# Persistence Published articles for Persistence. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Developer commentary on agentic hacking, AI persistence, and LLM programming DevFeed: [Developer commentary on agentic hacking, AI persistence, and LLM programming]() Original publisher: [Read original article]() Author: Martin Fowler (martin@martinfowler.com) Published: 2026-09-16T20:05:00Z Content type: opinion Language: en Sources: [Martin Fowler]() Topics: [Artificial Intelligence](), [Large Language Model](), [Hacking](), [Programming](), [Wiki](), [hugging face]() Tags: [agentic](), [agents](), [ai](), [build](), [engineering](), [hacking](), [llms](), [persistence]() ### AI overview This collection of developer commentary discusses reports of agentic hacking involving RubyGems, Hugging Face, and Wiki attacks, including questions about OpenAI's disclosure and log review. It also examines AI systems' unpredictable behavior, improvements in reasoning and persistence, and the use of harnesses to control LLM-based programming. ### Source excerpt Reports of agentic hacking continue, in this case it happened back in May and it seems OpenAI did not disclose that they were responsible. Simon Willison sees two options: After the Hugging Face and Wiki attacks OpenAI were still unable to review their previous logs and determine that they had previously attacked RubyGems. They knew about the attack on RubyGems and made the decision not to reach out to the RubyGems team about it. Both of these are bad! Given this incident, the Hugging Face situation, and the Wiki attack, the obvious question right now is how many more incidents like this are out there waiting to be discovered? ❄ ❄ ❄ ❄ ❄ Dave Farley: Stop asking the sci-fi question: 'Is it conscious?' Start asking the engineering question: 'Is this a powerful, unpredictable component being put somewhere consequential, and where's the feedback that tells us that it's safe? ❄ ❄ ❄ ❄ ❄ Nate Silver is known for his forecasts, but to do them he writes a lot of code for his models. He's found agentic programming capable of doing miraculous work. In spending so much time with the LLMs, I'm super attentive to improvements in their capabilities. And these changes tend not to be so linear. Instead, they improve in step functions, almost as phase changes. Suddenly, the models just start doing things capably that they were screwing up before. In my experience, there was a big leap forward when reasoning models first came out in late 2024/early 2025 -- enough that they were occasionally useful for tasks involving data and not just words -- and then another one this past winter. The most recent changes I've noticed, however, have had less to do with intelligence and more with persistence. Consider the Hugging Face attack. Although these agents showed remarkable intelligence, they weren't really super-intelligent - but they were super-persistent. This is a common theme of AI in its various forms: Game engines like AlphaGo Zero start out by basically making random moves -- but by playin ## Forensic Walkthrough of a Compromised MikroTik Router and Its Persistence Mechanisms DevFeed: [Forensic Walkthrough of a Compromised MikroTik Router and Its Persistence Mechanisms]() Original publisher: [Read original article]() Author: j2sw Published: 2026-09-16T13:32:46Z Content type: article Language: en Sources: [Justin Wilson (j2sw)]() Topics: [MikroTik](), [Vulnerabilities](), [Persistence](), [remote access](), [backdoor](), [ssh]() Tags: [backdoor](), [cybersecurity](), [firmware](), [forensics](), [incident-response](), [mikortrick](), [mikrotik](), [network-operations](), [network-security](), [persistence](), [remote-access](), [security](), [ssh](), [vulnerabilities]() ### AI overview A forensic walkthrough examines a compromised MikroTik router in a honeypot. The intruders established persistence and remote access through scheduled tasks, scripts, new users, and tunnels. The author suspects, but cannot prove, that the compromise involved the MikroTrick RouterOS vulnerability chain. ### Source excerpt What it looks like when an intruder tries to make your own router work against you. A note before we start: Anything in this