---
title: "Agentic AI"
description: "26 posts about Agentic AI, summarised, each linking to the original."
---

# Agentic AI
> 26 posts about Agentic AI, summarised, each linking to the original.

## Articles

### [Building secure mobile apps in the agentic era (2026) - Shopify](https://yomu.fyi/post/building-secure-mobile-apps-in-the-agentic-era-2026-shopify.md)
- Company: [Shopify](https://yomu.fyi/company/shopify.md)
- Author: Michael Hughes
- Published: Nov 18, 2026
- Type: Explainer

Shopify’s mobile app team now ships 11x more code per week than a year ago, averaging 434,000 lines changed per week during a 12-week rewrite. To maintain security at this pace, they built a system that maintains a map of each app’s attack surface, organized into surfaces such as deep links, WebViews, and stored data. Each surface has a skill derived from OWASP MASVS/MASTG and lessons from past incidents, which is used to inventory and review one item at a time. A daily pipeline compares the latest inventory with the last accepted map, reviews each changed item with read-only code access, and publishes a new map and changes page. Agents also monitor security news for industry-wide vulnerabilities and check them against the map and code, posting briefings only when the app is affected.


### [Kotlin Toolchain 0.14: Agentic Development and iOS Publishing](https://yomu.fyi/post/kotlin-toolchain-0-14-agentic-development-and-ios-publishing.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Kirill Likhodedov
- Published: Nov 17, 2026
- Type: Announcement

Kotlin Toolchain 0.14 introduces a kotlin publish testflight command that guides users through first-time iOS app publishing to TestFlight and supports automated updates afterward. The CLI wizard now generates MCP server configuration and an AGENTS.md file to improve agentic development workflows. A new Kotlin agent skills collection reduces token consumption in Toolchain-based projects, as shown by internal AI evals. The release adds native Java 8+ API desugaring for Android via a single desugaring: enabled line in module.yaml. Dependency resolution is optimized, making first project imports over 3x faster by reducing unnecessary repository requests and increasing parallel downloads.


### [Agent harness engineering: Separating model capabilities from enforcement](https://yomu.fyi/post/agent-harness-engineering-separating-model-capabilities-from-enforceme.md)
- Company: [Red Hat](https://yomu.fyi/company/red-hat.md)
- Author: Mark Vaykhansky, Uri Shaket
- Published: Nov 12, 2026
- Type: Explainer

A breach at Hugging Face showed that capable agents can chain vulnerabilities across trust boundaries when the surrounding harness gives them paths around security controls. The post argues that capabilities models can reliably learn should be absorbed into the model, while enforcement properties such as permissions, identity, isolation, and budgets must be hardened as infrastructure outside the model. It classifies harnesses into four types—workflow, coding, research, and agent runtime—each with different architectures for different jobs. The better the model gets, the more dangerous a weak harness becomes, because teams often remove enforcement scaffolding as the model absorbs planning and tool use. The post recommends diagnosing gaps in an existing agent’s harness before applying tooling such as a gateway layer, sandboxed containers, or distributed tracing.


### [Databricks Doubles Down on AI Products, Opens New 235,000-Square-Foot San Francisco Headquarters](https://yomu.fyi/post/databricks-doubles-down-on-ai-products-opens-new-235-000-square-foot-s.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Patrick Wendell, Matei Zaharia
- Published: Nov 11, 2026
- Type: Announcement

Databricks opened its new 235,000-square-foot headquarters at One Sansome Street in San Francisco, a space that matches the scale of its AI ambitions. The company is investing over $1 billion into its San Francisco operations by early 2028 and has committed to hosting Data + AI Summit there through 2030. Its AI Research team, based in San Francisco, develops KARL custom models, MemCurrent persistent memory for agents, and Adaptive Instructed Retriever for efficient search. Products like Lakebase, a serverless Postgres database for AI agents, and Genie, an autonomous enterprise agent, are built at the new headquarters. Databricks has grown to over 12,000 employees worldwide, serves over 20,000 customers including 70% of the Fortune 500, and achieved a $7 billion revenue run-rate.


