The OpenNash Field Notes

Working notes from the frontier of agent deployment.

Work with OpenNash

Need an agent that still works after launch?

We build production AI agents with the evals, routing, tools, review loops, and runbooks that keep them useful in the real world.

hello@opennash.com Book a call ->
FIG 01 This week's feature November 9, 2026
Strategy 11 min read

Why AI Adoption Stalls: Redesign the Workflow Before You Change the Model

Every company can buy the same AI models, yet the value goes to a few. Census, BCG, MIT, and OpenAI data show why. Redesigns at Uber and Ramp show the pattern: checks after each step and people on exceptions.

Read the field note ->
FIG 02 Recently published 3 notes

Mistral Large 4 Benchmark on FrontierCode 2.0: How Far Open Weight Coding Models Trail on Mergeable Code

Mistral reports Large 4 at 19.7% on Cognition's FrontierCode 2.0. Claude Opus 5.5 scores 45.5%. Here is what each coding workload can use.

Read more ->

Who Signs for the AI's Work? Accountable Sign-Off for CPAs, Lawyers, Insurers, and Advisers

When AI drafts the audit workpaper, the brief, the claim decision, or the client letter, the licensed professional still signs and stays liable. What regulators and courts require, and a checklist for a review worth signing.

Read more ->

How Brain-Computer Interfaces Work: Neuralink, Synchron, Precision, and the Science Behind Them

How BCIs turn neural spikes into cursor moves and speech. Neuralink's N1 and R1 robot, Synchron, Precision, Paradromics, BrainGate, the data math, and the AI decoders.

