Newsletters

Create a dedicated ChatGPT/Claude Project for each major initiative — dump all context (docs, Slack threads, meeting notes) into it as your external memory

Use AI to synthesize scattered information before making decisions, not just to generate content

The biggest PM bottleneck isn't ideas or skills — it's context management across dozens of sources

AI as a 'second brain' works best when you feed it real, messy context rather than clean prompts

Think of Claude Code as 'Claude Local' or 'Claude Agent' — it's not just for coding

Running AI locally lets it handle larger files, longer tasks, and direct computer interactions that cloud chatbots can't

Non-technical people are using it for file organization, data analysis, content creation, and workflow automation

The key unlock is giving AI access to your actual files and system — context is everything

Apply the 'barbell' strategy to your reading: consume either up-to-the-minute content or truly timeless pieces

The best product thinking often comes from outside product management — strategy, psychology, design, and economics

Timeless essays compound in value — re-reading them at different career stages reveals new insights

Build a personal library of foundational reading that shapes how you think, not just what you know

Vibe coding works best when you start with a genuine personal or work problem, not a hypothetical project

The most useful vibe-coded tools are small, specific utilities — not ambitious platforms

Try multiple AI coding tools simultaneously (Cursor, Replit, Claude Code) since they have different strengths

The barrier to building software has effectively dropped to zero — the bottleneck is now identifying what to build

An AI model learns patterns from massive datasets, similar to how a child learns language through exposure

Tokens are the units AI models process — roughly ¾ of a word in English. Context windows limit how much the model can 'remember'

RAG (Retrieval-Augmented Generation) lets AI access external knowledge, solving the problem of outdated or missing training data

Temperature controls randomness: low temperature = predictable outputs, high temperature = creative/varied outputs

PM job openings are up 53.6% from the 2023 low and still climbing — the worst is behind us

AI-related PM roles are the fastest-growing segment, making AI literacy a career accelerator

The market favors PMs who can demonstrate impact with data, not just process expertise

Despite recovery, competition remains fierce — differentiate through specialization and visible work

Writing evals is quickly becoming a core PM skill — every AI product needs systematic quality measurement

Start by defining clear, measurable criteria for what 'good' output looks like before building anything

Build diverse test datasets that cover edge cases, not just the happy path

Evals should run automatically and continuously — not just once before launch

PMs can now build functional prototypes in minutes using AI coding tools — no engineering resources needed

Use prototypes to test ideas with users before writing a single PRD or involving your engineering team

Start with simple prompts describing what you want, then iterate — don't try to spec everything upfront

The best prototyping workflow: describe → generate → test with users → iterate → hand off to engineering

PMs are expected to be on both maker and manager schedules — accept this unfairness and build systems to cope

Use AI to handle the routine parts of PM work (status updates, meeting notes, data pulls) so you can focus on judgment calls

The 'great flattening' means fewer layers and more scope per PM — those who leverage tools will thrive

Build repeatable systems for recurring work instead of doing everything from scratch each time

Duolingo's growth breakthrough came from gamification mechanics (leaderboards, streaks, notifications), not marketing spend

Leaderboards alone drove 17% more learning time — competition is a powerful motivator when designed well

Streaks create daily habits that compound into long-term retention — the fear of losing a streak is stronger than the desire to learn

Treat growth as a product discipline with its own team, metrics, and experimentation culture — not a side project