### [BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model](https://yomu.fyi/post/botpsy-on-device-personal-assistant-with-a-sub-billion-parameter-agent.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Milan Gritta, Jihye Back, JinUk, Cho, ohjoon kwon, Patrik Lambert, Mathias Buus, Amril Nurman
- Published: Nov 11, 2026
- Type: Announcement

Tether AI Research introduces BotPsy, a 0.8B-parameter model designed for on-device function calling and agentic workflows, built on Qwen3.5-0.8B. It achieves state-of-the-art tool-use and agentic performance for its size, outperforming every model up to 1.2B parameters on five of six benchmarks, including beating Qwen3.5-2B-Think on BFCL v4, ACEBench, τ-bench, ToolSandbox, and BrowseComp-Plus. BotPsy roughly doubles the agentic capability of its base model, with BFCL v4 rising from 23.4 to 44.0 and τ-bench from 23.2 to 51.5. It demonstrates improved safety, with unsafe responses dropping from 36% to 21% in automated red-teaming, and more efficient reasoning, using about 3× fewer reasoning tokens on GPQA Diamond than its base model. The model is released with open weights under Apache 2.0 for research and educational use in bf16, Q8, and Q4 precisions.


### [A2UI 1.0: Build delightful, fast, reliable, agentic experiences](https://yomu.fyi/post/a2ui-1-0-build-delightful-fast-reliable-agentic-experiences.md)
- Company: [Google](https://yomu.fyi/company/google.md)
- Author: Vishal Motwani, Google A2UI Team
- Published: Nov 11, 2026
- Type: Announcement

A2UI 1.0 is the first stable release of an open protocol for generative UI, enabling AI agents to respond with interactive screens instead of text. The protocol includes bidirectional function calls, accessibility support, and extensions for metadata. Developers can use open-source renderer SDKs for web, Android, and iOS, along with agent SDKs for Python, TypeScript, and Dart. Companies such as Atlassian, Alation, Bloomberg, and ByteDance are already building with A2UI. The protocol supports fast, reliable UI generation through A2UI Express and system-one decision models, with server-side data binding to keep sensitive values secure.


### [MCP or CLI? Yes.](https://yomu.fyi/post/mcp-or-cli-yes.md)
- Company: [Sentry](https://yomu.fyi/company/sentry.md)
- Author: David Cramer
- Published: Nov 10, 2026
- Type: Announcement

Sentry ships a unified experience that bridges CLI and MCP, arguing both can coexist rather than one replacing the other. The CLI handles filesystem tasks like grepping raw event logs from session replays, which MCP cannot do efficiently without server-side pre-aggregation. MCP provides native tool discovery and steering that the CLI lacks, allowing agents to discover behavior without explicit instructions. The implementation shares authentication by launching the MCP server in stdio mode with a shim that passes CLI credentials through a different token header to bypass OAuth constraints. This approach works around current MCP limitations, including inconsistent support for deferred tool calling, instructions metadata, and events across different harnesses.


### [Transforming fleet analytics for Novus Hi-Tech through agentic AI](https://yomu.fyi/post/transforming-fleet-analytics-for-novus-hi-tech-through-agentic-ai.md)
- Company: [AWS](https://yomu.fyi/company/aws.md)
- Author: Sameer Shrivastava
- Published: Nov 9, 2026
- Type: Problem & solution

Novus Hi-Tech needed a system to act on fleet data from thousands of vehicles across 10 countries, not just display it. They collaborated with the AWS PACE team to build FleetGPT, a multi-agent system using Strands Agents SDK and AWS managed AI services. FleetGPT replaces fixed dashboards with a natural language interface for dynamic querying, investigation, and insight generation, and automates real-time safety interventions via AI-powered voice calls. The system achieved a 46 percent reduction in road-behavior violations and a 55 percent drop in drowsiness alerts per 1,000 km. Warning validation time dropped from minutes to seconds, and fleet managers gained self-service analytics without engaging the data team.


