AI agents

AI Weekly: Mistral Large 4 and the Decision Model Wave

Stop Calling Everything Made With AI Slop

How to Build a Golden-Question Benchmark for Your Data Agents

Build a golden-question benchmark for data agents: frozen Iceberg tags, execution-based scoring with Arrow, and CI gates....

When a Data Agent Gives a Wrong Number: Incident Response for Agentic Analytics

A plan for when a data agent gives a wrong number: severity levels, containment, Iceberg time travel reproduction, and blast radius queries....

Multi-Tenant Data Agents: Serving Many Customers From One Lakehouse

Serve many customers from one lakehouse without leaking between them: isolation models, Polaris grants, and tenant-scoped agent state....

Release Management for Data Agents: Shipping Prompts, Tools, Models, and Metrics Without Breaking Answers

Ship prompts, tools, models, and metric changes without breaking answers: agent manifests, shadow mode, canaries, and rollback....

Semantic Caching for Data Agents Without Serving Stale or Leaked Numbers

Cache data agent results safely with canonical query keys, Iceberg snapshot freshness checks, and permission-aware cache keys....

Small Language Models for Data Agents: Where They Win and Where They Fail

Where small language models win inside a data agent and where they fail, plus a routing architecture that escalates only when needed....

Sonnet, GPT Sol and a Shelved Astra — AI Weekly September 23 to 30, 2026

Turning an Analytics Question Into a Verified Agentic Graph

A complete AGS 1.0 graph for governed metric questions, with verification gates, deterministic check scripts, and reconciliation against a semantic la...

AI Weekly: Opus 5.5, GPT-6 Sol and Luna, and MCPA

Week of September 16 to 23, 2026 Continue reading on Medium »...

How Apache Ossie Is Deciding What Agents and BI Tools Can Ask a Semantic Layer

How Apache Ossie's layered query design gives AI agents both a constrained dimensional interface and a grain-safe SQL interface for semantic layers....

AI Weekly: DeepSeek Cuts Prices as Agents Go Hosted

AI Weekly: DeepSeek Cuts Prices as Agents Go Hosted

Week of September 10 to 17, 2026...

Agentic Data Architecture

A six-layer reference architecture for agents on company data: planners, tool boundaries, identity, the semantic layer, and what breaks when a layer i...

Context Engineering for Data Agents

Why text-to-SQL accuracy collapses on enterprise schemas, the five kinds of context an agent needs, where each one hides, and how to make the semantic...

Guardrails for AI on Company Data

The control surfaces that actually contain damage once an agent is fooled: identity, permissions, audit trails, and prompt injection at the query laye...

The Data Team of the Agentic Era: Generalists Owning End-to-End Workflows

The case for generalists owning end-to-end data workflows with agents, the counterargument, and how to make the transition work....

Open Standards for Agentic Harnesses

Every team that gets serious about AI agents hits the same wall, usually around month three....

Agent-Driven Storage Tiering for Apache Iceberg: Moving Cold Data Without Breaking Queries

A background agent can move cold Iceberg partitions to cheaper tiers without breaking live queries. Heatmaps, path-safe moves, and restore paths....

Securing the Agentic Lakehouse Gateway: Preventing Prompt Injection and Data Exfiltration

Agentic lakehouse gateways face prompt injection and exfiltration through query results. A threat model and defenses for the layer in front of data....

Metric Contracts in Code: Testing, Versioning, and Serving Business Logic to Multi-Agent Systems

Metric contracts in code let teams test, version, and serve business logic to multi-agent systems without each agent inventing its own SQL....

Query Routing at Machine Scale: Dynamic Workload Distribution Across Lakehouse Engines

Route each lakehouse query by shape, not by sender. Signals, rules, and how to keep dashboards, batch jobs, and agents from sharing one engine....

The Agent Is Now a Named Coworker, and It Needs a File Format

Named, persistent agents need a file format. Open Agent Profile, Buzz, Grok Bot, and Hermes Bot Mode show why a portable agent identity matters....

Your Agent Should Answer the Phone: A Field Guide to AI Gateways on Slack, Discord, Telegram, Signal, and Teams

A field guide to AI gateways on Slack, Discord, Telegram, Signal, and Teams: architecture, auth, cost, and the failure modes that matter....

Graphs in AI Engineering Have Solved Three Problems. The Fourth Is the Plan.

Knowledge graphs, GraphRAG, and LangGraph solved three problems. The fourth is the work itself: a reviewable graph of bounded agentic loops....

