Field Notes
Build products, not tool stacks.
Honest notes from building the operating system for product companies - the math, the joins, and the parts most vendors won't say out loud.
-
Context engineering vs prompt engineering: the ceiling
Prompt engineering is not dead, it is no longer sufficient. What separates it from context engineering, and the ceiling neither discipline can raise alone.
ai-agentsmcpcomparisons
-
A2A vs MCP: which protocol do you actually need?
A2A vs MCP, decided: A2A standardizes agent-to-agent messaging, MCP standardizes agent-to-tool access. Which one you need right now, and why.
mcpagents
-
Remote MCP Server: The OAuth Surface You Have to Publish
A remote MCP server is an HTTPS endpoint with an OAuth surface in front of it. What you publish, what you validate, and what an agent reaches once it is in.
mcpagents
-
MCP Resources and Prompts: What They Do and When to Use Them
MCP resources are application-controlled, prompts are user-controlled. That difference, not your server code, decides whether either ever reaches the model.
mcpai-agents
-
Product Management Software With a Free Trial, Ranked
How to tell which product management tools you can start today: card requirements, trial length, and what happens to your workspace when the trial ends.
pricingproduct-management
-
Which Product Management Tools Have MCP Servers? (2026)
A verified 2026 survey of 16 product tools and their MCP servers - who ships one, what scope it exposes, and the single-vendor silo gap every one shares.
mcpai-agentsproduct-management-tools
-
AI Agent Integration: Ranking the Options by Agent Reach
Four ways to integrate an AI agent with your systems, ranked by what the agent can actually reach and write back to - not by how the connection is made.
integrationsai-agentsmcp
-
MCP Gateway: the four routing shapes, and how to pick one
A gateway makes many MCP servers manageable. It does not reduce how many you run, and it cannot join records across them. The four shapes, compared.
mcpagentsspine
-
AI for Product Managers: What It Does Well and What It Fails
A job-by-job audit of a product manager's week: which recurring tasks AI genuinely does well, which it fails at, and the data precondition that decides.
ai-agentsproduct-management
-
MCP Connector: Three Meanings, and Which One You Need
The phrase mcp connector means three different things, including a car part. Here is the disambiguation, plus the question that decides which one you need.
mcpintegrations
-
User Research Platforms: 4 Jobs, Ranked, and How to Pick One
User research platforms split into four jobs that do not substitute for each other. Ranked by need, plus the one thing none of them fix.
customer-researchproduct-tools
-
MCP Tools: One Phrase, Three Meanings, and How Selection Works
The phrase MCP tools collides three different meanings. Here is the disambiguation, plus how a model actually picks a tool: name, description, schema.
mcpai-agents
-
MCP Apps in Practice: Which Tools Deserve a UI
We shipped three MCP Apps. The two decisions page one skips: which tool earns an interactive ui:// view, and what it does on a host with no app support.
mcpagents
-
MCP Client: What It Is, and Why the One You Pick Matters
An MCP client is the part of your AI app that connects to one MCP server. Which client you run quietly decides how much of any server you can reach.
mcpagents
-
Product Prioritization Framework: Rank Features by Revenue
A product prioritization framework ranks features by request volume, revenue at stake, effort, and confidence, then verifies the bet after the feature ships.
prioritizationproduct-management
-
Best Product Management Tools for Startups (2026)
The product management tools startups actually shortlist in 2026, ranked - where each breaks as you hire, and when a spreadsheet wins.
product-managementstartupstools
-
MCP OAuth: What You Approve When You Click Authorize
The Authorize screen on a remote MCP server is a grant made on behalf of your workspace. What the scope limits, how long it lives, and who can revoke it.
mcpagentstrust
-
Product Management Software Cost for Startups (2026)
The real cost math of a startup product stack: four separate per-seat tools vs one flat price, the arithmetic behind both, and when point tools win.
pricingstartupsproduct-management
-
MCP Inspector: How to Test an MCP Server You Did Not Write
MCP Inspector shows a server's tool surface with no model in the way. An ordered procedure for evaluating a server you did not write, ending in a verdict.
mcpagents
-
OKR Software for Product Teams: 5 Options, Ranked
A ranked, honest guide to OKR software for product teams: dedicated OKR platforms, HR suites, work tools, and the connected-spine option - and when each wins.
