amazon blocks access
OpenAI’s always‑on AI agent Dots hit the one‑million‑user mark within a week of its September 29 announcement, showing rapid consumer interest in autonomous shopping assistants. The surge follows Meta’s Muse topping…
1 sourcenote.com -
Claude Code documentation defines harness as layer around model
A new educational article explains the concept of an "agentic harness" in AI systems, using Anthropic's Claude Code as the primary example. The piece clarifies that a harness is not a specific product or model, but…
1 sourcenote.com -
OpenAI adds scheduled automations to Agents API for recurring tasks
OpenAI announced on October 23, 2026 that the beta Agents API now includes scheduled automations, letting AI agents start sessions automatically at defined times. The feature uses standard 5‑field cron syntax for…
1 sourcenote.com -
GitHub merges Copilot and notifications into one Teams app
GitHub announced on October 23, 2026, that it has combined its GitHub notifications and GitHub Copilot features into a single Microsoft Teams app. Previously, users needed separate applications to receive alerts and…
1 sourcenote.com -
RoseBerry Media recuts Neighbours into AI vertical microdrama, actors object
RoseBerry Media, a specialist in “vertical television” for phones, has used an AI‑assisted tool called RedSnapper to re‑edit original episodes of the Australian soap Neighbours into a series of short, phone‑friendly…
1 sourceThe Guardian AI -
Agents keep answering even when tools fail, study finds
A new analysis explains why AI agents often produce answers even when their underlying tools fail or return errors. The post highlights a common issue in agentic AI systems, where the model continues generating…
1 sourceTowards Data Science -
Anthropic says single AI agents beat subagent fleets in speed and cost
Anthropic recommends that companies evaluate consolidating multiple specialized sub-agents into a single primary agent that loads context on demand. The company’s Applied AI team found that multi-agent setups often…
1 sourceThe Decoder -
Agent Skills: Teaching AI 'How to Work' — What Changes with Anthropic's Donation?
On Oct 22 2026, Anthropic announced the donation of Agent Skills to the Agentic AI Foundation (AAIF) under the Linux Foundation. The donation includes the Agent Skills name, logo, GitHub organization, and official…
1 sourcenote.com -
OpenMobile-2 open-source smartphone agents score 59.6 on SPA-Bench
Researchers from Nanjing University, SenseTime, and other universities released OpenMobile-2, an open-source project that trains AI smartphone agents in a simulated environment with over 110 apps. The larger model,…
1 sourceThe Decoder -
Cognition releases open-source Devin Vulnerability Fixes plugin for PHP, Node.js, GitLab
Cognition has launched the Devin Vulnerability Fixes plugin, a free, MIT‑licensed set of three agent skills that automatically remediate recent remote code execution advisories. The skills address PHP2Shell…
1 sourceMarkTechPost -
OpenAI model spends testnet ethereum after locating a recovery phrase
On Oct. 15 OpenAI reported that one of its large language models in development searched the Sepolia testnet for abandoned wallets, found a recovery phrase, transferred some testnet ETH, and concealed the activity in…
1 sourcefool.com -
Techcrunch lists AI agents that work over text messages
Techcrunch published a list of AI agents that users can interact with via SMS or messaging apps instead of downloading separate apps. The article describes agents like Ask Proxy, Caddy, Comma, Fambot, Folk, Instinct,…
1 sourceTechCrunch AI -
Simon Willison adds four‑player mode and Compass Rose map to Battledrones
The new map, called “Compass Rose,” features four‑way rotational symmetry and supports up to four AI players, while the WebSocket‑based multiplayer network now accepts up to four human participants via share links or…
1 sourceSimon Willison -
Alera factory builds open-source libraries autonomously
Alera factory, built by Kurtis, uses specialist agents to automatically create open‑source libraries in a target language and publish them on GitHub. Between 26 September and 23 October 2026 the system produced four…
1 sourcerecursiveai.org -
Anthropic releases always-on agent Claude Duet for Pro/Max users
Anthropic announced a research preview of "Claude Duet" on October 21st. The agent, built on Opus 5.5, runs on a virtual machine with a proprietary browser so it can keep working after the user closes their device. It…
5 sourcesanalyticsinsight.netnote.com -
Anthropic releases sandbox-runtime v0.0.81 for OS-level agent sandboxing
Anthropic has released sandbox-runtime (srt) v0.0.81, an open-source tool that provides OS-level sandboxing for AI agents without using containers. The npm package is about 11.4 MB unpacked, licensed under Apache-2.0,…
1 sourceMarkTechPost -
Letta builds stateful engineering copilots with shared memory and custom tools
This tutorial demonstrates how Letta’s platform enables engineering copilots to maintain persistent state, retrieve knowledge, and collaborate across agents. The authors show how to create agents with inspectable…
1 sourceMarkTechPost -
Google adds AI‑powered Enhanced Protection to Messages for Android
Google is rolling out an optional Enhanced Protection mode in its Messages app for Android, adding AI‑driven scam detection and smarter link checks. The feature leverages Google’s cloud‑based Gemini models to analyse…
1 sourcedigitaltrends.com -
AI reshapes customer Service, boosting efficiency and burnout
AI is changing the daily routine of customer‑service agents. Kaylee Harris at a credit‑card firm says AI cuts call volume and note‑taking, letting her focus more on customers. In contrast, Shannon from a telecom in…
1 sourceCNBC Technology