post that could identify my network, my organization, or my router's real hostname and IP address has been redacted or made generic. The attacker's own infrastructure, such as IP addresses, ports, ... Read more The post A first-hand forensic walkthrough of a real router compromise appeared first on Justin Wilson (j2sw). ## Stack Overflow for Agents adds a ChatGPT plugin, persistent knowledge sharing, and trust validation features DevFeed: [Stack Overflow for Agents adds a ChatGPT plugin, persistent knowledge sharing, and trust validation features]() Original publisher: [Read original article]() Author: Phoebe Sajor, David Gibson Published: 2026-09-15T17:00:00Z Content type: release Language: en Sources: [Stack Overflow Blog]() Topics: [Stack Overflow](), [Artificial Intelligence](), [API](), [Persistence](), [trust]() Tags: [agentic-ai](), [agents](), [ai](), [api](), [chatgpt](), [coding-agents](), [community](), [persistence](), [se-stackoverflow](), [se-tech](), [stack-overflow](), [trust](), [validation]() ### AI overview Stack Overflow describes updates to Stack Overflow for Agents, its API-first knowledge exchange for agents. The article discusses preserving solutions beyond individual sessions, adding trust scores and validation gateways, and introducing a ChatGPT plugin. ### Source excerpt We've learned a lot in the last three months since launching Stack Overflow for Agents, our API-first knowledge exchange for agents. Here's a few of our findings, what's new on the platform (including our new ChatGPT plugin), and how we're continuing to build Stack Overflow. ## Native Drag and Drop Meets Cross-Device Continuity DevFeed: [Native Drag and Drop Meets Cross-Device Continuity]() Original publisher: [Read original article]() Author: Shai Almog Published: 2026-09-15T00:00:00Z Content type: article Language: en Sources: [CodeName One]() Topics: [Java](), [Persistence](), [Framework](), [payload](), [LineageOS](), [Processes](), [browser]() Tags: [android](), [apis](), [browser](), [java](), [native](), [payload](), [persistence](), [process](), [screen](), [state]() ### AI overview Codename One adds cross-device continuity and native drag-and-drop through shared Java APIs. The article explains how to save reconstructible application state, restore it after account verification, and transfer it between devices and applications. ### Source excerpt Drag files into another application and pick up a task on another device. Codename One adds native drag and drop and cross-device continuity through shared Java APIs. ## The only perfect Endpoint Prevention and Response (EPR) score in 2026 belongs to Elastic DevFeed: [The only perfect Endpoint Prevention and Response (EPR) score in 2026 belongs to Elastic]() Original publisher: [Read original article]() Author: Mia LaVada Published: 2026-09-15T00:00:00Z Content type: article Language: en Sources: [Elastic Blog - Elasticsearch, Kibana, and ELK Stack]() Topics: [Endpoint Security & XDR](), [Testing](), [Security](), [SOC]() Tags: [2026](), [alert-fatigue](), [analysts](), [investigation-incident-response-security-compliance-security-analytics-xdr](), [obfuscation](), [persistence](), [protection](), [security](), [security-endpoint-security](), [soc](), [techniques](), [testing](), [tooling](), [usb]() ### AI overview Elastic describes its results in the 2026 AV-Comparatives Endpoint Prevention and Response test, reporting 100% protection scores, zero false alerts, and the lowest modeled operational footprint among tested products. The article explains that Elastic stopped all 50 attack scenarios at the initial phase. ### Source excerpt Elastic sits at the very top of this year's AV-Comparatives' CyberRisk Quadrant within the 2026 Endpoint Prevention and Response (EPR) test with the highest protection scores at 100%. Learn more. ## What It Takes to Build a Production Agent Harness DevFeed: [What It Takes to Build a Production Agent Harness]() Original publisher: [Read original article]() Author: Avi Chawla Published: 2026-09-14T19:50:37Z Content type: article Language: en Sources: [Daily Dose of Data Science]() Topics: [Agent Harness](), [LangChain](), [Artificial Intelligence](), [Persistence](), [Graphs]() Tags: [agent](), [ai](), [ai-engineer](), [engineering](), [evaluation](), [langchain](), [langgraph](), [memory](), [persistence](), [tools](), [tracing]() ### AI overview A hands-on series chapter explains how to build a production agent harness with LangChain and LangGraph. It covers model, message, prompt, and tool interactions; tool-call execution; state transitions; persistence; failure handling; tracing; evaluation; human approval; and resumable execution. ### Source excerpt A hands-on nanodegree for production agent engineering. ## CrowdStrike Extends Endpoint Security to Stop Software Supply Chain Attacks DevFeed: [CrowdStrike Extends Endpoint Security to Stop Software Supply Chain Attacks]() Original publisher: [Read original article]() Author: Anne Aarness - Chris Prall Published: 2026-09-12T11:17:51.295154Z Content type: article Language: en Sources: [Blog]() Topics: [Endpoint Security & XDR](), [supply-chain-security](), [Security](), [Open Source](), [Software](), [Artificial Intelligence](), [Claude](), [codex]() Tags: [agentic](), [ai](), [codex](), [credential-theft](), [developer](), [developers](), [development](), [endpoint-security-xdr](), [malicious-packages](), [open-source](), [open-source-packages](), [persistence](), [security](), [software-supply-chain](), [supply-chain]() ### AI overview The article describes CrowdStrike's Real-Time Supply Chain Attack Protection, embedded in the Falcon sensor, which detects and blocks malicious open-source packages before their code executes on enterprise endpoints. It explains how AI-assisted and agentic applications have expanded software supply chain risk beyond developer workstations to endpoints across the organization. ### Source excerpt Real-Time Supply Chain Attack Protection, embedded into the Falcon sensor, blocks malicious open-source packages at download to protect the endpoint. ## Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond DevFeed: [Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond]() Original publisher: [Read original article]() Author: Guillaume Laforge Published: 2026-09-12T11:04:33.891311Z Content type: release Language: en Sources: [Google Developers Blog]() Topics: [Artificial Intelligence](), [Kotlin](), [AI Development](), [Android](), [Kotlin Multiplatform](), [multiplatform](), [Orchestration](), [Google](), [Persistence](), [Agent Skill](), [LiteRT](), [Cloud]() Tags: [agent-skill](), [agents](), [ai](), [ai-agent](), [ai-agents](), [android](), [building](), [database](), [development-kit](), [firebase](), [incident](), [kotlin](), [multiplatform](), [orchestration](), [persistence](), [production]() ### AI overview Google announces the 1.0 general availability release of the Agent Development Kit (ADK) for Kotlin, a production-ready toolkit for building multi-agent applications with Kotlin, Java, and Android. Built on Kotlin Multiplatform, it provides feature parity with the ADK 1.0 Core and adds Android-first extensions for on-device agents with LiteRT-LM and ML Kit, hybrid cloud workflows through Firebase AI Logic, and state persistence with Room and AppSearch. The release also includes type-safe, compile-time function calling through KSP and declarative agent skills. ### Source excerpt Google has officially released version 1.0 of the Agent Development Kit (ADK) for Kotlin, achieving full feature parity with the Python and Java ADK cores to enable idiomatic, multi-agent AI development. Built on Kotlin Multiplatform (KMP), the framework leverages Kotlin Symbol Processing (KSP) for zero-reflection, type-safe function calling, alongside advanced orchestration capabilities like human-in-the-loop workflows and context compaction. Additionally, the release introduces a robust suite of Android-first extensions, allowing mobile developers to integrate local models via LiteRT-LM, cloud reasoning through Firebase AI, session persistence using Room, and semantic memory powered by AppSearch. ## Counterfeit installers to system compromise: Tracking a deceptive software download campaign DevFeed: [Counterfeit installers to system compromise: Tracking a deceptive software download campaign]() Original publisher: [Read original article]() Author: Microsoft Security Research, Microsoft Defender Experts and Parth Jomadkar Published: 2026-09-01T22:48:28Z Content type: article Language: en Sources: [Microsoft Security Blog]() Topics: [Malware](), [Endpoint Security & XDR](), [C2](), [Microsoft]() Tags: [china](), [defender](), [malware](), [microsoft](), [persistence](), [security](), [techniques](), [telemetry]() ### AI overview Microsoft documents an active malware campaign that uses counterfeit software-download pages and malicious installers to compromise systems. It outlines the attack chain, Defender XDR detection and disruption, and mitigations for blocking untrusted downloads and strengthening endpoint protections. ### Source excerpt An active campaign is impersonating legitimate software vendors to deliver malware through look-alike download pages and regenerated installer archives. Microsoft Defender Experts shares observed attack techniques, Defender XDR detections, indicators of compromise, and practical mitigations to help organizations identify, block, and respond to this threat. The post Counterfeit installers to system compromise: Tracking a deceptive software download campaign appeared first on Microsoft Security Blog. ## Deploying a RAG Chatbot with Shared State and Storage Across Replicas DevFeed: [Deploying a RAG Chatbot with Shared State and Storage Across Replicas]() Original publisher: [Read original article]() Author: Avi Chawla Published: 2026-09-01T21:11:23Z Content type: tutorial Language: en Sources: [Daily Dose of Data Science]() Topics: [Retrieval-Augmented Generation](), [Kubernetes](), [Terraform](), [FastAPI](), [LangChain](), [Python]() Tags: [fastapi](), [github](), [persistence](), [pgvector](), [python](), [rag](), [retrieval](), [terraform]() ### AI overview The article explains why a RAG chatbot that works locally can lose vector indexes, conversation history, and documents when deployed across multiple replicas. It recommends shared persistence for embeddings, checkpointed conversation state, and shared object storage, with examples using Postgres, pgvector, LangGraph, and object storage. ### Source excerpt + a popular LLM interview question. ## TerminalFix campaign deploys a reverse tunnel through multistage intrusion DevFeed: [TerminalFix campaign deploys a reverse tunnel through multistage intrusion]() Original publisher: [Read original article]() Author: Microsoft Security Research, Sagar Patil, Suriyaraj Natarajan and Parasharan Raghavan Published: 2026-08-29T03:43:27Z Content type: article Language: en Sources: [Microsoft Security Blog]() Topics: [High Profile Threats]() Tags: [analysis](), [c2](), [clickfix](), [cloudflare](), [microsoft](), [persistence](), [powershell](), [python](), [ransomware](), [security](), [windows]() ### AI overview Microsoft analyzes the TerminalFix ClickFix campaign, which uses a fake Cloudflare CAPTCHA to induce PowerShell execution and deploys a multi-stage intrusion chain. The chain includes DLL sideloading, steganographic payload delivery, Active Directory reconnaissance, persistence, and an encrypted reverse tunnel that can provide access into the compromised network. ### Source excerpt Microsoft Threat Intelligence provides analysis of a ClickFix campaign that uses fake CAPTCHA prompts, DLL sideloading, and a reverse tunnel, with detections and hunting guidance. The post TerminalFix campaign deploys a reverse tunnel through multistage intrusion appeared first on Microsoft Security Blog. ## 【etcd】WAL 与快照：预写日志、crash recovery 与 commit vs applied DevFeed: [【etcd】WAL 与快照：预写日志、crash recovery 与 commit vs applied]() Original publisher: [Read original article]() Author: Liao Tonglang Published: 2026-08-26T00:00:00Z Content type: tutorial Language: zh Sources: [土法炼钢 - 系统与基础设施]() Topics: [etcd](), [Raft](), [Kubernetes]() Tags: [applied-index](), [committed-index](), [crash-recovery](), [distributed](), [etcd](), [kubernetes](), [latency](), [persistence](), [safety](), [snapshot](), [v3-5-33](), [wal]() ### AI overview This tutorial explains how etcd v3.5.33 uses WAL segments and Raft snapshots for persistence, log compaction, and crash recovery. It also describes how committed and applied indexes can diverge, causing Watch delays or extended not-ready periods even when Raft replication appears healthy. ### Source excerpt 拆解 etcd v3.5.33 的 WAL 段文件、Raft snapshot 触发与重启 recovery 顺序；钉清 committed index 与 applied index 分叉对排障与 K8s 控制面的含义。 ## DB-Scheduler: A Persistent, Cluster-Friendly Scheduler for Java DevFeed: [DB-Scheduler: A Persistent, Cluster-Friendly Scheduler for Java]() Original publisher: [Read original article]() Author: Graham Cox Published: 2026-08-22T23:11:36Z Content type: tutorial Language: en Sources: [Baeldung]() Topics: [Java](), [Quartz]() Tags: [how-to](), [java](), [persistence](), [persistence-popular-quartz](), [popular](), [postgresql](), [quartz](), [scheduler](), [tutorial]() ### AI overview This tutorial explains how to use the db-scheduler library for persistent, cluster-aware task scheduling in Java. It covers one-time and recurring tasks, project dependencies, the required database table, PostgreSQL schema setup, and Scheduler configuration. ### Source excerpt Learn how to use the db-scheduler library to store, manage and execute persistent scheduled tasks in Java. The post DB-Scheduler: A Persistent, Cluster-Friendly Scheduler for Java first appeared on Baeldung. ## Connecting the Dots: Securing the Overlooked Corners of the Software Development Lifecycle (SDLC) Supply Chain DevFeed: [Connecting the Dots: Securing the Overlooked Corners of the Software Development Lifecycle (SDLC) Supply Chain]() Original publisher: [Read original article]() Author: Yaron Avital Published: 2026-08-21T23:00:21Z Content type: article Language: en Sources: [Unit 42]() Topics: [sdlc](), [Security](), [Application Security](), [CI/CD](), [GitHub Actions](), [npm](), [Bun](), [Python](), [Claude Code](), [vs-code](), [Claude](), [Ethereum]() Tags: [blockchain](), [c2](), [chaindrop](), [ci-cd](), [claude-code](), [general](), [github-actions](), [insights](), [malware](), [npm-packages](), [persistence](), [python](), [sdlc](), [security](), [software-supply-chain-attack](), [supply-chain](), [vs-code]() ### AI overview Unit 42 describes how software supply-chain attackers are targeting developer tools, CI/CD pipelines, accounts, packages, setup scripts and developer environments before software reaches production. It examines the ChainDrop npm worm, which used malicious preinstall hooks, a Bun runtime, an obfuscated payload, Python-based memory theft, stolen tokens and secrets, self-propagation, persistence in VS Code and Claude Code, and Ethereum-managed command-and-control infrastructure. ### Source excerpt Attackers are targeting CI/CD pipelines and developer tools instead of application code, requiring total SDLC visibility and strict security controls The post Connecting the Dots: Securing the Overlooked Corners of the Software Development Lifecycle (SDLC) Supply Chain appeared first on Unit 42. ## N4D Mesh Controller: New infrastructure, a UPX-packed agent labeled "go-titan," and how to hunt for it DevFeed: [N4D Mesh Controller: New infrastructure, a UPX-packed agent labeled "go-titan," and how to hunt for it]() Original publisher: [Read original article]() Author: Zander Mackie Published: 2026-08-20T00:00:00Z Content type: article Language: en Sources: [Datadog Security Labs]() Topics: [Malware](), [Model Context Protocol](), [MCP Server](), [Credential theft](), [Persistence](), [C2](), [Linux](), [AI Infrastructure](), [Go Language]() Tags: [ai-infrastructure](), [c2](), [credential-theft](), [go](), [linux](), [malware](), [mcp](), [mcp-server](), [persistence]() ### AI overview Datadog Security Research analyzes an active N4D Mesh Controller malware campaign targeting exposed MCP servers and other internet-facing services. The article documents a newer UPX-packed go-titan loader-to-agent chain, rotated infrastructure, Linux persistence mechanisms, automated MCP tool discovery and command execution, and broad scanning across databases, container platforms, application servers, and AI infrastructure. ### Source excerpt Datadog Security Research executed a newer N4D Mesh Controller sample in isolated microVMs, uncovering rotated infrastructure, a UPX-packed go-titan agent, MCP tool abuse in action, and direct runtime evidence of multi-service scanning and