okrproduct-managementcomparisons
-
MCP Server Not Working? Diagnose It in Five Layers
Your MCP server will not connect. A five-step diagnostic order that isolates which layer failed - client, process, transport, auth, or tool surface.
mcpmcp-serveragents
-
Product Requirements Document: How to Write One
What a product requirements document is, what to put in one, and a copyable PRD template - plus the part every guide skips: keeping the doc from going stale.
prdproduct-management
-
Too many MCP servers? It's a shape problem, not a count
Connecting five or six MCP servers made your assistant worse. The fix isn't a gateway - it's deciding which questions you actually need answered.
mcpagents
-
MCP Security: Are MCP Servers Safe? A Decision Procedure
MCP security depends on which server. A stranger's local server and a hosted OAuth endpoint are different risks, and here is a decision procedure for both.
mcpsecurity
-
RICE Framework Template: The Sheet, and Keeping It Honest
A copyable RICE scoring template with fixed scales and worked rows, plus the setup discipline that keeps scores comparable past the first session.
prioritizationproduct-management
-
Local MCP Server: Running One on Your Own Machine
How to run a local MCP server over stdio, where it stops paying at team scale, and when local is genuinely the better answer.
mcpagents
-
MCP Server Examples: What They Actually Do (2026)
Most MCP server lists sort by vendor. This one sorts by reach - what each example server can see, what a call returns, and what it structurally cannot.
mcpagents
-
Using AI to Triage Customer Feedback at Scale
AI triage for customer feedback is not support-ticket triage. What the classification layer does reliably, where it is wrong, and what it must attach to.
insightsaicustomer-research
-
Open Standards vs Proprietary AI Tools: What to Ask a Vendor
Proprietary AI connectors lock your integration layer, not just your data. What to ask a vendor about open standards - and when a closed integration still wins.
mcpintegrationstrust
-
Which product tools have the most integrations?
Integration counts are marketing numbers and none of them are comparable. What actually decides whether connecting a tool helps - and how to compare honestly.
integrationsproduct-stack
-
AIOProductOS is now a connector in Anthropic's directory
Our spine MCP is approved and listed. 71 tools over one endpoint that give your AI the whole customer record - revenue, feedback, roadmap and outcomes.
mcpannouncements
-
Continuous Product Discovery Without a Research Team
Continuous product discovery fails on plumbing, not motivation: a recruiting pipeline that refills itself, notes that land somewhere, and a ranked decision.
customer-researchdiscoveryproduct-management
-
How to Communicate a Roadmap to Stakeholders, Not Present It
Roadmap communication is a system, not a meeting: one source filtered per audience, a rhythm between reviews, and a protocol for saying something slipped.
roadmapstakeholderscommunication
-
Product Analytics Without a Data Team: The Four Lies
Product analytics without a data team is a judgment problem, not a tooling one - the weekly routine to run, and the four ways self-serve analysis lies.
product-analyticsproduct-opsproduct-management
-
Switching Product Management Tools: What It Really Costs
Switching product management tools costs more in lost history than lost data. A decision matrix, a sequencing playbook, and the honest case for staying.
trustproduct-opsmigration
-
Customer Feedback Without a Research Firm: DIY vs Pay For
Which parts of a research firm's job a small product team can self-serve this month, which two are worth paying for, and how to keep DIY feedback alive.
customer-researchinsights
-
Outcome-Based Roadmaps: Wire the Outcome to the Work
How to write an outcome-based roadmap that stays wired to the shipped work and the metric judging it - plus what to do when the number doesn't move.
roadmapproduct-managementmetrics
-
Feature Adoption Benchmarks: What Good Looks Like
Why published feature adoption rate benchmarks contradict each other, and how to set a defensible internal bar per feature class before you ship.
product-analyticsfeature-adoptionmetrics
-
How to Audit Your Product Stack (Costs, Overlap, Data Flow)
Most stack audits only count invoices. Add overlap and data flow - the seams between your tools - and run a repeatable product stack audit in a day.
tool-sprawlsaas-costsproduct-ops
-
Product Roadmap Templates That Don't Rot
Three copyable product roadmap templates - Now/Next/Later, outcome-based, and quarterly - plus the maintenance rules that keep a roadmap true after week three.
roadmapproduct-managementproduct-ops
-
What Customers Say vs What They Actually Need
Customers are unreliable narrators of their own needs. How to triangulate a stated request against usage, support and revenue - then check if shipping worked.