Read more ->
FIG 03 The full index All posts

News

RSS feed
Engineering The $1,280 Root Exploit: Why CIOs Must Change in the Age of AI Strategy The Model Swap Test: What to Own, Rent, and Keep Replaceable in Your AI Stack Market Is AI Showing Up in Productivity and Jobs Data? A Cross-Check of the New Evidence Engineering AI Security Agents: What Binance, PHP2Shell, and Hugging Face Changed for Enterprise Defense Engineering Weight-Space Programming: LLMs Can Now Patch Model Weights Like Software Engineering AI Made Code Cheap. Verification Is Now the Cost of Legacy Modernization Engineering Always-On AI Agents: The Router and Worker Architecture Behind Muse, Dots, Duet, and DeepSeek Engineering Approval Fatigue: Why People Catch Only 1 in 7 Dangerous AI Agent Actions Engineering Background Coding Agents Need an Agent-First API Layer: What Ramp, Stripe, and Spotify Built Engineering The CI Bottleneck From AI Coding: Why Anthropic, Cursor, and GitHub Are Rebuilding the Pipeline Engineering Why Enterprise AI Stalls on Legacy Systems, and How OpenNash Modernizes Them Economics Anthropic IPO Valuation: Damodaran's $1.475 Trillion Model and What Matt Levine, Ben Thompson, and Other Top Analysts Say Market Google Cloud vs AWS vs Azure: Revenue, Run Rate, TPUs, and the Race to No. 1 Engineering How to Compare AI Code Reviewers: GitHub's ReviewBench Method Engineering Cross-Tenant Leaks in AI Infrastructure: Cloudflare's KV Cache Data Leak and Vercel's Sandbox Findings Engineering PHP2Shell Post-Mortem: What PHP, Cloudflare, Fastly, AWS, and Wiz Learned From the AI-Found PHP Bugs Engineering Verifying AI Generated Code: How Meta and Shopify Make Agents Prove Their Work Engineering The Bitter Lesson for AI Agents: What to Build Into a Harness and What to Plan to Delete Market Anthropic's Co-Founders: Who They Are, What They Believe, and the Case For and Against Economics Anthropic's IPO S-1 Explained: Revenue, Compute, Cash, and the Top Risks Engineering Formal Verification for AI Agents: How to Prove Agent Output Is Safe Engineering Reusable Prompts for Teams: How to Build a Prompt Library That Stays Current Engineering Small Model vs Large Model: When a Specialized AI Model Beats a Frontier Model Field guide Why Is My AI Assistant Getting Worse? How to Diagnose a Sudden AI Model Quality Drop Field guide What Is Prompt Caching? How It Works and When It Saves Money Field guide Should You Block AI Crawlers? How AI Search Finds Your Website and What to Allow Field guide Internal AI Workspace: Why Ramp Glass and Shopify River Exist Engineering MCP Search and Execute: How Code Mode Gives Agents Better Tools Field guide Jev 1.14 Update: What Changed, What's Experimental, and When to Use It Market Behind the Meter Generation for Data Centers: How AI Operators Are Buying Power Instead of Waiting for It Engineering Quantized Model Tool Calling Accuracy: Where FP8 and INT4 Break Agents Market RL Environments for LLM Training: What Buyers of Task-Specific Models Should Know Field guide Overseas AI Data Center Risk: The Checklist Enterprise Buyers Should Borrow Engineering Flexible Load Data Center Interconnection: Running Inference Fleets as Curtailable Compute Engineering LLM Nondeterminism at Temperature 0: Batch Effects That Move Your Eval Scores Engineering Long Context vs RAG Cost: Finding the Crossover Where Stuffing the Context Wins Engineering Two Agents, One Record: Idempotency Keys and Optimistic Locking for Agent API Writes Engineering Who Owns the Transcript: Server-Side Conversation State in LLM APIs vs. Client-Held Context Field guide Frontier Model Access Programs: Planning Your Roadmap Around Cyber Capability Thresholds Field guide Know Your Token Mix: Measuring Where Enterprise Inference Spend Goes Engineering Harness A/B Testing for Agents: Prove the Scaffold Change Helped, Not the Model Engineering Fault Injection for AI Agents: Chaos Testing Model APIs, Tools, and Retrieval Engineering The Agent Kill Switch: Containment Engineering for Agent Fleets Field guide AI Agent Social Engineering Defense: What UK Government Tests Change About Identity