gitlab·Feb 19, 2026Original article

Agentic AI, enterprise control: Self-hosted Duo Agent Platform and BYOM (opens in new tab)

GitLab 18.9 introduces critical updates designed to provide regulated enterprises with governed, agentic AI capabilities through self-hosted infrastructure and model flexibility. By combining the Duo Agent Platform with Bring Your Own Model (BYOM) support, organizations in sectors like finance and government can now automate complex DevSecOps workflows while maintaining total control over data residency. This release transforms GitLab into a high-security AI control plane that balances the need for advanced automation with the rigid sovereignty requirements of high-compliance environments. ## Self-Hosted Duo Agent Platform for Online Cloud Licenses The Duo Agent Platform allows engineering teams to automate sequences of tasks, such as hardening CI/CD pipelines and triaging vulnerabilities, but was previously difficult to deploy for customers under strict online cloud licensing. This update makes the platform generally available for these environments, bridging the gap between cloud-based licensing and self-hosted security needs. * **Usage-Based Billing:** The platform now utilizes GitLab Credits to provide transparent, per-request metering, which is essential for internal chargeback and regulatory reporting. * **Infrastructure Control:** Enterprises can host models on their own internal infrastructure or within approved cloud environments, ensuring that inference traffic is routed according to internal security policies. * **Deployment Readiness:** By removing the requirement to route data through external AI vendors, the platform is now a viable option for critical infrastructure and government agencies. ## Bring Your Own Model (BYOM) Integration Recognizing that many enterprises have already invested in domain-tuned LLMs or air-gapped deployments, GitLab now allows customers to integrate their existing models directly into the Duo Agent Platform. This ensures that organizations are not locked into a specific vendor and can leverage models that have already passed internal risk assessments. * **AI Gateway Connectivity:** Administrators can connect third-party or internal models via the GitLab AI Gateway, allowing these models to function as enterprise-ready options within the GitLab ecosystem. * **Granular Model Mapping:** The system provides the ability to map specific models to individual Duo Agent Platform flows or features, giving admins fine-grained control over which agent uses which model. * **Administrative Ownership:** While GitLab provides the orchestration layer, administrators retain full responsibility for model validation, performance tuning, and risk evaluation for the models they choose to bring. For organizations operating in high-compliance sectors, these updates offer a path to consolidate fragmented AI tools into a single, governed platform. Engineering leaders should evaluate their current model investments and leverage the GitLab AI Gateway to unify their automation workflows under one secure DevSecOps umbrella.

AI AgentsMachine LearningGitlabDevSecOps+3

spotify·Feb 19, 2026·3 min readCurated summary

Background Coding Agents: Predictable Results Through Strong Feedback Loops (Honk, Part 3) | Spotify Engineering

Spotify argues that unsupervised coding agents become reliable only when surrounded by strong, automated feedback loops. Its “Honk” system uses component-specific verifiers, mandatory pre-PR checks, and an LLM judge to catch build failures, test failures, scope creep, and functionally incorrect changes. The conclusion is that constrained, sandboxed agents with rich verification are more predictable than flexible agents operating independently. ## Failure Modes at Scale - Agents may fail to produce a pull request, which is inconvenient but usually manageable. - They may produce PRs that fail CI, leaving engineers to repair incomplete work. - Most seriously, they may produce PRs that pass CI but are functionally wrong and potentially reach production. - These failures are more likely when components lack tests, agents modify code beyond the prompt, or agents cannot correctly run builds and tests. - Reviewing invalid or nonsensical PRs can become a significant engineering time sink. ## Verification Loops - Honk uses independent verifiers that provide incremental feedback while the agent works. - Verifiers activate automatically based on the repository contents; for example, a Maven verifier runs when a root-level `pom.xml` is present. - The agent sees an abstract MCP tool rather than the implementation details of Maven, test runners, or build systems. - Verifiers handle formatting, compilation, testing, and output parsing, returning concise error messages instead of consuming the agent’s context with raw logs. - All applicable verifiers run before a PR is opened. In Claude Code, this is enforced with a stop hook. - If verification fails, the PR is blocked and the user receives an error. ## An LLM as a Judge - Deterministic checks cannot detect every problem, especially when an agent makes unnecessary refactors or disables flaky tests. - Honk therefore sends the original prompt and proposed diff to a separate LLM judge. - The judge runs after the regular verifiers and can veto changes that exceed the requested scope. - Across thousands of sessions, the judge rejects roughly one quarter of proposed changes. - Agents successfully correct about half of the vetoed changes. - Spotify has not yet built formal evaluations for the judge, but observed that scope violations are its most common reason for rejection. ## Constrained Agents and Sandboxing - The agent has limited responsibilities: inspect the relevant code, edit files, and invoke verification tools. - Surrounding infrastructure handles prompt creation, pushing code, and user communication through systems such as Slack. - Restricting the agent’s capabilities improves predictability and provides security benefits. - Agents run in heavily sandboxed containers with limited permissions, few installed binaries, and almost no access to surrounding systems. - Spotify reports that agents solve increasingly complex tasks reliably when these feedback loops are present, but often produce unusable code without them. ## Future Expansion - Spotify plans to support more hardware and operating systems. - Current verifiers run only on Linux x86, limiting support for systems that require macOS, such as iOS applications, or ARM64 environments. - The company also intends to integrate Honk more deeply with existing CI/CD pipelines. The practical recommendation is to treat autonomous coding as an infrastructure and verification problem, not merely a prompting problem: keep agents narrowly scoped, isolate them securely, and require layered automated checks before accepting their changes.

