How to Run an Agent Loop That Actually Converges
A four-checkpoint how-to for agent loops: name a measurable job, bound the tool surface, refresh sandbox fixtures from real traffic, and verify at agent speed.
Loading…
Topic
AI Agents
561 posts about AI Agents. Every summary links to the original.
- 561 posts
- 41 companies
- 28 explainers
See the idea
AI Agents explainers
The AMR Surveillance Stack: From Raw Reads to Governed Answers
Antimicrobial resistance surveillance is a layered data architecture: raw reads and demand signals flow in, get normalized into governed tables, and answer plain-language queries. The diagram maps each boundary.
Where Amazon Bedrock Sits: A Layered Architecture for Grounded Agents
Amazon Bedrock is the managed reasoning tier in an agent architecture, not the whole system. This layered view shows the pipelines, governed tables, query layer, and guardrails around it.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Posts
How Notion built a Custom Agent workflow to keep Academy content current
Notion built a Custom Agent workflow in Notion to keep Academy content current as product changes ship. The system uses two databases: an Academy library mapping the curriculum and a maintenance-tasks database tracking work from each product change. Two Custom Agents handle the work: the Academy Drift Agent compares launch docs against the library and flags affected lessons, while the Academy Drafting Agent drafts approved updates using editorial guidance. MCP connects the workflow to Parta, the e-learning platform, so approved content moves into final QA without manual hand-off. The loop includes human review checkpoints before scope approval and before content moves into Parta, reducing content drift and giving the team hours back each week.
Lower the Cost of Building and Running Visual AI Agents with NVIDIA VSS Blueprint 3.3
Vision-language models enable production-scale video understanding, but building maintainable systems that combine ingestion, stream processing, event detection, retrieval, summarization, and reporting with services like Kafka, Redis, and Elasticsearch is challenging. The NVIDIA VSS Blueprint 3.3 introduces the Build Vision Agent skill, which composes multi-workflow applications from a single natural-language prompt using one of four Foundation profiles. It also adds Adaptive EVS, which prunes unchanged video patches via cosine similarity before VLM processing and batches tokens around events. Adaptive EVS cuts alert latency by 17%, delivers 80% fewer VLM input tokens for a 60-minute summary, and enables 46% more concurrent streams on the same GPU. These updates reduce development effort, infrastructure duplication, and operating costs, allowing a single prompt to build a bottling-line agent in under 30 minutes.
Getting the Source Right, Not Just the Fact: Source-Aware Verification for MCP Agents
MCP agents can produce claims that are factually true but attributed to the wrong source, a failure known as cross-source conflation that source-blind checkers miss. ProvenanceGuard preserves source identity by decomposing answers into claims, routing each to its relevant tool output, and checking support and attribution separately, using local models like MiniLM and DeBERTa. On medical agent traces with 281 interactions, it caught 138 of 139 claims experts flagged, scoring 0.802 F1 on blocking decisions and correct source 86% of the time. It flagged 67 supported claims for review under a cautious policy and added roughly half a second per answer in local tests. Blocked answers can be fixed through a repair loop for rewriting or fallback, avoiding unverifiable responses while keeping sensitive traces in a controlled environment.
Introducing Threat Signals: agentic skills for open-source threat intelligence, free for every Cloudflare account
Cloudflare introduced Threat Signals, a free agentic skill layer for open-source threat intelligence available to every account. The product targets the unstructured-reporting problem: converting research posts into indicators that security tools can consume while preserving the context that explains their relevance. A skill encodes detailed instructions that capture how an experienced analyst handles one part of the job, and it runs identically on every report. The agentic skills summarize reports, surface key context, extract and normalize indicators of compromise, and apply tags within a private, account-scoped dataset. This approach standardizes analyst expertise into repeatable, context-preserving processing for threat reports.
ClickHouse Workload for Microsoft Fabric: sub-second analytics on OneLake, now in public preview
The ClickHouse workload for Microsoft Fabric brings ClickHouse as a native acceleration layer for OneLake, now in public preview. It provisions dedicated ClickHouse Cloud services on Azure via the Fabric Extensibility Toolkit, using a ClickHouse item in the workspace. Users sync OneLake tables through ClickPipes as point-in-time snapshots with no ingestion charges. The embedded SQL console and external tools (Power BI, Copilot Studio, GitHub Copilot) can then query billions of rows with sub-second response times at high concurrency. Planned future features include write-back to OneLake, continuous sync, permission sync, and deeper Power BI integration.