Engineering Coding Agents in Continuous Integration: Cost Controls, Flake Budgets, and What to Never Automate Field guide Thinking Fast and Slow for AI: System 1 vs System 2 and TypeSafe's Jev, Explained Field guide Continuous Agent Verification Is Becoming a Buying Requirement, Not a Nice-to-Have Field guide Direct-to-Chip or Nothing: The Liquid Cooling Retrofit Decision for AI Racks Field guide Your Tools Are Blowing the Context Window: Budgeting Tool Output Before It Reaches the Model Market If the Model Hub Changes Hands: Mapping Your Hugging Face Dependency Before an Acquisition Does It for You Engineering Where Agent Code Actually Runs: Choosing a Sandbox for AI Agent Code Execution Engineering Agent Instruction Files Are Source Code: Versioning AGENTS.md and CLAUDE.md Like Config Field guide Thinking Budgets Are a Product Decision: Tuning Reasoning Effort per Task, Not per Model Engineering Durable Agents: Checkpointing, Resumption, and Why Long Tasks Need a Workflow Engine Field guide Agent Simulation Testing: Put a Simulated User in Front of Your Agent Before Real Traffic Does Research Agent Memory Benchmarks: What They Actually Measure and Where They Mislead Field guide Output Contracts: Schema Validation Is the Cheapest Agent Reliability Layer Engineering Agent Skills Over Prompts: Package Procedures So Agents Load Only What the Task Needs Field guide Your Coding Agent Is an Unmanaged Endpoint: Runtime Security Controls for Copilot, Claude Code, and Codex Engineering Agentic RAG (Part 6): Permission Filters, Checked Citations, and Abstention Engineering AI Agent Evaluation and Observability (Part 7): Traces, Graders, and Release Gates Engineering AI Agent Architecture (Part 1): Where Model Judgment Ends and Code Begins Engineering How an AI Agent Harness Works (Part 2): Run Loop, Event Log, and CLI Engineering AI Agent Tools and Function Calling (Part 3): Tool Contracts and Safe Writes Engineering Context Engineering and Agent Memory (Part 5): What the Model Sees at Each Step Engineering MCP for AI Agents (Part 4): When a Tool Protocol Earns Its Cost Engineering AI Agent Security and Deployment (Part 8): Trust Boundaries, Approvals, and Staged Rollout Market Rent or Own: GPU-as-a-Service Pricing and Contract Economics in 2026 Engineering Model Deprecation Migration: Build the Runbook Before the Retirement Email Arrives Engineering Choosing an Inference Server: vLLM vs SGLang vs TensorRT-LLM Under Agentic Load Field guide Open-Weight Model License Terms Are a Procurement Problem, Not a Footnote Engineering Subagent Design Boundaries: Context Isolation, Not an Org Chart Field guide AI for Science Data Infrastructure: What the NSF's $83M Award Signals Market Building the AI Investment Business Case for the Board: Capex, Opex, and Depreciating Models Engineering Context Compaction for Agents: Keeping Long-Horizon Sessions Coherent Past the Window Field guide Autonomous Attack Agents Are Real: What the Anthropic Espionage Disclosure Changes for Defenders Engineering Agent Identity Management: Why AI Agents Should Never Hold Keys Field guide EU AI Act Enforcement Is Live: What August 2026 Changes for Providers and Deployers Engineering Reviewing AI Generated Code at Scale: Redesigning the Merge Path Engineering Prefill, Decode, Speculate: The Three Levers of Disaggregated Inference Serving Field guide Where Enterprise AI Delivers Real ROI: Six Domains That Work Engineering Measuring AI Coding Assistant Productivity: Throughput, Revert Rate, and the Metrics That Do Not Lie Field guide AI Data Center Power Constraints: Capacity Planning When the Grid Sets the Ceiling Field guide Leaderboard Gaps Smaller Than Error Bars: How to Compare Reasoning Models Honestly Engineering Agent Run Variance: Why pass@k Hides the Failures That Matter Field guide AI Voice Agent Cost Per Minute: What $0.05 to $1.00 Actually Buys in 2026 Field guide Freshdesk AI Pricing in 2026: The Real Cost of Copilot, Sessions, and Per-Agent Seats Field guide Custom AI Agent Development Cost: What a Production Deployment Really Costs Field guide Best Sierra AI Alternatives