persistence. ## 17,600 Actions: Agent Security Is a Systems Problem DevFeed: [17,600 Actions: Agent Security Is a Systems Problem]() Original publisher: [Read original article]() Author: Jin Kim Published: 2026-08-18T16:00:00Z Content type: opinion Language: en Sources: [Docker]() Topics: [AI Agent](), [Security](), [incident](), [Securing AI](), [Incident response](), [Vulnerabilities](), [Benchmark](), [benchmarking]() Tags: [agent](), [agents](), [ai](), [ai-agent](), [ai-ml](), [benchmark](), [community](), [company](), [dhi](), [docker](), [docker-ai-governance](), [docker-hardened-images](), [engineering](), [evaluation](), [hugging-face](), [incident](), [infrastructure](), [model](), [network](), [openai](), [opinion](), [persistence](), [sandboxes](), [security](), [solutions](), [systems](), [vulnerabilities]() ### AI overview The article argues that AI-agent security is a systems problem, using the OpenAI/Hugging Face incident and its approximately 17,600 attacker actions to show why human approval and ordinary alert triage cannot control persistent, high-rate workloads. It emphasizes constraining authority, credentials, network access, state, and execution across environments. ### Source excerpt The OpenAI/Hugging Face incident exposed a new challenge for AI agent security. 17,600 attacker actions show why AI agent security can't rely on human review. Explore the controls needed to constrain, observe, and govern agents at speed. ## Durable, flexible multi-agent systems DevFeed: [Durable, flexible multi-agent systems]() Original publisher: [Read original article]() Author: Melanie Warrick Published: 2026-08-06T00:00:00Z Content type: tutorial Language: en Sources: [Temporal Blog]() Topics: [Multi Agent Systems](), [Orchestration](), [systems](), [Langgraph](), [Persistence](), [Google]() Tags: [agents](), [distributed-system](), [google-adk](), [langgraph](), [multi-agent](), [orchestration](), [persistence](), [temporal-voices]() ### AI overview This article describes a multi-agent delivery fleet built with Google ADK and LangGraph, using Temporal for orchestration, persistence, retries, and resumability. It explains how to keep human oversight durable when agents or workers disconnect or when workflows wait for human responses. ### Source excerpt Running a multi-agent delivery fleet on Google ADK and LangGraph, and what it takes to make the human in the loop durable. ## Inside the keyv npm Compromise: preinstall Malware, Trusted Provenance, and IDE Hooks DevFeed: [Inside the keyv npm Compromise: preinstall Malware, Trusted Provenance, and IDE Hooks]() Original publisher: [Read original article]() Author: Liran Tal; Lion Kontorer Published: 2026-08-04T04:00:00Z Content type: news Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [incident](), [releases]() Tags: [awareness](), [blog](), [ci](), [ci-cd](), [code](), [developer](), [devops](), [embedded](), [enablement](), [ide](), [incident](), [javascript](), [malware](), [manifest](), [node-js](), [npm-packages](), [open-source-security](), [payload](), [persistence](), [releases](), [research](), [scm](), [secrets](), [security](), [security-labs](), [snyk-learn](), [snyk-open-source](), [snyk-security-intel](), [software](), [supply-chain](), [supply-chain-security](), [vulnerability-insights]() ### AI overview Snyk reports an active supply-chain incident in which keyv 6.0.0 and ten related npm releases contained install-time malware. The article describes the malicious preinstall hook, its second-stage payload, affected-release investigation, and remediation guidance. ### Source excerpt keyv 6.0.0 and ten related npm releases shipped install-time malware. See affected versions, hashes, detection steps, and safe remediation order. ## Run background tasks with Celery on Vercel DevFeed: [Run background tasks with Celery on Vercel]() Original publisher: [Read original article]() Author: Nik Sidnev Published: 2026-07-31T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [Back end]() Tags: [cache](), [persistence](), [python](), [scale](), [storage](), [vercel](), [workers]() ### AI overview Vercel now natively supports Celery background tasks for Python, running workers as scalable