Read original(opens in new tab)

AI AgentsLarge Language ModelsCI/CDModel Context Protocol+4

github·Feb 13, 2026·3 min readCurated summary

Automate repository tasks with GitHub Agentic Workflows

GitHub Agentic Workflows bring coding agents into GitHub Actions, allowing developers to describe repository tasks in Markdown instead of complex YAML. They can automate issue triage, documentation, testing, code cleanup, CI investigation, and reporting while preserving human oversight through permissions, sandboxing, logging, and review. The post presents the technology, now in technical preview, as an extension of CI/CD rather than a replacement for deterministic build and release pipelines. ## Markdown-Defined Repository Automation - Developers describe desired outcomes in plain Markdown and add the workflow to a repository. - The workflow runs in GitHub Actions using configurable coding agents such as: - GitHub Copilot CLI - Claude Code - OpenAI Codex - Because workflows operate within GitHub Actions, they benefit from repository context, audit logs, permission controls, and sandboxed execution. ## Examples of Continuous AI GitHub describes these workflows as “Continuous AI”: AI-powered automation integrated throughout the software development lifecycle. - **Issue triage:** Summarize, label, and route new issues. - **Documentation maintenance:** Update READMEs and documentation after code changes. - **Code simplification:** Find opportunities for improvement and open pull requests. - **Test improvement:** Evaluate coverage and add valuable tests. - **Quality hygiene:** Investigate CI failures and suggest targeted fixes. - **Reporting:** Produce recurring reports on repository health, activity, and trends. These tasks are difficult to implement with traditional deterministic YAML workflows because they require interpretation, judgment, and code changes. ## Relationship to CI/CD - Agentic workflows are intended to augment, not replace, existing CI/CD systems. - Traditional pipelines remain responsible for deterministic builds, tests, and releases. - Agentic workflows handle higher-level tasks involving analysis, recommendations, and repository maintenance. - GitHub Actions provides the infrastructure needed for controlled execution and observability. ## Guardrails and Human Control - Security is presented as a core design requirement, particularly against unintended behavior and prompt injection. - Workflows run with read-only permissions by default. - Write operations require explicit approval through “safe outputs,” which are designed to make changes pre-approved and reviewable. - The overall approach emphasizes inspectability, defined boundaries, and human review rather than unrestricted autonomous changes. ## Adoption Across Teams - GitHub Next reports using workflows to replace repetitive chores and assemble useful information for developers. - Home Assistant uses them to analyze large numbers of issues and identify important trends. - The Cloud Native Computing Foundation applies them to documentation automation and organizational reporting. - Carvana uses them for engineering work spanning multiple repositories. GitHub Agentic Workflows are best viewed as a controlled way to add AI judgment to repository operations. Teams should begin with focused, reviewable maintenance tasks and expand usage as they gain confidence in the workflows’ behavior and safeguards.

Read original(opens in new tab)

AI AgentsCI/CDGitHub ActionsMarkdown+3

cloudflare·Feb 12, 2026·2 min readCurated summary

Introducing Markdown for Agents

AI agents increasingly need structured, efficient access to web content, making traditional HTML a costly format for machine consumption. Cloudflare’s Markdown for Agents lets enabled websites serve HTML pages as Markdown when clients request `text/markdown`, reducing token usage and parsing overhead. The post argues that websites should treat AI agents as first-class visitors alongside humans and search engines. ## Why Markdown Matters for AI - Markdown conveys document structure with far less surrounding markup than HTML. - A Markdown heading such as `## About Us` uses roughly 3 tokens, compared with 12–15 tokens for an equivalent HTML heading. - The post’s HTML uses about 16,180 tokens, while its Markdown version uses approximately 3,150—a reduction of about 80%. - Converting HTML to Markdown inside an AI pipeline adds computation, cost, and complexity, and may not preserve the publisher’s intended structure. ## How Markdown for Agents Works - Cloudflare-enabled zones can respond to content negotiation requests containing: ```http Accept: text/markdown ``` - Cloudflare fetches the original HTML from the origin, converts it to Markdown at the network edge, and returns the converted response. - Clients can request Markdown with `curl`, while Workers-based agents can use a `fetch()` request with `Accept: "text/markdown, text/html"`. - Responses use `Content-Type: text/markdown` and include `Vary: accept`. - Existing coding agents, including Claude Code and OpenCode, already send compatible `Accept` headers. ## Token Estimates and Agent Workflows - Converted responses include an `x-markdown-tokens` header. - Agents can use this estimate to: - Determine whether content fits within a context window - Plan chunking strategies - Manage processing costs and limits ## Content Signals - Markdown responses include: ```http Content-Signal: ai-train=yes, search=yes, ai-input=yes ``` - These signals indicate that the content may be used for AI training, search results, and AI input, including agentic applications. - Cloudflare says future versions will support custom Content Signal policies. ## Availability - Cloudflare enabled Markdown for Agents on its Developer Documentation and Blog. - AI crawlers and agents can test the feature by requesting those pages with `Accept: text/markdown`. Web publishers can make their content more accessible to AI systems by supporting Markdown negotiation, while agents should request `text/markdown` whenever available to reduce tokens, parsing work, and processing cost.