for Mid-Market Companies: Cost, Control, and Time to Value Engineering AI Materials Discovery Works When the Lab Closes the Loop Engineering DeepSeek's Lessons: How Efficient AI Teams Turn Constraints Into Systems Market Should Intel Second-Source Nvidia? A Practical Test of the Proposal Market Japan AI Strategy: How to Turn Strong Research Into Exportable Products Market The Lisp Machine Boom and Bust: Five Lessons for AI Platform Buyers Engineering Lumentum and AI Data Center Lasers: The Economics Behind Co-Packaged Optics Market The Nvidia Way: An Operating System for Fast, Technical AI Teams Engineering Superconducting AI Data Centers: What Fits in the Shoebox and What Does Not Market TSMC and the AI Chip Shortage: A Better Capacity Planning Model Field guide AI Agents for Private Equity Portfolio Companies: A 100-Day Value Creation Plan Engineering AI Agent Guardrails: Why System Prompts Are Not Policy Enforcement Engineering AI Agent Testing: Build a Sandbox Before Production Engineering AI Agent Tool Selection: Why More MCP Tools Can Make Agents Worse Engineering Why AI Data Agents Still Fail in BigQuery, dbt, and Airbyte Field guide AI Workflow Automation Should Start With Process Discovery Engineering Prompt Injection in AI Agents: An Enterprise Security Playbook Engineering RAG Evaluation: Diagnose Retrieval and Generation Failures Separately Engineering RAG Poisoning: How to Secure an Enterprise Knowledge Base Engineering How to Reduce AI Agent Cost With Plan Caching Engineering SFT vs. RL: Which Fine-Tuning Method Should Enterprises Use? Field guide What Can AI Agents Actually Automate in the Enterprise? Field guide AI Agent CRM Integration: A Production Guide for Zendesk, HubSpot, Salesforce, and NetSuite Field guide When Should AI Support Escalate to a Human? A Decision Framework for Deflection vs Escalation Field guide Cross-Channel Customer Support Memory: Why Your AI Forgets Customers Between Chat, Email, and Voice Field guide An Operating Model for Enterprise AI: From One Workflow to a Repeatable System Field guide How to Automate Tier 1 Support: A Practical Playbook Beyond FAQ Bots Field guide When to Scope the Next AI Workflow: Stability Gates Before Expansion Field guide The Weekly Edge-Case Review: How Good AI Workflows Improve Without Drifting Field guide Train the Operators and Reviewers: The Human Rollout Plan for AI Workflows Engineering AI Agent Writeback Patterns: Updating CRM, Helpdesk, ERP, Docs, Email, and Slack Safely Engineering Retrieval Paths and Source Priority: How AI Agents Decide Which Context to Trust Field guide Duplicate Data Breaks AI Agents: How to Clean Fields, Records, and Source Truth for AI Workflows Engineering Building a BAA-Ready AI Agent Stack: Architecture, Vendors, and the PHI Boundary Engineering Enterprise AI Strategy After Model Performance Parity: Why Routing Beats Picking a Winner Engineering Agentic AI Is Making Memory a Serving Bottleneck Field guide Migrating From Zendesk AI to a Custom Agent: Cost, Risk, and a 90-Day Cutover Plan Engineering Your Agent Failed in Prod. Can You Replay It? Field guide Domain-Specific AI Agents Beat Generic Chatbots for Real Work Field guide AI Agents for Financial Compliance: Automating Multi-Document Review Field guide The Production AI Playbook: What Enterprise Agents Need Before Launch Engineering Production Evals for AI Agents: Measuring Workflow Behavior Field guide Spec-Driven AI Workflows: Why Pilots Fail Before the Agent Is Built Engineering Agent Tokenomics: How to Cut AI Cost Without Breaking Quality Engineering Harbor Framework: Agent Evals for Real Business Work Field guide What Is Agentic Knowledge? A Practical Definition for AI Agents and Enterprise Knowledge Work Engineering Best Libraries for Caching in an Agent Knowledge Base: Redis, GPTCache, and LangChain Compared Field guide The AI Knowledge Retrieval Platform Buyer Framework: Consolidating Wikis Without Migrating the Mess Field guide Self-Hosted AI Customer Support: When On-Prem Beats Cloud (and Data Stays Yours) Field guide Per-Resolution vs Flat-Fee Chatbot Pricing: Which