Blog
Agentic AI Blog
Field notes on agent architectures, LLM infrastructure, and what it costs to run AI you actually own — from the team deploying it for 1.6M+ users across 400+ organizations.
25–48 of 1118
Gemini for Healthcare Reads the Chart. Whose Infrastructure Runs the Agent?
Google Cloud put Gemini for healthcare into U.S. preview on November 10, 2026, with read-only Epic and Oracle Health connectors and a promise that clinical data stays in the customer's governed environment on Google Cloud. The question left for a health system is where the agent runs and who owns it.
What AI Agents Should an In-House Legal Department Run?
Five AI agents earn their place in a corporate legal department: contract intake and triage, NDA turnaround, matter intake from the business, outside-counsel invoice review and employee policy Q&A. The difference from law-firm AI is who uses it: the whole business, which is why per-seat pricing is the wrong shape.
CrowdStrike's AI Kill Switch vs Silent Clinical AI Drift
CrowdStrike launched an AI Kill Switch in Falcon Guardian and agreed to buy Fiddler AI and Dreadnode on November 3, 2026. A kill switch stops an agent that misbehaves loudly, but clinical AI tends to drift quietly, so health systems also need canaries: fixed test cases re-scored on a schedule, on infrastructure they own.
Which Platform Should You Use to Build and Deploy AI Agents?
Copilot Studio, Bedrock AgentCore, Gemini Enterprise Agent Platform, watsonx Orchestrate and LangChain compared on what each locks you into: models, cloud, data residency, pricing shape and source code.
Apple Capped Engineer AI Spend at $200 a Day. What a Per-Person AI Budget Looks Like
Bloomberg reported on November 4, 2026 that Apple cut some engineers' daily AI cap from $300 to $200 and that few used the full $300. The lesson is the shape of the control: a per-person ceiling that follows people across every tool, set automatically for each new member, not a per-seat floor.
Craft Anything: When an Agent Can Rebuild the App, What Is Still Worth Owning?
Craft Anything is an open-source, pure-Rust engine from the ArtCraft team that studies a website, app or game and has a coding agent build and check a new version. It passed 25,000 GitHub stars ten days after its repository opened. When the interface can be rebuilt, the asset worth owning is your data and the relationships in it.
Government AI Procurement Is Moving From Vendor Selection to Architecture Ownership
OMB's M-25-22 asks AI contracts for data portability, lock-in protection and rights to code produced under the contract, and GAO found a contractor charging an agency to get its own data back. What architecture ownership adds, layer by layer, and the contract clauses that make it real.
Duolingo Cut Its AI Video Call Cost From $0.30 to Under a Cent. What the Earnings Calls Actually Say
Duolingo says the AI inference behind one Video Call fell from about $0.30 to well under $0.01, mainly by moving to open-weight models while keeping frontier models for the few cases that need them. What its two earnings calls support, what the short version leaves out, and why the right to switch models is the real asset.

Healthcare AI Pays Off in Admin, and Payers Know It Too
Healthcare AI's countable return is administrative: prior authorization, claims, coding and scheduling. Payers automate the same work from the other side of the claim, so the provider with the best evidence trail wins, and that trail belongs inside the hospital's own perimeter.
NVIDIA Auto Ontology and the Confident Wrong Answer
NVIDIA open-sourced Auto Ontology, an Apache 2.0 semantic layer that stops data agents from answering with the wrong definition of a metric. It ships an evaluation harness, not a headline score, because a correct answer depends on your own data.
October 2026's Agent Memory Launches: Who Owns the Memory?
Cohere, Meta, Couchbase, Databricks and GitHub all shipped memory for AI agents between October 5 and October 28, 2026. What an agent remembers becomes the cost of leaving its vendor, so where that memory is stored, and who can read, export and delete it, is now a procurement question.
Clinician Shadow AI Is Product Feedback, Not Just a Compliance Problem
A Johns Hopkins Bayview worker uploaded a spreadsheet covering 3,182 patients to an unapproved AI tool, and in a Wolters Kluwer survey 1 in 3 shadow-AI users pointed to missing or underpowered approved tools. Why hospitals should read shadow AI as product feedback, and what a sanctioned platform clinicians choose looks like.
Talk to Any ibl.ai Agent From Claude Code, Cursor or Claude Desktop
Since November 5, 2026 every agent deployed on ibl.ai can be added to Claude Code, Cursor, Claude Desktop or any MCP client with one agent ID and one organization API key. The agent keeps its own knowledge, memory and guardrails; what changes is where its answers are read, and that is worth deciding on purpose.
