The Delvyn Studio Blog
Product thinking strategies, agent specification best practices, and insights for AI-native teams that build the right things.
AI Made Building Faster. It Didn't Make You More Sure What to Build.
AI Made Building Faster. It Didn't Make You More Sure What to Build.
AI accelerated delivery. It did not accelerate the judgment that decides what to deliver. The 89%/6% split Cagan surfaced in July is not a technology failure — it is a governance failure. The tool that closes it is not another AI agent. It is the operating model that tells the agent what to build.
An Operating Model Is Not an Operational Layer
An Operating Model Is Not an Operational Layer
Some PM tools now claim to be the operational layer for the product operating model. That framing sounds sharp, but it quietly reduces the operating model to a process add-on bolted onto delivery. The operating model is upstream of all that.
The Hard Part Was Never the Building
The Hard Part Was Never the Building
Cagan just published a fresh definition of the product role, built on Benedict Evans' argument that most people aren't tool builders. The three skills he lands on are all judgment, not execution. Which raises an operational question nobody in the tooling category is answering: if judgment is now the scarce input, where does it actually live?
Every PM Tool Has an Agent Mode. None of Them Ship the Operating Model.
Every PM Tool Has an Agent Mode. None of Them Ship the Operating Model.
The PM tooling category is racing to ship agent modes. Productboard Spark drafts specs and runs competitive analysis. Pendo's Leo monitors usage and generates in-app guides autonomously. These are genuinely useful capabilities. But every one of them automates output. None of them ship the operating model the agent needs to produce output worth acting on.
When Feasibility Risk Collapses, Discovery Is the Risk Left
When Feasibility Risk Collapses, Discovery Is the Risk Left
Marty Cagan's value, usability, feasibility, and viability model still works. But AI coding agents changed the bottleneck. Discovery quality now decides whether speed creates value or waste.
MCP Finalizes the Pipe. Your Operating Model Is Still the Payload.
MCP Finalizes the Pipe. Your Operating Model Is Still the Payload.
The MCP 2026 Release Candidate finalized on July 28, giving every coding agent a standard way to read product context. The transport problem is solved. The payload problem — most teams have no written operating model for an agent to read — has not moved at all.
The AI Productivity Paradox Isn't a Paradox
The AI Productivity Paradox Isn't a Paradox
AI made delivery faster and outcomes didn't follow. Cagan calls it the AI Productivity Paradox and says it isn't really a paradox. It's what happens when you accelerate the feature factory instead of redesigning around discovery.
The Synthetic C-Suite Is a Symptom, Not a Solution
The Synthetic C-Suite Is a Symptom, Not a Solution
WorkBoard and Profit.co both arrived at C-suite-titled AI agents within weeks of each other. The convergence is telling. A synthetic Chief of Staff does not build the decision chain strategy needs — it papers over the fact that one was never built.
Great Products Are Not Enough. Neither Is Great Governance.
Great Products Are Not Enough. Neither Is Great Governance.
Cagan makes governance a mission's immune system. But governance holds the exit; the operating model holds the day — and the moment strategy reverts to feature-factory habits between board meetings, the mission is already dying.
GPT-5.6 vs Claude Fable 5: Pick the Model, Feed It Your Product Context
GPT-5.6 vs Claude Fable 5: Pick the Model, Feed It Your Product Context
GPT-5.6 and Claude Fable 5 both pivoted to work agents in a month. For product teams, the winner isn't the model, it's whether the agent has your context.
Tessl Shipped 10,000 Specs. None Help Your PM Write One.
Tessl Shipped 10,000 Specs. None Help Your PM Write One.
Tessl proved spec-driven development is real infrastructure. It also exposed the upstream gap: product managers still need a better place to write the spec first.
Spec-Driven Development Went Mainstream. The Specs Still Start Blank.
Spec-Driven Development Went Mainstream. The Specs Still Start Blank.
Spec-driven development is now the default for serious AI teams. But every spec still starts from a blank prompt with none of your product context.
Give Claude Your Full Product Context (Without Copy-Pasting)
Give Claude Your Full Product Context (Without Copy-Pasting)
Delvyn Studio's MCP server gives Claude real-time access to your entire product context. Stop pasting strategy docs into chat windows.
Marty Cagan says your product coach should be a foundation model. Here is what that actually looks like.
Marty Cagan says your product coach should be a foundation model. Here is what that actually looks like.
Cagan argues foundation models need project instructions and strategic context to be effective coaches. That is literally what spec-driven product management provides.
Claude Code + Linear + Delvyn Studio: The AI-Native Product Workflow
Claude Code + Linear + Delvyn Studio: The AI-Native Product Workflow
Claude Code writes code. Linear tracks issues. Neither carries your product strategy. Agent specs and learning captures close that gap.
Why AI Agents Need Product Context, Not Just Prompts
Why AI Agents Need Product Context, Not Just Prompts
You can give Copilot a perfect prompt and still get the wrong feature. The missing ingredient isn’t better prompting — it’s product context.
From PRD to Agent Spec: What Changes When AI Builds Your Features
From PRD to Agent Spec: What Changes When AI Builds Your Features
Traditional PRDs are long, narrative documents that humans skim and AI agents can’t use. Agent specifications flip the format: structured, context-rich, and optimized for machine consumption.
The Product Thinking Layer: What It Is and Why It Matters
The Product Thinking Layer: What It Is and Why It Matters
Every product team has tools for execution and roadmapping. Few have a system for the thinking that happens before execution starts. The product thinking layer connects vision to implementation.