customer-researchinsightsproduct-management
-
EU Data Residency for Product Tools: What GDPR Teams Need
GDPR doesn't require EU storage - it regulates transfers. A concrete audit for finding where each tool in your product stack holds EU personal data.
trustgdpr
-
MCP vs API integrations: what changes for a product team
MCP vs API, decided for product teams: who consumes each, what an MCP server changes, when a plain API still wins, and why neither joins your data.
mcpintegrations
-
Why Customers Churn: How to Run a Churn Analysis in a Week
Exit surveys only capture the customers who answer. A runnable churn analysis that reconstructs why customers leave from usage, billing and support evidence.
product-analyticsmetricscustomer-research
-
How to Say No to Feature Requests (With Data)
How to say no to feature requests without damaging trust: score the request on accounts, revenue, and what it displaces - then send the evidence with the no.
product-managementprioritization
-
How Many Customer Interviews Are Enough? A Saturation Test
How many customer interviews are enough? Skip the magic number - run a saturation test after each one, and size the round to the decision you're making.
customer-researchinsights
-
Product Manager Tools for Remote Teams: Fewer Surfaces
The standard remote PM stack is six tools that never reference each other. A better test: can a teammate reconstruct the why from the artifact, with no meeting?
product-opstool-sprawlproductivity
-
Connect Your AI Assistant to Jira or Linear via MCP
How to connect your AI assistant to Jira or Linear via MCP in a few steps - and when a single tool's MCP beats one spine MCP over a joined record.
mcpai-agents
-
Vendor Lock-In in Product Tools: How to Keep Your Data Yours
Vendor lock-in in SaaS traps your data, not just your workflow. A concrete portability checklist for product teams - and the honest trade-off in consolidating.
trustsaas-costs
-
Free vs Paid Product Management Tools: What's Worth It?
An honest guide to free vs paid product management tools - the real total cost, seat caps, data lock-in, and when free actually wins.
product-managementpricingtool-sprawl
-
Now Next Later Roadmap vs Timeline Roadmap: Which to Use
Now Next Later roadmap vs timeline roadmap: a clear decision rule by audience and commitment type, plus a copyable template. It is not either/or.
roadmapproduct-managementprioritization
-
Product Metrics That Matter (and the Vanity Ones to Drop)
The product metrics that matter tie to a decision and an outcome. A field guide to the numbers worth tracking - and the vanity ones to retire.
product-analyticsmetrics
-
Confluence MCP: setup, real uses, and the wall you'll hit
How to connect an AI assistant to Confluence via MCP, what it's genuinely good for, and the questions a docs silo can never answer.
mcpconfluenceai-teammate
-
How to Do Customer Interviews Without Feeling Awkward
How to do customer interviews when the asking feels awkward: copyable scripts for the invite, the silence, and the sales question, plus what to do after.
customer-researchinsights
-
A Loom Alternative Built for Repeatable Product Demos
Most Loom alternative lists compare screen recorders. Here is the other lane: scripted, re-runnable product demos you regenerate instead of re-record.
demo-videoproduct-marketingstudio
-
How to Build an MCP Server (What We Learned Building Two)
The tutorials stop at hello-world. What decides whether your MCP server is usable: tool surface, return shapes, transport, auth, and versioning.
mcpmcp-serverai
-
How to Make a Product Demo Video (Scripted and Repeatable)
How to make a product demo video: a copyable script skeleton, length guidance by demo type, and how to re-shoot it after every release.
demo-videoproduct-marketingstudio
-
AI Agent Workflows for Product Managers: 5 That Work Today
Five AI agent workflows product managers can run today - each bounded to the real product data it reads and writes, from feedback triage to weekly reporting.
ai-agentsworkflowsproduct-management
-
SaaS Tool Sprawl: What Your Stack Actually Costs
SaaS tool sprawl costs far more than license fees. The real bill is the integration tax between disconnected product tools - here's how to count it.
tool-sprawlsaas-costsvendor-consolidation
-
Amplitude vs Mixpanel vs PostHog: 2026 Comparison
Amplitude vs Mixpanel vs PostHog in 2026: how the three compare on pricing model, depth, replay, flags, and free tier - plus the gap all three share.
product-analyticsanalyticsamplitude
-
How much does Jira actually cost? The full-stack math
Jira's sticker price is the smallest line item. The real bill: Confluence, AI credits, marketplace apps, and the tools Jira makes you buy around it.
