Aakash Gupta
product-growth.com: $51K/mo sponsorpg.com: $31K/mo landpmjob.com: $28K/mo youtube.com/@growproduct: $20K/mo
Page 1 • Showing 20 tweets
Cursor's $7,225 invoice broke the internet. So I mapped pricing across 50 AI startups to figure out what's actually working. Six models emerged, four case studies, and one decision tree for picking yours. 🔗 https://news.aakashg.com/p/how-to-price-…
“The biggest beneficiaries of vibe coding are going to be… the shape rotators, not the wordcels” — Palmer Luckey
I got an expert to give me a masterclass on AI evals: (63 mins of stuff most builders learn the hard way) 0:00 - Why your product is lying to you 22:03 - Full eval flow walkthrough 35:42 - Evals ARE your PRD (not QA) 57:52 - Failures at Grammarly, Copilot & Klarna
Claude Cowork has become my most used AI tool. It does everything: Excel. PowerPoint. Browser automation. All in parallel. Plus scheduled tasks that run while I sleep. Here are 6 workflows that replaced hours of my week, with full prompts: https://news.aakashg.com/p/you-should-be…
You Should Be Using Claude Cowork: Complete Guide
.@Replit 25x'ed ARR to >$150m in less than 1 year. Here's your complete guide to using it in <13 mins: 1:01 - Design Mode 6:26 - Fast v Agent Mode 7:19 - Figma Import 7:48 - Connectors 10:26 - Agent 3 11:25 - Deploy
The strongest people I know can admit when they're wrong without their ego shattering. Wrong opinion doesn't mean wrong person. Failed attempt doesn't mean failed life. They separate their mistakes from their identity. That detachment is power.
PMs at Notion are shipping pull requests. Product managers writing and merging code into production. A year ago that sentence would’ve gotten you laughed out of a product leadership Slack. Figma’s PM org reportedly cut their average feature cycle from 6 weeks to 3 after standardizing Claude Code workflows internally. Linear’s team has talked publicly about PMs prototyping directly in their codebase. This tells you everything about where PM hiring is headed. The gap between “PM who can go from insight to working prototype” and “PM who writes a spec and waits 2 sprints” is becoming the new senior vs junior divide. Except it has nothing to do with years of experience. The last time a tool shift redrew PM job descriptions this fast was when Figma made every PM a passable wireframer around 2019-2020. That took 2 years to become table stakes. Claude Code is moving faster because the output is code, not mockups, and code ships.
Claude Code + MCP = Vibe PMing Here's your complete guide with @frankdotlee, Principal AI PM at @Amplitude_HQ: 3:45 - Setting Up Claude Code + MCP 11:08 - Top 5 Use Cases for PMs 40:35 - Biggest Mistakes
You should be researching your users more. So I got an expert to give me a masterclass in how to do it with AI: Step 0: Load Context Into Claude - 8:22 Step 1: Per-Participant Analysis - 16:12 Step 2: Verification - 26:06 Step 3: Auditing AI - 51:31
OpenAI just dropped the best app they've ever made. Codex turns ChatGPT into a delegation engine that works while you sleep. I wrote the complete guide: setup in 5 min, 4 workflows, built-in automations, and mid-turn steering with GPT-5.3 ↓ 🔗 https://news.aakashg.com/p/openai-codex-…
Netflix made $45.2 billion in 2025. YouTube made $44 billion in ad revenue. On a like-for-like basis, these two are nearly identical. YouTube gets to $60 billion by bundling in YouTube TV, YouTube Premium, YouTube Music, and NFL Sunday Ticket. Comparing that combined number to a single streaming service is like comparing Alphabet’s total revenue to Disney+ and declaring Google won the streaming wars. The number that actually matters: YouTube paid creators $100 billion over the past four years while spending $0 on content production. Netflix spent $16 billion on content in 2024 alone and has to keep writing that check or the library evaporates. Google bought YouTube for $1.65 billion in 2006. It now earns that back every 10 days. YouTube built a model where the content creates itself, the audience sells to advertisers, and the platform collects rent on every transaction. Netflix writes a $16 billion check every year just to keep the catalog from shrinking.
NotebookLM is the most underrated AI tool Google has made. I used one notebook to go from zero knowledge to a working prototype, customer discovery materials, and a product deck in a single afternoon. Here's your complete guide: https://news.aakashg.com/p/complete-guid…
.@AnkythShukla on the 25x cost trap killing AI products in production: "Most of the PMs want to do is they think that if in the prototype I have used maybe the most advanced model, I should use the same model in the production as well. And then because of cost considerations, the management has to go ahead and pull the plug." "But now there is a good possibility, which is that maybe another model, which is a cheaper model, can go ahead and produce a similar kind of output." "You can see that the best model GPT 5.1, which is mostly going to be used in your prototype because it's very intelligent. The output is $10 per million token. But for maybe a model such as GPT nano, the output is maybe 0.4, right, which is only 40 cents." "You will only get the confidence of using this when you have created the right kind of evaluations." From the @aakashgupta podcast.
Instead of scrolling, watch this masterclass on Claude Code from the creator
Most PMs hear “Claude Code” and think “that’s for engineers.” That’s the wrong frame. What 41 skills and 7 sub-agents actually give you is a team of specialists that execute PM workflows on command. Need competitive analysis? One prompt. PRD review? Delegated to a sub-agent that returns a summary without eating your main context window. The https://code.claude.com/docs file is the real unlock. It functions as an operating constitution that tells Claude how your specific workflows run, what your standards are, and when to invoke which skill. Every conversation starts pre-loaded with your accumulated decision-making framework. Here’s what took me 6 months to learn: the compounding isn’t in any single skill. It’s in the interactions between them. A sub-agent that researches feeds into a skill that writes, which triggers another that formats for the specific output channel. You build it once, and every future task runs through infrastructure you’ve already laid. Most “PM productivity tools” give you templates. This gives you execution capacity.