Read original(opens in new tab)

AI AgentsCloudflareCloudflare WorkersMarkdown+4

microsoft·Feb 11, 2026·4 min readCurated summary

How we built the Microsoft Learn MCP Server

Microsoft Learn MCP Server gives AI agents direct, standardized access to current Microsoft documentation through the Model Context Protocol (MCP). Rather than requiring custom APIs, scraping, or embeddings, agents can dynamically discover and use tools for searching documentation, fetching full articles, and finding code samples. Microsoft’s experience shows that successful MCP systems depend not only on retrieval quality, but also on agent-oriented tool design, operational resilience, clear descriptions, and defensive compatibility practices. ## Purpose of Learn MCP Server - Provides trusted, up-to-date Microsoft Learn content to GitHub Copilot and other AI agents. - Uses Streamable HTTP Transport so MCP-compatible clients can connect to a remote server. - Supports three tools: - `microsoft_docs_search` for titles, relevant content sections, and source URLs. - `microsoft_docs_fetch` for retrieving complete article content. - `microsoft_code_sample_search` for locating language-specific code examples. - Grounds agent responses in official Microsoft documentation rather than relying solely on model memory. ## Why MCP Instead of a Traditional API - Conventional APIs require each client to implement: - Authentication and request formatting. - Documentation and integration logic. - Error handling and compatibility maintenance. - MCP allows clients to discover available tools and schemas at runtime. - The same server can support many agents without custom integrations. - Runtime discovery helps clients adapt to evolving tool contracts and reduces hardcoded assumptions. ## Architecture - The remote MCP server sits in front of the Microsoft Learn knowledge service. - It uses the official C# MCP SDK and runs on Azure App Service. - Clients communicate through Streamable HTTP Transport. - The server uses the same content vector store as Ask Learn, providing shared: - Freshness guarantees. - Relevance ranking. - Index coverage. - Ask Learn delivers retrieval directly to users, while Learn MCP Server exposes that capability through a protocol usable by external agents. ## Designing Tools Around Agent Workflows - Internal retrieval APIs expose many low-level options, such as `topK`, index selection, thresholds, filters, and search modes. - Learn MCP Server hides that complexity behind intuitive search-and-fetch operations. - Tool contracts should reflect how agents work rather than mirror backend APIs. - Keeping retrieval details internal prevents implementation choices from leaking into the agent-facing interface. ## Operating a Remote MCP Service - A public MCP server has distributed-systems concerns despite using JSON-RPC: - Cross-region deployment. - Dynamic scaling. - CORS. - Session affinity. - Statelessness. - Data protection. - Operational design and SDK collaboration are as important as implementing the tools themselves. ## Tool Descriptions Shape Agent Behavior - Tool and parameter descriptions act as instructions for language models. - Small wording changes can significantly affect whether agents select a tool and how successfully they use it. - Microsoft created automated evaluation tooling to test descriptions against observed agent behavior and success metrics. - Updated descriptions can be delivered when clients refresh their MCP sessions. ## Combining Search and Fetch - Search and fetch are more effective together than independently. - A typical workflow is: - Search for the most relevant Learn article or section. - Fetch the full Markdown page for additional context. - Use that content to produce a better-grounded answer with stronger citations. - Explicitly describing this follow-up pattern improved downstream results. ## Handling Hardcoded Clients - Some MCP clients treat discovered tools like fixed APIs and hardcode schemas. - Renaming the `question` parameter to `query` caused 2–5% of requests to fail. - Supporting both names during a deprecation period reduced disruption. - Public MCP services must evolve defensively, even though the protocol supports dynamic discovery. - Tools such as MCP Interviewer can help identify schema and behavioral problems before deployment. ## Using Data to Guide Improvements - Usage data showed that most requests involve: - Coding tasks. - Explanations. - Troubleshooting. - The team prioritized retrieval and description changes around these intents. - Documentation-level agent instructions also encourage use of Learn tools when Microsoft technologies are involved. Microsoft Learn MCP Server replaces the manual process of searching, opening, and copying documentation into a development environment. The practical recommendation is to connect compatible agents to the server so they can retrieve official Learn content directly, while MCP tool authors should design simple contracts, measure real agent behavior, and preserve compatibility as their services evolve.

Read original(opens in new tab)

AI AgentsModel Context ProtocolGithub CopilotRAG+4

spotify·Feb 9, 2026·3 min readCurated summary

Our Multi-Agent Architecture for Smarter Advertising | Spotify Engineering

The post argues that fragmented advertising workflows, not backend infrastructure, are the core problem. Although buying channels share services and data, their planning and optimization logic is repeatedly reimplemented across channels and surfaces, causing drift and technical debt. The proposed solution is a shared agentic decision layer that interprets advertiser goals, orchestrates existing Ads APIs, and applies consistent reasoning across products. ## Fragmented Workflows Across a Shared Backend - Direct, Self-Serve, and Programmatic buying use largely consolidated infrastructure but retain different workflows and decision logic. - Spotify Ads Manager, Salesforce, Slack, and internal tools contain overlapping automation. - Budget allocation, inventory selection, reach, efficiency, and STR decisions are repeatedly implemented in different places. - Incremental workflow changes therefore create duplicated maintenance work and inconsistent behavior. ## Why Conventional Workflow Services Fall Short - Hard-coded state machines and REST services are poorly suited to combinatorial planning tasks. - Campaign planning depends on: - User and advertiser characteristics - Available inventory and audiences - Business priorities - Forecasts, performance, and optimization goals - A workflow optimized for one channel or “happy path” will not adapt well as requirements change. - Improvements to decision logic must be replicated across every product surface, increasing the risk of divergence. ## The Missing Intent Layer - Existing systems can perform individual actions such as creating line items, running forecasts, and retrieving insights. - They do not consistently translate high-level objectives into: - A sequence of tool calls - Explicit tradeoffs - Validation and safety checks - An objective such as maximizing reach in Brazil while protecting video inventory and meeting STR requires coordinated reasoning across multiple capabilities. ## A Modular Agentic Architecture - Campaign planning and management are modeled as cooperating specialized agents. - Agents use shared signals, including: - Inventory - Audiences - STR - Quality and risk - Historical performance - They jointly optimize advertiser goals and Spotify’s business constraints. - Existing Ads services become tools that agents orchestrate, rather than capabilities being rebuilt in each workflow. - A long-running orchestration layer delegates tasks while agents share context and evaluation logic. - The same decision engine can support every buying channel and surface. ## Engineering Implications - APIs need to be designed as agent tools, rather than only as CRUD interfaces. - Testing must include behavioral evaluation in addition to unit and integration tests. - Observability should explain what an agent decided and why, not merely track latency and errors. - Safety requires guardrails for semi-autonomous decisions, beyond ordinary input validation. - The approach avoids both duplicated deterministic workflows and a brittle, centralized rules engine for probabilistic, ML-heavy advertising logic. The overall recommendation is to centralize campaign decision-making in a reusable agentic platform while keeping existing services as specialized tools. This should reduce duplicated workflow logic, make improvements consistent across products, and allow advertising workflows to evolve without repeatedly rebuilding them.