Big improvements to Seer Agent
Seer Agent, Sentry’s AI debugging assistant, now performs write actions through the Sentry API, such as setting up custom dashboards, creating monitors and alerts, and assigning, resolving, and archiving issues. It defaults to the user’s permission level and explicitly requests permission for tasks outside its write scope, with permissions not persisting beyond a single chat. The agent can also incorporate infrastructure data from external monitoring platforms like Datadog and Google Cloud Platform via Seer Connectors, using API keys instead of OAuth. This allows Seer Agent to diagnose issues by connecting front-end and back-end data beyond Sentry’s native scope. These updates evolve Seer Agent from a conversational rubber duck into a proactive debugging assistant that can act on behalf of users.
86% of brands think their AI works. Only 51% of customers agree.
A Twilio research report identifies a 35-point satisfaction gap: 86% of brands believe their AI customer service satisfies customers, but only 51% of consumers agree. The discrepancy arises because companies measure metrics such as conversation completion surveys and containment rates, which systematically exclude abandoned or unresolved interactions. 71% of consumers abandon a conversation when the AI fails to recognize them, yet those exits rarely appear in satisfaction data. 49% of consumers say AI never resolved their issue. To close the gap, Twilio recommends tracking abandonment by conversation stage, resolution of the underlying issue across sessions, repeat contact within seven days, containment split by outcome, and escalation quality.
The Intelligent Data Platform for the AI Era
MongoDB announced the largest platform expansion yet, including MongoDB 9.0, Atlas Infinite, and Atlas Agent Engine. MongoDB 9.0 delivers up to 2x higher throughput, 35% faster reads, and 30% faster updates over version 8.0. Atlas Infinite is a new deployment option that allows storage and compute to scale independently, with time to scale more than 96% faster and each shard holding 10x more storage. Atlas Agent Engine provides built-in memory, retrieval, identity, and governance for production agents, operating on the same data platform as operational data. The company argues that as AI moves from generating answers to taking actions, the data layer must provide live, fresh, and retrievable operational data rather than separate AI infrastructure.
Your Database Just Became Your AI Control Point
Agentic AI transforms the database from a passive record store into an active control point that serves real-time context and enforces policy at the moment an agent acts. According to IDC’s April 2026 survey of 715 organizations, 30% already run agentic AI in production and another 50% have established spending plans. Real-time data delivery is now baseline: 64.6% of enterprise use cases require real-time or near-real-time delivery, and 91.2% of streaming scenarios need end-to-end latency under 200 milliseconds. IDC identifies four requirements for agentic AI: a real-time operational foundation for unstructured data, elastic scale for bursty agent traffic, accurate always-current retrieval via integrated vector databases, and a governed runtime with durable memory and identity-bound policy. By 2029, IDC predicts 60% of enterprise data platforms will unify transactional and analytical workloads in hybrid processing architectures to power agentic AI.
Lakebase Search: State-of-the-art full text and vector search for Postgres
Traditional OLTP systems like Postgres face challenges meeting the low-latency, high-accuracy search demands of AI agents, often requiring standalone search engines with ETL pipelines. Lakebase introduces lakebase_vector and lakebase_text, two extensions that bring scalable approximate nearest neighbor search and BM25 full-text search directly into Lakebase Postgres. lakebase_vector overcomes pgvector’s memory-bound HNSW index by using hierarchical IVF clustering and binary quantization, achieving 71-millisecond P99 latency at 97 percent recall on the 100-million-vector LAION benchmark. The extensions decouple storage from compute for pay-per-use scaling, enable fast parallel index builds, and support native hybrid search that fuses vector similarity with keyword relevance in a single SQL query. Early adopters like Conexiom reduced infrastructure costs by three times while achieving five times higher throughput, and the service is generally available on AWS and Azure.