### [Using AI to Build AI: How Agentforce Engineers Scale Automation Without Sacrificing Reliability](https://yomu.fyi/post/using-ai-to-build-ai-how-agentforce-engineers-scale-automation-without.md)
- Company: [Salesforce](https://yomu.fyi/company/salesforce.md)
- Author: Fereydoon Vafaei
- Published: Nov 9, 2026
- Type: Problem & solution

The Agentforce team at Salesforce faced the challenge of using AI to help build and maintain the Atlas Reasoning Engine without compromising reliability. The central risk was that AI could diagnose problems, modify the engine, test fixes, and review its own work while being confidently wrong. To address this, the team built an integrated engineering knowledge base using semantic retrieval to connect fragmented knowledge across incidents, fixes, and discussions. They implemented layered AI evaluation with an agent-as-a-judge model and human oversight, using different models to review work from different perspectives to avoid self-preference bias. As a result, typical bug-resolution time dropped from two to four weeks to one or two days, and system availability reached 100% in weekly reports during the observed period.


### [ClickHouse Managed Postgres goes GA and Xata acquisition brings agentic capabilities to the service](https://yomu.fyi/post/clickhouse-managed-postgres-goes-ga-and-xata-acquisition-brings-agenti.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Nov 9, 2026
- Type: Announcement

ClickHouse announced the general availability of ClickHouse Managed Postgres, a high-performance, enterprise-grade Postgres service backed by local NVMe storage. The service delivers 2–10× performance improvements for OLTP workloads and supports hundreds of production customers, including AI-native startups and enterprises. Native CDC through WalShadow and ClickPipes, along with the pg\_clickhouse extension, unify transactions and analytics between Postgres and ClickHouse. The company also acquired Xata to add instant branching, cloning, automatic idling, and scale-to-zero capabilities powered by Xatastor’s copy-on-write storage. These agentic features will be integrated into Managed Postgres to support development workflows, agent sandboxing, and agent memory within a single platform.


### [Agent Swarms on Sprites](https://yomu.fyi/post/agent-swarms-on-sprites.md)
- Company: [Fly.io](https://yomu.fyi/company/fly-io.md)
- Published: Nov 6, 2026
- Type: Explainer

Running multiple AI agents in a swarm on a traditional single machine creates conflicts over dependencies, file systems, and network access. Sprites addresses this by giving each agent its own fully isolated virtual machine with independent filesystems and network policies. Checkpointing allows any agent’s environment to be rolled back to an earlier state in seconds, making speculative execution practical. Each Sprite has configurable network policies that deny access to unauthorized domains at the DNS level, preventing data leaks. Environments pause when idle and resume on demand, so the entire swarm survives closing the laptop; the orchestrator creates workers, assigns tasks, and collects results through HTTP endpoints, shared databases, or direct file reads from each worker’s console.


### [Closing the Loop: The Missing Ingredient in Agentic Development](https://yomu.fyi/post/closing-the-loop-the-missing-ingredient-in-agentic-development.md)
- Company: [Fly.io](https://yomu.fyi/company/fly-io.md)
- Published: Nov 6, 2026
- Type: Problem & solution

AI code agents that require constant permission prompts create a new form of friction similar to old copy-paste workflows. The Sprite environment solves this by giving agents unrestricted access to a fully isolated computer where they can install dependencies, start servers, and test endpoints without human intervention. Preinstalled tools such as test runners, service logs at known paths, and checkpoint-based rollback enable agents to write code, run it, observe failures, and fix issues autonomously. Additional tools like curl, jq, sqlite3, and network debugging utilities support API testing, database inspection, and network troubleshooting. The result is a tight closed loop where the agent can build, test, and ship without prompting the developer, turning the developer into the idea person rather than middleware.


### [Make Computers Make Computers](https://yomu.fyi/post/make-computers-make-computers.md)
- Company: [Fly.io](https://yomu.fyi/company/fly-io.md)
- Published: Nov 6, 2026
- Type: Explainer

Sprites are isolated virtual computers that can create and manage other Sprites via an API, and Claude can use that API when authorized by a token from sprites.dev/account. This enables automated workflows like central OAuth token management, where an Auth Sprite handles token refresh and storage for other Sprites. For CI, an orchestrator Sprite spins up a worker for each test file, locks its network to only pypi and github, runs the test, collects results, and destroys the worker. Preview environments are created per pull request, with each PR getting its own Sprite that is redeployed on updates and destroyed on merge. Additional use cases include agent delegation with network-policy-locked workers, running untrusted code in a deny-all network Sprite, and team provisioning via a provisioner Sprite that creates and configures new Sprites with checkpoints enabling fast restoration.