Read original(opens in new tab)

AI AgentsMachine LearningAPI DesignMulti Agent Systems+3

github·Feb 5, 2026·3 min readCurated summary

Continuous AI in practice: What developers can automate today with agentic CI

Continuous AI extends CI into software-engineering tasks that require judgment, context, and interpretation rather than deterministic rules. It uses continuously running agents guided by natural-language instructions to review repositories, identify issues, and produce reviewable artifacts such as patches, issues, or reports. GitHub’s central argument is that AI should complement—not replace—traditional CI, while operating within explicit permissions and developer oversight. ## Why CI Isn’t Enough - CI is effective for binary, rule-based checks: - Tests pass or fail. - Builds succeed or fail. - Linters detect defined violations. - Many important engineering tasks depend on intent and context, including: - Finding discrepancies between documentation and implementation. - Detecting confusing accessibility text that passes linting. - Identifying behavioral changes caused by dependency updates. - Spotting subtle performance regressions, such as compiling a regular expression inside a loop. - Recognizing UI regressions that only appear during interaction. - GitHub describes this as a shift from AI-generated code toward AI handling cognitively demanding maintenance work. ## What Continuous AI Means - Continuous AI is a pattern, not a replacement for CI: - **Natural-language rules + agentic reasoning, executed continuously inside a repository.** - Developers describe expectations in natural language, especially when those expectations are difficult to encode with schemas, heuristics, or YAML. - Example workflows include: - Comparing documented behavior with implementation and proposing fixes. - Producing weekly reports on project activity, bug trends, and code churn. - Detecting performance regressions in critical paths. - Finding semantic regressions in user flows. - Workflows are refined collaboratively with agents by adding intent, constraints, and acceptable outputs rather than being authored as a perfect single instruction. ## Guardrails and Safe Outputs - Agents operate with read-only repository access by default. - They cannot modify content, create issues, or open pull requests unless explicitly authorized. - “Safe Outputs” defines the exact artifacts an agent may produce and the constraints governing them. - Agent activity is sanitized, logged, and auditable. - The goal is to keep the potential impact predictable even when agents make mistakes or behave unexpectedly. ## Natural Language Complements YAML - Deterministic problems should remain in CI, using YAML, schemas, tests, and heuristics. - Some expectations—such as determining whether documentation and code still express the same behavior—require semantic understanding. - Natural-language instructions let agents reason about intent without forcing that intent into brittle rules. - Continuous AI therefore expands automation into judgment-heavy tasks while preserving CI as the foundation for deterministic validation. ## Developers Remain in the Loop - Agents do not make unrestricted autonomous commits. - Depending on permissions, they can produce pull requests, issues, comments, discussions, or other reviewable artifacts. - Pull requests are especially useful because they fit existing developer review and collaboration practices. - The broader vision is to delegate recurring maintenance work while allowing developers to retain judgment, taste, and final control. Continuous AI is best adopted alongside traditional CI: use conventional automation wherever rules are sufficient, and use guarded, continuously running agents for tasks involving interpretation, synthesis, and evolving intent.

Read original(opens in new tab)

AI AgentsCI/CDGithubCode Review+3

google·Feb 5, 2026·3 min readCurated summary

How AI tools can redefine universal design to increase accessibility

Google Research proposes Natively Adaptive Interfaces (NAI), a framework that uses multimodal and agentic AI to make interfaces adapt to individual users rather than forcing everyone into a fixed design. Developed through co-design with disability communities, NAI aims to reduce the accessibility gap by embedding assistive capabilities directly into products. Early prototypes suggest that personalized, context-aware interfaces can improve experiences for disabled users while also benefiting the broader population. ## Community-led co-design - Google follows the principle “Nothing About Us, Without Us,” involving people with disabilities as co-designers from the beginning. - Partnerships include RIT/NTID, The Arc of the United States, RNID, and Team Gleason. - These collaborations focus on real-world barriers and recognize the expertise of disability communities. - The approach also aims to create employment and economic opportunities for people who help shape the technology. ## Moving from reactive accessibility to adaptive interfaces - Google identifies an “accessibility gap” between the release of new features and the development of compatible assistive tools. - NAI addresses this by making accessibility native to the interface instead of adding it afterward. - Static navigation is replaced with dynamic, agent-driven modules that can interpret context and adjust the experience. ## Multi-system agents - An Orchestrator maintains shared context and delegates tasks to specialized sub-agents. - A Summarization Agent breaks down complex documents and assigns subtasks to expert agents. - A Settings Agent dynamically adjusts interface elements such as text size. - This structure lets users accomplish tasks without navigating complicated menus or searching for the right control. ## Multimodal interaction - Gemini-based prototypes combine voice, vision, and text rather than limiting accessibility to text-to-speech. - Live video can be converted into interactive audio descriptions. - Users can ask follow-up questions about specific visual details as events unfold. - Conversational interaction provides situational awareness and may reduce cognitive load. ## Proven prototypes - **StreetReaderAI** - Supports blind and low-vision users navigating physical spaces. - Combines an AI Describer that analyzes visual and geographic information with an AI Chat system for questions. - Maintains context so users can ask about previously encountered locations, such as the position of a bus stop. - **Multimodal Agent Video Player (MAVP)** - Makes audio description interactive rather than static. - Users can change the level of detail or ask questions during playback. - Uses an offline “dense index” of visual descriptions and retrieval-augmented generation (RAG) for fast responses. - **Grammar Laboratory** - Developed by RIT/NTID with Google.org support for American Sign Language and English learners. - Provides grammar instruction through ASL videos, English captions, spoken narration, and written transcripts. - Uses adaptive AI to customize lessons according to each student’s language preferences and interactions. ## The curb-cut effect - Accessibility features designed for people with significant constraints can benefit many other users. - Voice interfaces created for blind users may help sighted people who are multitasking. - AI synthesis and learning tools designed for people with learning disabilities can also support users who want information presented more clearly or flexibly. - NAI therefore treats accessibility as a source of better universal design, not as a specialized add-on. NAI’s central recommendation is to build accessibility into interfaces from the start, using multimodal AI, persistent context, and community-led design. The most effective systems will adapt to users while remaining accountable to the people whose needs they are intended to serve.