The Five Layers of an Agentic Lakehouse

The five layers of an agentic lakehouse: Storage, Catalog, Semantic, Gateway, and Agent Surface, and how one question travels through all of them....

Goal-Directed Data Quality Agents: Anomaly Quarantine on Apache Iceberg

Goal-directed data quality agents that watch Apache Iceberg tables, detect anomalies, and quarantine suspect data safely with snapshot isolation and b...

Managing the TCO of Agentic Analytics: Token Budgets, Query Throttles, and the Economics of Autonomy

Managing the total cost of agentic analytics: token budgets, query throttles, unit economics, and the FinOps discipline that keeps AI spend under cont...

Metric Contracts in 2026: Standardizing Business Logic Across Multi-Agent Frameworks

Metric contracts in 2026: versioned, testable definitions of business logic that let multi-agent frameworks compute revenue identically, with OSI inte...

Query Routing at Machine Scale: Multi-Engine Workload Distribution for the Agentic Lakehouse

Query routing at machine scale for the agentic lakehouse: engine selection, acceleration substitution, admission control, and placement....

Securing the Agentic Lakehouse Gateway: A Threat Model for Prompt Injection, Exfiltration, and the Firewall That Reads Sentences

A threat model for the agentic lakehouse gateway covering prompt injection, exfiltration, and the firewall that reads sentences....

Semantic Layer Federation: One Meaning for Data That Lives Everywhere

Build a federated semantic layer across multi-cloud data so one set of governed metric definitions serves BI tools, dashboards, and AI agents identica...

Building Stateless AI Tool Gateways with FastMCP, the 2026 MCP Spec, and Kubernetes

How to build stateless AI tool gateways with FastMCP, the 2026 MCP specification, and Kubernetes, and why statelessness finally makes MCP scale....

The Plan and the Worker: Two Open Specifications for Agent Harnesses

Two open specifications, the Agentic Graph Specification and the Open Agent Profile, turn agent plans and agent identity into portable, reviewable fil...

Budgeting for Agentic Analytics When Every Question Costs Something Different

Budgeting for agentic analytics when every question costs something different: token economics, query economics, instrumentation, and the cost control...

The Five Layers of an Agentic Lakehouse and Where the MCP Server Sits

The five layers of an agentic lakehouse and where the MCP server sits: storage, catalog, semantic layer, MCP gateway, and agent surface, plus identity...

Wiring Analytical Queries to Transactional APIs in Closed-Loop Decision Agents

Wiring analytical queries to transactional APIs in closed-loop decision agents: conditional writes, sagas with compensations, decision records, and bl...

Surviving Optimistic Commit Collisions When Hundreds of Agents Write to Iceberg

Surviving optimistic commit collisions when hundreds of agents write to Iceberg: which conflicts are real, commit buffers, partitioning, and the patte...

Why Agentic AI Needs a Governed Semantic Layer Behind the Model Context Protocol

Why agentic AI needs a governed semantic layer behind the Model Context Protocol: metric consistency, access control, Apache Ossie for portable....

Moving From Supply Chain Dashboards to Decision Loops With the Model Context Protocol

Moving from supply chain dashboards to decision loops with MCP: sense, decide, act, and verify, with typed action tools, idempotency keys, and graduat...

Metric Contracts as the Interface AI Agents Actually Need

Metric contracts as the interface AI agents need: calculation, inclusion rules, grain, temporal semantics, ownership, semantic versioning, and testing...

Guardrails for Analytics Agents That Do More Than Answer Questions

The risk isn't agents going rogue, it's agents acting correctly on bad input at machine speed....

Building Agent Telemetry Tables in Iceberg That Survive an Audit

A practical guide to building agent decision traces in Apache Iceberg that support audit reconstruction, governance review, and cost attribution....

What Agentic Analytics Actually Costs, and How to Keep It Bounded

Agent analytics generates two cost streams that scale on different variables. Here's the arithmetic, the levers that actually move the number, and how...

When the Query Optimizer Starts Managing Its Own Materializations

Autonomous materialized view management replaces quarterly review meetings with workload-driven scoring, and it's essential when AI agents generate....

Why AI Agents Fail on Raw Data, and What to Give Them Instead

Agents fail on raw lake data because business rules live in people's heads. Data products with semantic contracts fix this at the source....

The Five Layers Between Your Lakehouse and a Trustworthy Agent

Agent reliability is a property of the stack the model sits on. Five layers with distinct owners and failure modes turn the agent is unreliable....