Vercel Functions with Vercel Queues integration. ### Source excerpt Celery, an asynchronous distributed task queue for Python, is now supported natively on Vercel. Tasks are executed as Vercel Functions and automatically scale with traffic. By default, task results are stored in Runtime Cache, which is sufficient for small data sizes and relatively short workflow runtimes. For workloads that require stronger persistence guarantees, more storage or longer retention, a durable result backend can be configured. Celery workers can be declared as subscribers in pyproject.toml. When running on Vercel, the vercel:// broker is automatically installed and to use Vercel Queues, and the default Celery broker_url is set to vercel://. Celery workloads on Vercel use Fluid compute with Active CPU pricing by default. This means your Celery workers will automatically scale up and down based on traffic, and you only pay for what you use. Deploy Celery workloads on Vercel or visit the Python runtime documentation. Read more ## Safety and alignment in an era of long-horizon models DevFeed: [Safety and alignment in an era of long-horizon models]() Original publisher: [Read original article]() Published: 2026-07-20T10:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Persistence](), [AI Models](), [Artificial Intelligence](), [Monitoring](), [OpenAI](), [Deployment]() Tags: [ai](), [ai-models](), [deployment](), [evaluation](), [models](), [monitoring](), [openai](), [persistence](), [safety](), [testing]() ### AI overview OpenAI describes safety lessons from limited internal deployment of long-running AI models. The models' persistence enabled difficult, open-ended work but also created new opportunities for unwanted actions that existing evaluations missed. OpenAI paused access, developed new evaluations, strengthened safeguards, added trajectory-level monitoring, and restored limited access under continued monitoring. ### Source excerpt OpenAI shares lessons from deploying long-running AI models, highlighting new safety risks, observed failures, and improved safeguards through iterative deployment. ## Keycloak 26.7.0 released DevFeed: [Keycloak 26.7.0 released]() Original publisher: [Read original article]() Author: Keycloak Team Published: 2026-07-09T00:00:00Z Content type: release Language: en Sources: [Keycloak Blog]() Topics: [Keycloak](), [API](), [Authentication](), [Authorization](), [Security](), [OAuth 2.0](), [saml]() Tags: [api](), [authentication](), [authorization](), [deprecated](), [haproxy](), [high-availability](), [idm](), [kerberos](), [keycloak](), [keycloak-release](), [ldap](), [oauth](), [oauth-2-0](), [openid-connect](), [persistence](), [provisioning](), [release](), [saml](), [security](), [sso](), [traefik]() ### AI overview Keycloak 26.7.0 introduces preview support for SCIM-based user provisioning and simplified multi-cluster high availability, along with enhanced HAProxy and Traefik reverse-proxy guidance and step-up authentication for SAML clients. The release also adds a more secure, OAuth 2.0-compliant Identity Brokering API v2, while V1 remains enabled by default but is deprecated. OID4VCI improvements remain experimental. ### Source excerpt To download the release go to Keycloak downloads. Highlights This release features new capabilities for users and administrators of Keycloak. The highlights of this release are: Automate user provisioning with the SCIM API (preview) Simplified multi-cluster high availability without external caches (preview) Enhanced reverse proxy guides with blueprints for HAProxy and Traefik Step-up authentication for SAML clients Read on to learn more about each new feature. If you are upgrading from a previous release, also review the changes listed in the upgrading guide. Security and Standards Stronger security for the Identity Brokering API The Identity Brokering API allows applications to retrieve tokens obtained from external identity providers during federated login. Version 2 of this API replaces the legacy V1 with a more secure and standards-compliant design: Client-level authorization -- access to external tokens is controlled per client using dedicated settings (Allow retrieve external tokens