Read original(opens in new tab)

AI AgentsGenerative AIAccessibilityGemini+4

aws·Feb 2, 2026·3 min readCurated summary

AWS Weekly Roundup: Amazon Bedrock agent workflows, Amazon SageMaker private connectivity, and more (February 2, 2026) | Amazon Web Services

The AWS Weekly Roundup highlights new capabilities for AI agents, private connectivity, encryption management, and resilience testing. Major launches include Bedrock server-side tools and longer prompt caching, SageMaker Unified Studio support for PrivateLink, and S3 encryption changes without data movement. Additional updates strengthen event-driven architectures, observability, zero-trust access, and AI-assisted AWS deployments. ## AI Agents and Developer Workflows - Amazon Bedrock’s Responses API now supports server-side tools such as web search, code execution, and database updates within AWS security boundaries. - Bedrock also offers a one-hour prompt-cache TTL for select Anthropic Claude models, improving performance and reducing costs for long-running, multi-turn agents. - AWS MCP Server deployment SOPs, currently in preview, let agents deploy applications from natural-language prompts using CDK, CloudFormation, and CI/CD workflows. - The deployment preview supports React, Vue.js, Angular, and Next.js through tools such as Kiro, Cursor, and Claude Code. - CloudWatch Application Signals integration with Kiro provides AI-assisted investigation of service health, SLO compliance, and observability issues. ## Private Connectivity and Zero-Trust Security - SageMaker Unified Studio now supports AWS PrivateLink, allowing VPC traffic to remain within the AWS network instead of traversing the public internet. - IAM policies can govern private SageMaker connectivity for stricter security and compliance requirements. - AWS Verified Access guidance demonstrates centralized zero-trust application access across multi-account environments using IAM Identity Center and AWS RAM. - AWS Network Firewall adds predefined web categories for identifying and controlling generative AI application traffic, with full-URL filtering available alongside TLS inspection. ## Storage, Encryption, and Database Performance - Amazon S3’s `UpdateObjectEncryption` API changes encryption for existing objects without moving or re-uploading data. - Supported operations include switching from SSE-S3 to SSE-KMS, rotating customer-managed KMS keys, and standardizing encryption with S3 Batch Operations. - Amazon Keyspaces table pre-warming prepares tables for predictable high-throughput workloads, reducing throttling and cold-start delays during traffic spikes. - Pre-warming works with on-demand and provisioned capacity, including multi-Region tables. - DynamoDB MRSC global tables now integrate with AWS Fault Injection Service, enabling simulated Regional failures and validation of replication and application resilience. ## Event-Driven Systems and Observability - EventBridge’s event payload limit increased from 256 KB to 1 MB, allowing events to carry richer JSON, telemetry, ML, and generative AI data without external storage or fragmentation. - Lambda’s enhanced observability for Kafka event source mappings adds CloudWatch logs and metrics for polling, scaling, processing state, permissions, and failures. - The feature supports both Amazon MSK and self-managed Apache Kafka sources. ## CloudFormation and Community - AWS’s 2025 CloudFormation review covers improved troubleshooting, drift-aware change sets, stack refactoring, StackSets, the CloudFormation language server, and IaC MCP tooling. - AWS Community Day Romania will take place April 23–24, 2026, featuring technical sessions, AWS experts, and networking opportunities. Together, these updates point toward more private, observable, resilient, and AI-assisted AWS operations. Teams should evaluate the new capabilities against their security, scalability, and automation needs, particularly Bedrock agent tooling, S3 encryption updates, PrivateLink connectivity, and resilience testing.