jirapricingtool-stack
-
Product analytics pricing 2026: the three meters that get you
Event-metered, MTU-metered, or flat: how analytics vendors actually charge, why renewals surprise teams, and how to model the bill before signing.
analyticspricingamplitude
-
Product event tracking without a data team
Ship product event tracking in an afternoon with a few-KB SDK - no data team, no warehouse - and get analytics already joined to revenue.
sdksproduct-analyticsproduct-ops
-
RICE vs WSJF vs Value/Effort: Pick a Prioritization Framework
RICE vs WSJF vs Value/Effort compared: what each scores, when to use each, and why the framework matters less than the data feeding its inputs.
roadmapproduct-managementprioritization
-
The weekly signal memo: kill the status meeting
How we replaced Monday status theater with a machine-written memo from our own product data - what's in it, why it's deterministic, how to build yours.
dogfoodingai-teammateproduct-ops
-
AI agents vs copilots in product tools
Copilots suggest in-context; agents claim and ship whole tasks for review. An honest comparison of where each wins in product management tools.
agentsaicomparisons
-
Best Project Management Tools for Product Teams (2026)
The listicles conflate project and product management. Here's how to pick tools that track execution AND the customer, feedback, and revenue behind the work.
project-managementproduct-managementtool-comparison
-
Why Is Nobody Using My New Feature? A Diagnosis Guide
Low feature adoption is a funnel problem, not a promotion problem. Diagnose where adoption broke - aware, tried, adopted, retained - and fix the real cause.
feature-adoptionproduct-analyticsproduct-management
-
Best MCP Servers for Product Teams (2026)
An honest comparison of MCP servers for product teams - Linear, Notion, and Productboard tool-MCPs versus a spine-MCP that exposes the join across systems.
mcpcomparisonsproduct-management
-
Can AI Write Your PRD? What Works and What Fails
AI reliably writes a PRD's structure but invents the substance - customers, revenue, decisions - unless it's grounded in your real product data.
ai-product-managementprd
-
The Real Cost of Per-Seat Pricing for Product Teams
Per-seat SaaS pricing hides four real costs for product teams: dead seats, collaboration rationing, AI credit fees, and a bill that scales with headcount.
pricingsaas-costs
-
A drop-in in-app chat SDK for SaaS
A drop-in in-app chat SDK that gzips to a few KB and logs every conversation to a spine, so chats join to the customer's revenue and shipped work.
sdkschatsaas
-
How to Measure Feature Adoption (and Prove It Worked)
How to measure feature adoption with the adoption-rate formula, the denominator traps to avoid, and how to prove the feature actually moved an outcome.
product-analyticsanalyticsproduct-ops
-
How to Rank Your Roadmap by Revenue (Not Votes)
How to prioritize a roadmap by revenue: attach a real dollar figure to each item - churn risk, expansion, deals at stake - then rank by revenue, not votes.
roadmapproduct-managementrevenue
-
The AI teammate that claims a task and ships it
Walk the loop: an AI agent claims a task, works over MCP against the joined spine, opens a PR, a human reviews, and the outcome lands on the same record.
agentsmcpai-teammates
-
Airtable vs Jira for Product Management: Which Is Better?
Airtable vs Jira for product management: Jira wins on agile delivery, Airtable on flexible ops. But both are per-seat and disconnect work from revenue.
jiracomparison
-
Surveys vs Interviews: Which Customer Research Method to Use
Surveys measure breadth, interviews explain the why. When to use each, how to combine them, and how to stop research from dying in a doc.
customer-researchinsights
-
MCP for product managers: what it changes
MCP lets a PM ask across feedback, revenue, and work in one place instead of four tabs. What it changes day to day - and when a single-tool MCP is enough.
mcpai-product-managementproduct-management
-
How to Give an AI Assistant Access to Your Product Data
A vendor-neutral guide to safely connecting an AI assistant to your product data over MCP - what it can see, what it's allowed to do, and how to scope it.
mcpai-agentsproduct-data
-
Is Your PM Tool Training AI on Your Data?