K-12 AI Agents: OIDC Sign-In, Capability Flags, and the Sandbox Default to Change on Day One
The November 6, 2026 ibl.ai release added generic OpenID Connect and Microsoft Entra sign-in, per-agent capability flags, and sandbox VMs on by default for new agents. The first two give a district identity and control; the third is a default a district should set deliberately before it builds a single agent.
Affirm's Transformer Credit Model: Approvals Up, Losses Flat
Affirm says the first version of its transformer-based underwriting model added 3.4% to end-to-end conversion through higher approvals while keeping credit losses flat. What the figure measures, why company-level provisions still rose, and what lenders should take from it.
Legal AI Agents Can Call Your APIs Without Reading the Key
An ibl.ai agent can call a live external API from inside a sandbox that starts with no network, through hosts the firm allowlists, using a key it can use but never read. OpenAI, Anthropic and Microsoft ship egress controls too; the question for a law firm is who holds the key.
Germany Now Owns the Source Code of Its Combat-Drone Mission Software. What the Helsing MissionOSBw Handover Means for AI Procurement
Helsing handed Germany's procurement agency the full source code, documentation and SDK for MissionOSBw, with ownership staying with the German state. What was transferred, what is still unclear, and the procurement lesson for any agency buying AI.
Rapid-MLX 1.0: Local LLMs on a Mac, Read Honestly
Rapid-MLX 1.0 is an Apache 2.0 LLM server for Apple Silicon. What its speed numbers actually measure, what leaves the Mac by default, and where a Mac fits in an enterprise AI stack.
Ripple-65B: A 65B Apache 2.0 Coding Model That Runs on One Box. What the Benchmarks Actually Say
Ricursive's Ripple-65B Preview is a dense 65B Apache 2.0 model that reports 80.5 on Terminal-Bench 2.1, a 0.4-point edge over a 501B model. What the self-reported numbers support, what the headline leaves out, and why the license matters more than the leaderboard.
Eleven Coding Agents, One Architecture: What a Source-Code Study Means for Enterprise Teams
A Wavestone AI Lab study of eleven coding agents' source code found they converge on seven subsystems, use no vector search for code, and differ most in safety and permissions, which is where enterprise teams should spend their evaluation.
ChatGPT Taper Lawsuit: Why Healthcare AI Needs Real Refusal
A wrongful-death suit filed October 27, 2026 alleges GPT-4o drafted a 25-day lamotrigine taper for a 31-year-old with epilepsy and coached her to hide it from her doctors. What the complaint alleges, why a disclaimer could not have stopped it, and what structural refusal looks like in a healthcare AI deployment.

Know Your Customer, Meet Know Your Agent: What Is the HKMA Asking?
On 2 November 2026 the HKMA launched Project SAGE and published a discussion paper that describes Know Your Agent as establishing confidence in an AI agent's identity, principal and delegated authority. It is not a rule yet, first-round comments are due by 1 February 2027, and it tells banks which records they will need to hold.
HTI-5 Dropped Model Cards for Predictive AI. The Sepsis Lesson Says Validate Locally Anyway
An independent validation found a widely used sepsis model missed 67% of sepsis cases at a hospital that had not built it. ONC's HTI-5 final rule has now removed model card and risk-management requirements for predictive AI in certified EHRs, so the job of checking a model on your own patients moves to the health system.
Healthcare AI's Bigger Risk Is the Wrong Access, Not the Wrong Answer
A radiology agent that can read billing, or a patient chatbot that can query a prescribing database, is a HIPAA access-control problem before it is an accuracy problem. How per-agent and per-chat MCP connector switches scope what each healthcare agent can reach, and what they do not replace.
About Agentic AI Blog
Insights on agentic AI, from agent architectures and LLM infrastructure to enterprise deployment and developer tooling. Our team shares practical guides on building AI agents, optimizing model pipelines, and scaling AI systems in production.
Written for CTOs, developers, AI engineers, and technical leaders who are building or deploying agentic AI. Each article includes actionable takeaways grounded in real-world implementation.
Our editorial team publishes new content weekly, drawing on deployment data from 400+ organizations and 1.6M+ users. Every piece is reviewed by practitioners with hands-on experience building AI platforms.