Read original(opens in new tab)

AI AgentsGenerative AIObservabilityAmazon Bedrock+4

cloudflare·Jan 30, 2026·3 min readCurated summary

Google’s AI advantage: why crawler separation is the only path to a fair Internet

Google’s dominance in search gives it a structural advantage in generative AI: publishers must allow Googlebot to preserve search visibility, while Google can also reuse that access for AI products. The authors argue that this blurs search indexing and AI data collection, deprives publishers of traffic and compensation, and disadvantages competing AI companies. They support the CMA’s proposed UK conduct rules but say the only fair solution is to separate crawling for search from crawling for generative and agentic AI. ## CMA’s Strategic Market Status designation - The UK’s Digital Markets, Competition and Consumers Act 2024 allows the CMA to designate firms with substantial, entrenched market power as having Strategic Market Status. - In October 2025, Google received this designation for general search and search advertising, where it holds roughly 90% of the UK market. - The designation covers AI Overviews and AI Mode, allowing the CMA to impose legally enforceable conduct requirements on Google’s search ecosystem. - The authors view the CMA’s consultation as an important first step toward clearer rules for AI crawling and publisher control. ## Problems with Google’s dual-purpose crawler - Publishers cannot realistically block Googlebot because doing so could reduce their visibility in Google Search and damage advertising revenue. - Google uses the same search access not only for indexing and referrals, but also to ground AI Overviews, AI Mode, and broader generative AI services. - These AI features may reproduce publisher content while sending little or no traffic back to the original sites. - This threatens ad-supported publishing models and can put Google in direct competition with the publishers whose content it uses. - Unlike other AI companies, Google can obtain large amounts of content without negotiating payment, because publishers are effectively unable to refuse its search crawler. ## Google’s crawling advantage Cloudflare’s data indicates that Googlebot accesses substantially more unique pages than other major AI crawlers: - About 1.7 times more than ClaudeBot and GPTBot. - About 3 times more than Meta-ExternalAgent. - About 3.3 times more than Bingbot. - About 5.1 times more than Amazonbot. - Nearly 15 times more than Applebot. - Nearly 167 times more than PerplexityBot. - More than 700 times more than CCBot. - More than 1,800 times more than archive.org_bot. - Googlebot crawled roughly 8% of the sampled unique URLs during the two-month observation period. ## Limits of robots.txt and the need for separate controls - Publishers are much less likely to block Googlebot in `robots.txt` because of its importance for search referrals. - `robots.txt` expresses preferences but does not technically enforce crawler behavior; publishers must rely on bots to comply. - Web Application Firewalls can technically block unwanted crawlers, but this does not solve the core problem when search and AI access are tied to the same Googlebot identity. - The authors therefore argue that publishers need a meaningful, independent way to permit Google Search indexing while refusing the use of their content for generative AI. The proposed CMA rules should go further by requiring effective separation between search crawling and AI crawling. Publishers should be able to opt out of generative AI use without sacrificing search visibility, creating fairer conditions for content creators and competing AI developers.

Read original(opens in new tab)

AI AgentsGenerative AIWeb Application FirewallWeb Crawling+3

cloudflare·Jan 29, 2026·2 min readCurated summary

Introducing Moltworker: a self-hosted personal AI agent, minus the minis

Moltworker adapts the self-hosted Moltbot personal AI assistant to run on Cloudflare without requiring users to buy a dedicated Mac mini. It combines a Cloudflare Worker, Sandbox SDK, Browser Rendering, R2, AI Gateway, and Cloudflare Access to provide a globally available, secured deployment. The result is a managed infrastructure layer around Moltbot’s standard Gateway runtime while preserving its integrations and persistent state. ## Running a Personal Agent on Cloudflare - Cloudflare Workers increasingly supports Node.js APIs natively, reducing the need for compatibility hacks and making it easier to run existing JavaScript and TypeScript packages. - An internal test of the 1,000 most popular NPM packages found that only 15 relevant packages failed to run in Workers. - Although much of Moltbot runs inside a container, improved Workers compatibility is useful for building agent logic closer to users. - Cloudflare’s Developer Platform provides the main infrastructure components: - **Sandboxes** for securely running untrusted code. - **Browser Rendering** for automated headless browser interactions. - **R2** for persistent object storage. - Cloudflare’s global network for scalability and security. ## Moltworker Architecture - Moltworker consists of: - An entrypoint Worker serving as an API router and proxy. - Cloudflare Access protecting the Worker and administration interface. - A Sandbox container running Moltbot’s standard Gateway and integrations. - R2 for persistent storage. - This structure separates the public API and administrative layer from the isolated environment where the agent executes. ## AI Gateway Integration - Cloudflare AI Gateway proxies requests between Moltbot and AI providers. - It provides: - Centralized request visibility. - Cost monitoring, logs, and analytics. - Provider and model switching without changing Moltbot code. - Fallback providers or models for improved reliability. - Secrets can be managed through: - **Bring Your Own Key (BYOK)**, where provider credentials are stored centrally. - **Unified Billing**, where users purchase credits and Cloudflare handles provider billing. - Integration requires creating an AI Gateway instance, enabling a provider such as Anthropic, and setting `ANTHROPIC_BASE_URL`; Moltbot itself does not need code changes. ## Sandbox-Based Execution - The Sandbox SDK runs agent code in isolated environments built on Cloudflare Containers. - It provides simplified APIs for: - Executing commands. - Managing files and directories. - Running background processes. - Exposing services. - Executing code in contexts such as Python. - The SDK abstracts container lifecycle, networking, filesystem, and process-management concerns behind TypeScript APIs. Moltworker offers a way to run a capable personal AI agent online with managed security, storage, browser automation, and model access—without maintaining dedicated hardware.