Engineering Multi-Agent AI Teams That Build and Test Themselves
Salesforce Marketing Cloud’s AI Delivery Lead Sohini Arya and team built Agent Designer, a system that automates the creation and testing of multi-agent AI teams. The orchestrator cannot write agent files itself; instead, seven specialist agents analyze architecture, failure modes, and compliance before the orchestrator synthesizes findings for human review. The system transforms an engineer’s request into a tested multi-agent team in 15-30 minutes, down from 2-4 hours manually, and helped approximately 200 engineers adopt the manager-of-agents model. Governance challenges included enforcing write-scope boundaries via prompts, reconciling independently produced design conflicts, and applying topology-specific budget formulas that account for flat, swarm, or team structures. Verifier agents return structured PASS/WARN/FAIL blocks, and self-healing is limited to two fix attempts and a $3 ceiling to prevent runaway repairs, with the orchestrator’s 500-turn budget divided across analysis, approval, generation, testing, and retries.
Introducing cf: the agentic CLI for the entire Cloudflare API
Cloudflare introduced cf, a new CLI that provides access to the entire Cloudflare API surface of over 3,000 operations, far exceeding Wrangler’s roughly 280 commands. Cf is designed primarily for AI coding agents, which account for nearly half of Wrangler usage, and defaults to JSON output with condensed formatting to minimize token consumption. The CLI includes a natural language search command, cf cli search, that helps agents discover the right operation without bloating context, and uses TypeScript-based configuration via cloudflare.config.ts for type safety. Cf is generated by Forge, Cloudflare’s internal pipeline that produces CLI commands from OpenAPI schema annotations, enabling consistent and comprehensive API coverage. The open beta is available globally via npm, and cf adopts Vite as its default local development server, with a migration path from Wrangler and maintenance support for 18 months after beta ends.
Air Teams: Bring Your Best Agentic Workflows to the Whole Team – and Automate Repeatable Work
JetBrains introduced Air Teams, a shared workspace for agentic development that extends beyond individual developers to teams and organizations. The system comprises four components: Automations for recurring tasks triggered by events or schedules, shared cloud environments with preconfigured tools and credentials, cloud tasks that run in parallel without tying up local machines, and team projects that centralize members, environments, and credits. Automations include templates for code review, bug fixes, and dependency updates, and they run under a project’s service account so they persist after a creator leaves. Shared environments are defined in a repository’s .air/cloud/startup.sh script and can be snapshotted for reuse, with secrets shared without exposing values. Air Teams is available to business customers, with plans to expand to individual customers later.
The road to the agentic browser: A Kitesurf update
Kitesurf, Cloudflare’s browser for AI agents built entirely on Workers, has received a major update. It now supports WebMCP, enabling agents to call site functions like searchFlights() instead of simulating clicks, and site owners can enable WebMCP with one switch. The browser passes over 730,000 Web Platform Tests, up 500,000 since launch, and adds support for CSS Layout, CSSOM, custom elements, URL-based module resolution, and improved iframe behavior. Efficiency optimizations reduce DOM traversal overhead, font fetching, and timer work, keeping CPU usage in line with launch benchmarks despite expanded capabilities. Kitesurf also gains full Browser Run API coverage, including Quick Actions from Worker bindings, and a terminal rendering mode using the Kitty protocol or ANSI text, with open-sourcing planned soon.
Holo4: powering generalist computer-use agents
Holo4 is a new series of agentic models designed to interact with software through any available interface: GUIs, code, MCP, and APIs. It comes in two sizes, a 27B dense model and a 35B-A3B Mixture of Experts model, both available on the H Models API. The models were trained through supervised and reinforcement learning on environments and tasks generated by the Agentic Task Factory, which produces verifiable tasks from documentation alone. On the OSWorld 2.0 benchmark, Holo4 27B scores 61.7% against 81.8% for Opus 5.5, achieving competitive performance at a fraction of the cost. The company also released Holotron4 Nano, an updated version of Holotron 3 built on the Nemotron 3 Nano Omni model using the same post-training stack.
NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring
The post compares early internet security—where browser sandboxing isolated pages—to the current need for trust layers in AI agents, citing incidents where agents escaped evaluation environments and misreported actions. NVIDIA introduces OpenShell, an open-source secure runtime that sandboxes each agent with kernel-level isolation and enforces verifiable policies defined by operators before and during execution. The platform follows five core principles, including out-of-band enforcement, controlling the path to the model as a kill switch, and scaling agent authority with reasoning visibility, to prevent agent drift. For hardware-level protection, NVIDIA Sentry extends monitoring into BlueField DPUs, which sit on the node’s only path to the model, providing continuous out-of-band observability and real-time policy enforcement at line speed. The architecture is optimized for NVIDIA Vera CPU and BlueField DPU systems, enabling fleet-wide agent security with full lineage tracking, and for existing Vera systems, enabling protections is just a software update.
Add Runtime Controls to AI Agents with NVIDIA OpenShell
AI agents with broad access to systems and data can cause failures like modifying production data or exposing secrets. NVIDIA OpenShell 0.1.0 is an open-source runtime that enforces permissions outside the agent workload using sandboxed execution, controlled service access, credential management, and formal policy analysis. It combines a Gateway for lifecycle management, a Supervisor that checks outbound requests against policy, and a Sandbox with kernel-level filesystem and process controls and no network path except through the supervisor. The supervisor inspects HTTP, GraphQL, and MCP traffic to allow reads while blocking writes on the same API, and real credentials are kept outside the agent workload and bound to authorized endpoints. OpenShell supports multi-tenant platforms, formal policy verification that can prove permissions remain within boundaries, and policy advisor for dynamic network changes, and is adopted by Cadence, Slack, and Gecko Robotics for chip design, enterprise automation, and physical AI.
AI-powered agents for contact center in Australia: containment quality, not deflection
The Commonwealth Bank episode in 2025 demonstrated that a deflection metric is not a resolution metric: the bank reversed 45 redundancies after call volumes rose despite a voice bot credited with diverting 2,000 calls weekly. By May 2026, on a rebuilt platform, the same automation program achieved 84.6% end-to-end resolution for self-service messaging interactions. A peer-reviewed 2024 study found Whisper’s match error rate significantly higher for Australian English than American English, an effect of roughly 1.3 percentage points, with non-native accents worst affected. Australian regulation as of September 2026 imposes no affirmative AI disclosure duty, but misleading conduct provisions apply and an automated decision-making transparency obligation commences December 2026. Twilio’s platform offers an Australian region with emergency calling, model-agnostic agent execution, and a memory layer that preserves context during handoffs, addressing the containment quality gap.
AI-powered agents for contact center in India: what they can and cannot do
AI voice agents in Indian contact centers conduct end-to-end customer conversations using speech recognition, a language model, and text-to-speech, differing from agent assist which only suggests actions. Indian deployments are past pilot: Meesho handles 60,000 calls daily with a 75% cost reduction, and Bajaj Finance reports 71% of DIY service volumes handled by AI at one-third the human cost. The four-stage loop—speech recognition, language model, speech synthesis, telephony—must complete before the pause feels wrong, making architecture more critical than model choice. The Voice of India benchmark shows no system achieves under 20% word error rate across all 15 languages, so engine choice must be per-language and swappable. The RBI FREE-AI committee recommends informing customers they are talking to AI, human handling of complaints, and a route to a person, though these are not yet binding.
AI-powered agents for contact center in Philippines: what the term actually means
In the Philippines, an AI-powered agent usually means a human agent assisted by AI for transcription, answer retrieval, live coaching, and wrap-up, not an autonomous software agent. The contact center sector employs 1.68 million people and earned $33.9 billion in 2025, with 56% of IBPAP members actively implementing AI. Published figures show a 22% reduction in call handling time from Concentrix and a 37% increase in agent satisfaction from TELUS Digital, but no benchmark measures Taglish speech. The IBPAP survey found 8% of operators reduced headcount, 13% increased it, and 24% saw job role shifts, while the IT-BPM sector added 70,000 net staff in 2025. Platforms like Twilio allow composable assist layers using models from OpenAI, Anthropic, or Azure, important where the language problem is unsolved and no Philippine rule requires disclosing AI use.
Related topics
All topics →Weekly digest
The best engineering reads, once a week.