Most PM tools bolt a model vendor behind an AI button. Here's how to verify - not assume - whether yours trains AI on your customer data.
ai-policydata-privacyai
-
Product management MCP: the complete 2026 guide
The complete 2026 guide to a product management MCP - what it is, why tool-scoped MCPs hit a silo wall, and what a spine-level MCP changes.
mcpproduct-managementai-product-management
-
A Segment alternative for product teams
Segment fans data out to a marketing stack; a product spine with 100+ built-in connectors joins it on one record. Honest comparison of where each wins.
connectorscomparisonsproduct-analytics
-
How to add session replay to your product
Add session replay with one async script tag. Replays tie to the same customer record as revenue and events - and mask PII by default.
sdkssession-replayproduct-analytics
-
Product Analytics Tools: How to Choose (2026)
How to choose product analytics tools in 2026: a five-criteria framework covering pricing models, time-to-value, and whether your data stays connected.
product-analyticsanalyticsproduct-ops
-
How to Write a Product Roadmap That Actually Works
How to write a product roadmap that actually works: start from outcomes, prioritize by revenue and customer requests, and pick the right format.
roadmapproduct-managementproduct-ops
-
5 agentic workflows that save PM time
Five concrete AI-agent workflows product teams run today: feedback triage, weekly signal memo, roadmap-drift check, PRD review, release receipt.
agentsai-teammatesproduct-ops
-
Best Product Management Tools for Small Teams (2026)
Honest 2026 guide to product management tools for small teams - verified pricing for Linear, ClickUp, Notion, Productboard, and the real cost of a full stack.
product-managementcomparisons
-
How to Turn Customer Feedback into Product Changes That Matter
Turn customer feedback into product changes by joining feedback with revenue, work, and outcomes - prioritize by revenue at stake, ship, then verify impact.
product-managementproduct-ops
-
Claude MCP & Cursor MCP - connect your product data
Set up Claude MCP and Cursor MCP with your product data: add a hosted MCP server URL, authorize over OAuth, then ask cross-system questions in plain language.
mcpaiconnectors
-
Connect Stripe to your product analytics - no data team
Connect Stripe so revenue joins product usage on one spine - no ETL, no warehouse. Then answer revenue-weighted funnels and who-actually-pays in minutes.
connectorsstripeproduct-analytics
-
The 1.7 KB product-analytics SDK
A product-analytics SDK that gzips to 1.7-5.2 KB, loads async, and lands events on a spine where they join to revenue and shipped work.
sdksproduct-analyticsperformance
-
Can an AI agent do product management?
The honest answer: an AI agent can run the mechanical parts of PM - triage, drafts, drift checks - but judgment and prioritization stay human.
agentsai-product-managementai-teammates
-
What is MCP? The Model Context Protocol in AI, Explained
MCP, the Model Context Protocol, is the open standard that lets AI assistants call real tools and data. What it is, how MCP servers work, and when to skip it.
mcpmcp-serverai
-
What is a product-data connector?
A product-data connector pulls data from an external tool and lands it on a shared spine, so it joins to revenue and work on one record - not a raw API pull.
connectorsintegrationsproduct-ops
-
100+ connectors, one shared spine
Connected, not consolidated: 100+ connectors land data from every tool onto one shared record, so product questions become joins instead of reconciliations.
connectorsspineproduct-ops
-
Atlassian AI training: how to opt out or export your Jira data
From Aug 17 2026 Atlassian contributes customer data to improve AI for all customers. Which plans are ON by default, and where the org-level setting lives.
atlassianjiraai-policy
-
Spine-MCP vs tool-MCP: why product MCPs go wrong
Productboard, Linear, and Notion shipped MCP servers in 2026 - each exposes its own data. A spine-MCP exposes the join across systems on one customer record.
mcpaiproduct-management
-
Context switching is quietly eating your product team
Every tool holds a partial view of your customer. Your team becomes the integration layer. Here's what that costs - and what structurally fixes it.
productivityproduct-opstool-sprawl
-
Jira alternative for small product teams - honest guide
Small teams often outgrow Jira before they grow into it. A fair comparison of Jira and AIOProductOS across pricing, setup, and the question Jira can't answer.
comparisonsjiraproduct-management
-
The real cost of a product team's tool stack (2026 math)
SaaS sprawl costs more than the license fees. Here is what a typical 8-person product team actually pays - and what the hidden costs look like.
tool-sprawlpricingproduct-ops
-
What is an AI teammate? Real agents vs agent-washing
"Agent" is 2026's most abused word. A precise definition, a table that cuts through the noise, and five questions to ask any vendor.
ai-teammatesmcpagents
New posts land here first - subscribe via RSS, or follow the build on LinkedIn, X, and YouTube.