Read original(opens in new tab)

AI AgentsCloudflare WorkersServerlessNode.Js+4

grammarly·Jan 29, 2026Original article

How to Use AI Agents: A Simple Guide to Getting Started (opens in new tab)

AI agents represent a shift from reactive, prompt-based AI to proactive, goal-oriented systems capable of planning and executing multi-step tasks with minimal oversight. By operating in a continuous loop of gathering context, selecting tools, and evaluating results, these agents can manage complex workflows that previously required manual follow-up. The most effective implementation strategy involves starting with small, repeatable processes and gradually increasing agent autonomy as reliability is proven through feedback and testing. ### The Mechanism of Agentic AI * Unlike traditional generative AI that responds to isolated instructions, agents possess "agency," allowing them to decide the next best action to reach a defined objective. * Agents function through an iterative operational cycle: they analyze relevant context, select an action, utilize available tools, and evaluate the outcome to determine if the goal is met. * Advanced writing agents, such as those integrated into workplace tools, can proactively suggest revisions for tone, logical progression, and specificity by maintaining contextual awareness across a document's lifecycle. ### Deploying Agents via Repeatable Workflows * Initial use cases should focus on contained, well-understood tasks rather than end-to-end process overhauls to ensure the agent’s logic can be easily monitored. * In research and organization, agents can be tasked with continuously gathering and categorizing sources, updating citations as new data becomes available. * Communication workflows benefit from agents that can reference historical conversation threads to draft follow-ups, summarize long discussions, and adjust meeting agendas dynamically. * Content creation agents can manage the transition from rough notes to structured outlines, applying specific tone and clarity feedback across multiple versions of a draft. ### Integration and Tool Selection * Effective deployment often requires no coding experience, as agentic capabilities are increasingly built into existing word processors, email clients, and project management platforms. * Using familiar software ecosystems reduces the technical barrier to entry and allows for easier scaling of the agent’s behavior over time. * Project management agents can be utilized to monitor task progress, adjust timelines based on changing conditions, and surface high-priority items automatically. ### Establishing Goals and Ownership * Success depends on defining specific end states rather than vague instructions; for example, asking an agent to "flag logical gaps and suggest supporting evidence" is more effective than asking it to "improve writing." * Defining clear ownership ensures the agent knows which parameters to prioritize, such as maintaining a consistent brand voice while revising for conciseness. * Testing should begin with small-scale scenarios, like a single recurring email update, to allow for the refinement of instructions and priorities based on real-world performance. ### Scaling Autonomy and Oversight * Once an agent demonstrates consistent accuracy in a narrow task, its scope can be broadened to include related steps, such as tracking data throughout the week to prepare a draft before being prompted. * Increased autonomy does not mean a lack of control; humans should remain in the loop to provide feedback, which the agent uses to refine its future decision-making logic. * The transition from prompts to progress is achieved by allowing agents to work across different tools and contexts as they prove their ability to handle more complex judgment calls. To get the most out of AI agents, treat them as collaborative partners by starting with a narrow focus and providing specific, goal-oriented feedback. Rather than handing off entire processes immediately, focus on delegating repeatable tasks where the agent’s ability to plan and adapt can yield the highest immediate value.

AI AgentsAutomationProject ManagementCommunication Tools+2

grammarly·Jan 29, 2026Original article

Agentic AI vs. generative AI: What’s the Difference and When to Use Each (opens in new tab)

While generative AI focuses on creating content like text and images through prompt-based prediction, agentic AI represents a shift toward autonomous goal achievement and execution. By combining the creative output of large language models with a continuous loop of perception and action, these technologies allow users to move from simply generating drafts to managing complex, multi-step workflows. Ultimately, the two systems are most effective when used together, with one providing the ideas and the other handling the coordination and follow-through. ### Distinguishing Creative Output from Autonomous Agency * Generative AI functions as a responder that produces new content—such as text, code, or visuals—by predicting the most likely next "token" or piece of data based on a user’s prompt. * Agentic AI possesses "agency," meaning it can take a high-level goal (e.g., "prepare a client kickoff") and determine the necessary steps to achieve it with minimal guidance. * While tools like Midjourney or GitHub Copilot focus on the immediate delivery of a specific creative asset, agentic systems act as proactive partners that can use external tools, manage schedules, and make independent decisions. ### The Underlying Mechanics of Prediction and Action * Generative models rely on Large Language Models (LLMs) trained on massive datasets to identify patterns and chain together original sequences of information. * Agentic systems operate on a "perceive, plan, act, and learn" loop, where the AI gathers context from its environment, executes tasks across different applications, and adjusts its strategy based on the results. * The generative process is typically a direct path from input to output, whereas the agentic process is iterative, allowing the system to adapt to changes and feedback in real-time. ### Practical Applications in Content and Workflow Management * Generative use cases include transforming rough bullet points into polished emails, summarizing long documents into flashcards, and adjusting the tone of a message to be more professional. * Agentic use cases involve higher-level orchestration, such as monitoring document revisions, consolidating feedback from multiple stakeholders, and automatically sending follow-up reminders. * In a project management context, an agentic system can draft a project plan, identify owners for specific tasks, and update timelines as milestones are met or missed. ### Navigating Technical and Operational Limitations * Generative AI is susceptible to "hallucinations" because it prioritizes probabilistic output over factual reasoning or logic. * Agentic AI introduces complexity regarding security and permissions, as the system needs authorized access to various apps and tools to perform actions on a user's behalf. * Current agentic systems still require human oversight for critical decision-making to ensure that autonomous actions align with the user's intent and organizational standards. To maximize efficiency, you should utilize generative AI for the creative phases of a project—such as brainstorming and drafting—while delegating administrative overhead and coordination to agentic AI. As these technologies continue to converge, the focus of AI utility is shifting from the volume of content produced to the successful execution of complex, real-world results.