### [How TUI scales threat modeling for 100+ teams with Amazon Bedrock](https://yomu.fyi/post/how-tui-scales-threat-modeling-for-100-teams-with-amazon-bedrock.md)
- Company: [AWS](https://yomu.fyi/company/aws.md)
- Author: Vushesh Babu Adhikari
- Published: Nov 5, 2026
- Type: Explainer

TUI’s four-person security team faced a bottleneck: manual threat modeling for 100+ digital product teams, each assessment taking 4-6 hours. They built a system using Amazon Bedrock with Claude Sonnet 4.5 and deployed on Amazon Bedrock AgentCore runtime after a 10-week proof of concept with AWS. Instead of loading all 192 security standards into prompts, they implemented metadata-driven selection, reducing context from 338 KB to 21 KB and improving consistency. The LangGraph-based agentic workflow uses self-correcting threat generation and SME comparison for accuracy and completeness. In production, assessment time dropped by up to 70%, sessions per month increased from 3-4 to 10-12, and 255 users registered in eight weeks, making threat modeling mandatory for all new services.


### [Five Ways NVIDIA Co-Design Maximizes the Value of Every Byte in the AI Factory](https://yomu.fyi/post/five-ways-nvidia-co-design-maximizes-the-value-of-every-byte-in-the-ai.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Shara Tibken
- Published: Nov 5, 2026
- Type: Explainer

The post explains how NVIDIA co-designs GPU architecture, interconnect, serving software, CPU, and storage to maximize useful work per byte of HBM in an AI factory. NVFP4 4-bit quantization reduces weight and KV cache footprint by up to 3.8x relative to FP16, freeing memory for larger batches and higher concurrency. Wide expert parallelism (Wide-EP) spreads MoE experts across many GPUs via NVLink, reducing redundant weight copies and increasing arithmetic intensity. NVIDIA Dynamo treats KV cache as a fleet-wide resource through KV-aware routing, disaggregated serving, and a tiered KV Cache Runner (KVCR) that preserves context across prefill/decode boundaries. The Vera CPU offloads KV-cache management during tool-call idle gaps, and CMX provides a pod-level flash tier for shared KV cache reuse, together delivering up to 30x throughput per megawatt on Vera Rubin NVL72 versus GB300 NVL72. The five levers combine to measure memory by useful tokens per byte rather than capacity alone.


### [DataGrip 2026.3: Agentic Development With JetBrains Air, Cross-Database Queries, DuckDB Support, and Major UX Improvements](https://yomu.fyi/post/datagrip-2026-3-agentic-development-with-jetbrains-air-cross-database.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Richie Mitish
- Published: Nov 5, 2026
- Type: Announcement

DataGrip 2026.3 includes JetBrains Air by default, enabling AI agents to run SQL with syntax highlighting and reference database objects using @dbObject notation, with results viewable in a full-featured data editor within the chat. A new Workspace data source based on DuckDB allows joining tables from different data sources and DBMSs using a special syntax, eliminating the need to copy data between systems. DuckDB receives complete support with a dedicated SQL dialect, code completion, introspection, and three connection types: default file, in-memory, and URL-only. The UI revamp adds a Database Output tool window with automatic tab following, updated default shortcuts (Cmd+2 for Agent Sessions, Cmd+4 for Files), and enhanced column management including pinning, auto-resize, and the Manage Columns popup. Additional improvements include MariaDB sequences and package support, PostgreSQL 19 compatibility, and better transaction control with manual mode commit/rollback button state for PostgreSQL, Greenplum, MySQL, MariaDB, and Oracle.