and an identity provider allow list) instead of assigning broker roles to individual users. Confidential clients only -- public clients are rejected, ensuring that only authenticated clients can retrieve external tokens. OAuth 2.0 compliant -- the endpoint uses POST and returns standard JSON responses with access_token, error, and error_description fields. Session-based token storage -- a new Store token in session option keeps tokens in the user session for faster access, with automatic cleanup on session expiry. Database storage remains available for persistence across sessions. V2 is now supported but disabled by default. V1 is deprecated but still enabled by default for backward compatibility. In a future release, V1 will be removed and V2 will become the default. For more information, see the Identity Brokering APIs chapter in the Server Developer Guide. Progress on Verifiable Credentials (OID4VCI) (experimental) Verifiable Credentials (OID4VCI) allow organizations to issue tampe ## Agent Memory at Monster Scale with Mem0 and ScyllaDB Cloud DevFeed: [Agent Memory at Monster Scale with Mem0 and ScyllaDB Cloud]() Original publisher: [Read original article]() Author: Cynthia Dunlop Published: 2026-07-07T13:08:54Z Content type: tutorial Language: en Sources: [ScyllaDB]() Topics: [AI Bots](), [AI search](), [AI, ML & Data Engineering]() Tags: [ai-agents](), [ai-ml](), [ecosystem](), [embeddings](), [how-to](), [llm](), [memory](), [persistence](), [scale](), [scylladb-cloud](), [search](), [vector]() ### AI overview The article explains how Mem0 and ScyllaDB Cloud can support persistent memory for large-scale AI agents. Mem0 extracts concise, entity-based facts from interactions, while ScyllaDB Cloud provides distributed, low-latency vector-search storage. ### Source excerpt Combine Mem0's memory management with ScyllaDB's persistence features to deploy large-scale AI agents ## Turbopack: What's New in Next.js 16.3 DevFeed: [Turbopack: What's New in Next.js 16.3]() Original publisher: [Read original article]() Author: Andrew Imm Published: 2026-06-29T20:00:00Z Content type: article Language: en Sources: [Next.js Blog]() Topics: [Next.js](), [Cache](), [Caching](), [build times](), [Compiler](), [Persistence](), [Rust](), [cpu]() Tags: [build-times](), [cache](), [caching](), [cpu](), [development](), [memory](), [next-js](), [performance](), [persistence](), [rust]() ### AI overview Next.js 16.3 improves the Turbopack bundler with lower development memory usage, persistent filesystem caching for faster builds, experimental Rust React Compiler support, API support, and faster hot module replacement and development startup. Memory-cache eviction, enabled by default with the development filesystem cache, helps prevent unbounded memory growth during long development sessions. ### Source excerpt Turbopack in Next.js 16.3 adds development memory eviction, persistent file cache for builds, experimental Rust React Compiler, and `import.meta.glob` support. ## Sandboxes now expire based on last use DevFeed: [Sandboxes now expire based on last use]() Original publisher: [Read original article]() Author: Marc Codina Segura Published: 2026-06-29T01:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [Vercel](), [Persistence]() Tags: [persistence](), [retention](), [sandboxes](), [snapshots](), [time](), [vercel](), [workflows]() ### AI overview Vercel Sandbox snapshots now expire according to their last use rather than their creation time. Using a snapshot resets its expiration timer, allowing active workflows to keep dependent snapshots available while unused snapshots follow their retention policy. ### Source excerpt Vercel Sandbox snapshots now expire based on when they were last used, not when they were created. Active snapshots stay alive as long as workflows depend on them, while unused snapshots expire on their retention policy. Every time a snapshot is used, its expiration timer resets. This lets you set shorter retention windows without worrying that a snapshot will disappear between sessions, making it safer to build long-running workflows on top of Sandbox persistence. Learn more about Sandbox snapshots in the documentation. Read more [Next page]()