AI AgentsLarge Language ModelsGenerative AIAutomation+2

microsoft·Jan 28, 2026·4 min readCurated summary

Diagnosing instability in production-scale agent reinforcement learning

Hugging Face has integrated its Post-Training Toolkit into TRL, bringing production-ready diagnostics to reinforcement learning and agent post-training pipelines. The work identifies a late-phase instability specific to tool-using, on-policy agents: variance can grow in post-tool contexts even while loss, reward, entropy, and global KL remain stable. Targeted tail, distributional, and effective-sample-size diagnostics can expose this failure before it becomes divergence. ## Production Monitoring for Long-Running Agents - Modern agent training runs over long horizons, uses external tools, and adapts continuously. - Failures often develop gradually rather than appearing as a single catastrophic event. - Standard aggregate metrics can hide rare but increasingly severe updates. - The proposed monitoring approach: - Computes diagnostics in-stream. - Separates text-only and post-tool interactions. - Aggregates statistics across workers. - Uses lightweight rolling windows and percentile tracking at fixed intervals. ## Tool-Conditioned Variance Amplification - Tool calls expand the state space through external transitions, exposing the policy to contexts it may rarely encounter in the reference distribution. - Training states can be modeled as: `d(s) = (1−α)·d_text(s) + α·d_tool(s)` - As the proportion of tool-conditioned states, `α`, increases, more updates occur where the reference policy assigns low probability to sampled actions. - This causes importance-weighted updates to develop increasingly large tails. - The mechanism is distinct from global entropy collapse or optimizer instability, though those factors may interact with it. ## Minimal Reproduction and Tail Diagnostics - A small on-policy experiment with an instruction-tuned open-weight model reproduced the pattern. - The 95th percentile of absolute per-token log-ratios, `|r|`, was tracked separately for text-only and post-tool contexts. - Findings included: - Text-only tail magnitudes remained stable or declined. - Post-tool tails grew steadily under fixed-policy baselines. - Drift-aware training substantially reduced tail growth. - Constraining tool outputs also suppressed the effect. - Aggregate loss, reward, and entropy remained stable while the tail was expanding. ## Distributional Shift in the Right Tail - Empirical CDFs across early, middle, and late training showed a change in distribution shape rather than a simple threshold crossing. - In tool-conditioned contexts: - The right tail flattened and stretched. - More probability mass moved toward high-magnitude updates. - Drift-aware baselines muted or reversed the shift. - This supports a distributional explanation rather than an artifact of choosing a particular percentile. ## Importance Ratios and Effective Sample Size - For ratio-based on-policy objectives, gradient variance is related to: `Var[ĝ] ∝ E[(π_θ(a|s) / π_ref(a|s))²]` - When `π_ref(a|s)` is small in tool-conditioned states, a small number of updates can dominate the estimator. - Larger batches and improved baselines may reduce noise but do not fix poor support overlap. - Effective sample size (ESS) provides a supporting signal: - ESS declines as importance weights become concentrated. - It is sensitive to window size and batch structure. - Its trends align with post-tool tail growth, but absolute values should not be over-interpreted. ## Delayed Failure and Misdiagnosis - Instability appears first in tool-conditioned contexts and may remain invisible in global metrics for a long time. - By the time aggregate metrics change, substantial variance amplification may already have accumulated. - The problem is often incorrectly attributed solely to optimizer behavior or inadequate global variance reduction. - Such interventions may delay failure without addressing the underlying support mismatch. - The mechanism is less pronounced when tool outputs are tightly constrained, policies are effectively frozen after tool calls, or interaction diversity plateaus early. The practical recommendation is to add slice-aware, tail-focused diagnostics to production TRL pipelines. Monitoring post-tool log-ratio percentiles, distributional changes, and supporting ESS trends can provide an early warning system for instability that global loss, reward, entropy, and KL metrics miss.

Read original(opens in new tab)

AI AgentsReinforcement LearningPost TrainingTrl+3

grammarly·Jan 28, 2026Original article

AI Assistants vs. AI Agents: What’s the Difference and When to Use Each (opens in new tab)

While AI assistants and agents often share the same large language model foundations, they serve distinct roles based on their level of autonomy and task complexity. Assistants operate on a reactive "prompt-response" loop for immediate, single-step tasks, whereas agents function as semi-independent systems capable of planning and executing multistep workflows to achieve a broader goal. Ultimately, the most effective AI strategy involves leveraging assistants for quick, guided interactions while utilizing agents to manage complex, coordinated projects that require memory and tool integration. ### Reactive vs. Proactive AI Architectures * Assistants are reactive tools that follow a "prompt-response" loop, similar to a tennis match where the user must always serve to initiate action. * Agents are proactive and semi-independent; once given a high-level goal, they can decompose it into actionable steps and execute them with minimal step-by-step direction. * In a practical scenario, an assistant might summarize meeting notes upon request, whereas an agent can organize those notes, assign tasks in a project management tool, and schedule follow-ups automatically. ### Technical Capabilities and Coordination * Both tools utilize Large Language Models (LLMs) to understand natural language, but agents incorporate advanced features like long-term memory and cross-app integrations. * Memory allows agents to retain feedback and results from previous interactions to deliver better outcomes over time, while integrations enable them to act on the user's behalf across different software platforms. * The two systems often work in tandem: the assistant acts as the front-facing interface (the "waiter") for user commands, while the agent acts as the back-end engine (the "kitchen") that performs the orchestration. ### Balancing Control and Complexity * AI assistants provide high user control and instant setup, making them ideal for "out of the box" tasks like grammar checks, rephrasing text, or answering quick questions. * AI agents excel at reducing cognitive load by managing "moving parts" like deadline tracking, organizing inputs from different stakeholders, and maintaining project states across various tools. * Grammarly’s implementation of agents serves as a technical example, moving beyond simple text revision to offer context-aware suggestions that help with brainstorming, knowledge retrieval, and predicting audience reactions. To maximize productivity, users should delegate isolated, high-control tasks to AI assistants while allowing AI agents to handle the background orchestration of complex projects. Success with these tools depends on maintaining human oversight, using assistant-led prompts to provide the regular feedback that agents need to refine their autonomous workflows.

AI AgentsLarge Language ModelsAutomationNatural Language Processing+2