Model Wins as You Scale Field guide How to Automate Tier 2 Support Tickets Without Breaking Escalation Field guide Tier 1 vs Tier 2 Customer Support: Definitions, Examples, and Where AI Fits Each Tier Field guide Patching AI Agent Failure Modes: Prompt, Tool, Retrieval, Routing, or Workflow? AI architecture The Four Ways to Scale AI Compute: Pretraining, Post-Training, Test-Time, and Sleep-Time Field guide Managed AI Workflow Operations: What Actually Happens After the Agent Goes Live Economics The AI Economy Is Real. Now It Has to Become Workflow ROI Field guide Audit Logs and Dashboards for AI Workflows: What Operators Need After Launch Field guide AI Customer Support in 2026: From Tier 1 Bots to Tier 3 Automation Field guide Launch AI Workflows Inside Existing Systems, Not a Separate Chatbot Field guide Every Production Failure Should Become an Eval Case: The AI Agent Feedback Loop That Compounds Engineering Loops and Evals: How You Know AI Agents Work in 2026 Field guide Measure the Outcome, Not the Prompt: Business Metrics for Governed AI Agent Workflows Field guide AI Workflow Pass/Fail Criteria: The Production Readiness Checklist Engineering AI Agent Traces and Tool Calls: What to Capture So Workflow Failures Become Fixable Engineering Reliable Agentic RAG Architecture: Lessons From Bayer's PRINCE System Economics Tokenomics After GLM-5.2: When AI Model Choice Becomes Unit Economics Field guide Workflow-Specific Evals: Why Generic AI Benchmarks Don't Prove Your Agent Works Field guide The AI Compute Buildout: Players, Dependencies, and Where We Are Field guide AI Workflow Approval Gates and Exception Routing: Stopping Silent Failure Field guide Keep Deterministic Steps Deterministic: When to Use Rules, Code, and LLMs in One Workflow Field guide AI Agent Allowed Actions: The Governance Document to Write Before Launch Field guide Permission-Aware AI Agents: Controlling Who Sees What Data Field guide Data and Context Engineering for AI Agents: The Work Before the Workflow Works Field guide The AI Workflow Readiness Scorecard: Where AI Can Help Safely Field guide The AI Workflow Baseline: Measure Cycle Time, Volume, Exceptions, Cost, and Quality Before You Build Field guide Find the System of Record: The First Rule of Reliable AI Workflow Design Field guide What AI Will Automate Next: Closed-Loop vs. Open-Loop Work Field guide Workflow Discovery for AI Automation: Map the Process Before You Build the Agent Security Mythos/Fable 5 ExploitBench: From Crash to Code Execution Evals Mythos/Fable 5 Evals: Awareness and Sandbagging AI Safety Mythos/Fable 5 Bio Risk: Why Anthropic Stops Short of CB-2 AI Safety Claude Fable 5 and Mythos 5: Same Weights, Different Safeguards Field guide Decagon Pricing, Deployment, and Ownership: What Buyers Should Check AI Safety Mythos/Fable 5 NLA: What Anthropic Found Inside Engineering Recursive Self-Improvement Needs a Testing Harness, Not Just a Smarter Agent Field guide AI Customer Service for Financial Services: Audit Trails, SOC 2, and What Regulators Will Actually Ask Engineering What Is AG-UI? The Protocol Layer Agentic Apps Were Missing Compare Agentforce vs Sierra vs Decagon vs OpenNash: AI Agent Fit Compared Services AI Agent Consulting Services: Build Production Agents You Own Pricing AI Agent Pricing 2026: Platforms, Retainers, and True Cost Field guide HIPAA-Compliant AI Customer Service: What Healthcare Buyers Actually Need to Verify Field guide Why You Need an AI Services Partner: AGI Ships, But It Does Not Self-Install Field guide Workflow Discovery for AI Automation: Map the Process Before You Build the Agent Field guide Decagon Alternatives in 2026: 7 Options Compared on Pricing, Deployment Time, and Ownership Field guide AI in Insurance: Where Deterministic Automation Saves Time First Field guide How to Eval AI Agents in 2026: From Pretty Demos to Production Evidence Engineering Macro Evals for Agentic Systems: Why One Failed Trace Is Not Enough Market Replacing Junior Jobs With AI Is a Bad Business Strategy Engineering Self-Improving AI Agents Are Not Magic: What OpenAI's Tax AI and Karpathy's