### [What’s New in PhpStorm 2026.3](https://yomu.fyi/post/what-s-new-in-phpstorm-2026-3.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Hanna Yakush
- Published: Nov 5, 2026
- Type: Announcement

PhpStorm 2026.3 introduces Air, a new workspace for agentic development that works with multiple coding agents such as Claude Agent and GitHub Copilot, using existing subscriptions without requiring a JetBrains AI subscription. It adds support for PHP 8.6 with inspections for deprecated constructors returning values and calls to mb\_ereg\* functions, plus quick-fixes to migrate code. Laravel development gains new inspections for unknown config keys, missing resource route methods, and unknown Livewire component methods, along with a Processes tab to manage dev server and queue workers. The terminal now runs on libghostty, improving support for keyboard shortcuts, emoji, and AI coding agents like Claude Code and Junie. Remote development features an integrated SSH connection flow and native OpenSSH support with certificate-based authentication, while cross-database queries are enabled through a DuckDB-based Workspace data source that runs in memory.


### [GoLand 2026.3 Is Now Available!](https://yomu.fyi/post/goland-2026-3-is-now-available.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Artem Pronichev
- Published: Nov 5, 2026
- Type: Announcement

GoLand 2026.3 introduces JetBrains Air, a central workspace to run coding agents from multiple providers like Claude and Codex, track multi-agent workflows, and review changes with line-specific feedback before shipping. It supports existing subscriptions for Claude, Gemini, GitHub Copilot, and others, allowing users to switch providers mid-session, and includes a free Junie Lite configuration for those without a subscription. A bundled profiler skill enables agents to capture performance profiles, identify resource-consuming call paths, and highlight source lines for investigation, building on the Go optimization tools from the previous release. Other additions include Git worktrees that copy local project configuration, Figma design context via a plugin, remote development with SSH and Light mode, Kubernetes pod file browsing, and a redesigned Home window for quick tasks. Performance fixes address the most reported freezes, and optional sound cues for builds, breakpoints, and agent events improve accessibility and awareness.


### [CLion 2026.3 Is Here](https://yomu.fyi/post/clion-2026-3-is-here.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Oleg Zinovyev
- Published: Nov 5, 2026
- Type: Announcement

CLion 2026.3 introduces JetBrains Air, a new layer for agentic development that lets developers run, track, and verify multi-agent workflows inside the IDE. The release adds agentic debugging of hard faults on ARM Cortex-M chips, where an agent inspects fault registers, memory, and disassembly to trace the root cause. Agent-assisted CMake project configuration is now available, with a notification button that opens a terminal agent to diagnose and fix load failures. The IDE also gains support for NVIDIA CUDA Tile C++ and ISPC, a Class Layout tool window for visualizing memory layout, and automatic compiler header retrieval for code insight without a build system. Debug profiles are now stored as JSON, and a dedicated TRACE32 debug profile is included, along with RTOS objects for AI agents and improved Git worktree support.


### [Why Actionable Banking Is the Critical Stepping Stone to Full Autonomy](https://yomu.fyi/post/why-actionable-banking-is-the-critical-stepping-stone-to-full-autonomy.md)
- Company: [MongoDB](https://yomu.fyi/company/mongodb.md)
- Published: Nov 4, 2026
- Type: Explainer

Banking customers often lack immediate feedback on transactions, but actionable banking powered by agentic AI provides real-time, proactive advice on financial steps. A Personetics survey of 2,000 bank customers found that 70% strongly want their bank to proactively analyze finances and 84% would switch to a bank offering timely advice. Actionable banking serves as the stepping stone to autonomous banking, with agents acting on events while customers retain decision authority, and MongoDB Atlas enables this through real-time triggers, unified customer context, and an auditable record. The post outlines three agentic workflows — proactive investment on large deposits, proactive lending on overdue balances, and proactive added-value services on new mortgages — each ending in push and discuss rather than execution. Engineering challenges of fragmented context, agents that forget, and ungovernable actions are addressed by MongoDB’s unified document model, Atlas Agent Engine with built-in memory, and a single control plane for governance.


[Older posts](https://yomu.fyi/topic/agentic-ai/page/2.md)