Autoresearch Teach Builders Engineering What the Anthropic Cookbooks Reveal About Building Agents That Actually Work Field guide AI Agent Deployment Is Not the Finish Line: The Operating Model After Launch Market Model-Agnostic AI Strategy: Why Model Routing, Traces, and Memory Beat Vendor Lock-In Market Verifiable Outcomes: Why the Best AI Agents Start With Boring Proof Field guide Open Source Just Caught Up to Opus 4.5: Why the Moat Is the Harness, Not the Model Market AI Receptionist for Law Firms: Intake, Conflicts, Escalation, and Audit Trails Market Best AI Voice Agents for Customer Support: Retell, Bland, Synthflow, Sierra, and Custom Compared Market Zendesk AI Agent Pricing: Licenses, Setup, and Resolution Fees Market Sierra AI Pricing: What Outcome-Based Really Costs and When to Walk Away Engineering AI Coding Harnesses: Stop Writing Code, Start Designing the System Around the Agent Engineering Notion AI Agent Architecture: 5 Rebuilds, 100+ Tools, and What Finally Worked Market Why You Need a Professional Services Partner to Deploy AI Agents (And What That Actually Looks Like) Market Build vs Buy AI Customer Support: The 3-Year Cost Sierra, Ada, and Salesforce Hope You Skip Market Intercom Fin Pricing Reality: When $0.99 Per Resolution Becomes a $1.2M Problem Market Ada AI Review 2026: Pricing Reality, Platform Limits, and When Custom Wins Market Salesforce Agentforce vs. Custom AI Agents: When the CRM Giant Isn't the Right Fit Market Why 'Configurable' Isn't 'Custom': What Businesses Actually Get from AI Agent Platforms vs. Purpose-Built Software Market AI Agent Audit Trails: What Regulated Industries Need Before Deploying Agents Engineering LLM Routing in Production: How to Pick the Right Model for Each Task Engineering Agent Reliability Engineering: SRE Patterns That Keep AI Agents Running in Production Market AI Agent UX: How to Design Interfaces That Users Actually Trust Field guide Human-in-the-Loop Agent Design: 5 Patterns for When AI Should Ask Permission Market Your Team's First AI Agent: 5 High-ROI Starting Points That Are Not Chatbots Market Build vs. Buy AI Automation in 2026: A Decision Framework for Technical Leaders Engineering Agent Memory Beyond RAG: Short-Term, Long-Term, and Episodic Memory Patterns That Work Engineering Evaluation-Driven Development: How to Ship AI Features That Actually Improve Over Time Market AI Labor Market Impact 2026: What to Automate Now and What Still Needs Humans Engineering The 8 Levels of Agentic Engineering: How Teams Move from Copilot to Background Agents Field guide AI Automation, SaaS Stock Drops, and the Future of Apps: What Actually Gets Repriced Field guide Why AI Agents Fail in Production: 7 Failure Modes and How to Prevent Them Engineering 9 Agentic Workflow Patterns Ranked: Which Ones Actually Work in Production Market Under $50 a Day: Cost Engineering for Production AI Workflows Market OpenClaw for Business: What It Is, How It Works, and When to Deploy It Podcast AI's Real Bottleneck Isn't Software. It's Materials. Market How to Pick an AI Agent Platform Without Getting Locked In Engineering 6 Agentic Knowledge Base Patterns: How AI Agents Are Replacing Static Wikis Market From Pilot to Production: The Enterprise AI Agent Readiness Checklist Engineering The Lethal Trifecta: Securing AI Agents Against Data Exfiltration (Enterprise Checklist) Engineering LLM Evaluation Beyond ROUGE: Building Custom Evals for Enterprise Agents Field guide Agentic Workflows vs Traditional Automation: When to Choose Each (2026 Guide) Resource CAISO Storage Daily Brief: Dispatch, Prices & Ancillaries (2026) Resource ERCOT Prices Today: Day-Ahead, Real-Time & Ancillary (2026) Market The Hidden Tax of Agents: Compound Error and Cost Explosions Resource Clinical Trial Sponsors List: Top 500 Recruiting (2026) Resource FDA 510(k) Cleared Companies: Complete List (2026) Resource Fortune 500 Full List (2025 Rankings) Resource Y Combinator Companies: Full List (2026) Podcast Personal AI Infrastructure: How Scaffolding Changes Everything Engineering Five Production Agent Patterns from Anthropic's Playbook