Thilak Sriram P

hey @WisprFlow @tankots the product is super slow in the past week. Please send fix asap

Yesterday I demoed Claude Code to 20+ PMs and 30+ folks at our company Product Talk. The #1 question: where do I start. My 4-step setup, works for any knowledge worker (PM, designer, marketer, analyst, lawyer): 1. Install. Run npm install -g @anthropic-ai/claude-code, then type claude inside your work folder. That folder becomes the AI's home base. 2. Add a CLAUDE.md to that folder. 5 lines on who you are, what you work on, who your stakeholders are. This is the briefing it reads every session. 3. Connect one MCP. Just one. Granola, Figma, Jira, Gmail, Trello, Drive, your data warehouse. Pick the tool you open 10 times a day. The unlock is immediate. 4. Write your first slash command. Mine started as /prep-my-day. 8 lines telling Claude to pull my calendar, unread Slack mentions, and top 3 priorities every morning. You stop prompting daily and start running a routine. Day 1 you won't feel much. Week 4 you won't remember how you worked without it. Reply if you want my CLAUDE.md template. Claude Code is my personal OS. It runs my morning briefing. Sharpens my product thesis before every executive readout. Pressure-tests the bets I'm placing. I share real techniques from how I actually run it. Built by a product leader. Useful for anyone (product managers, lawyers, accountants, HR, analysts, doctors) who works with information, not just for engineers. #ClaudeCode #AIProductivity #KnowledgeWork @AnthropicAI

Claude Code is my operating system. 50+ custom skills. 7 automated hooks. 5 data integrations. All on one CLAUDE.md file. Today's piece: the 7 hooks that run my AI without me asking. - I used to forget to log decisions. - I used to forget to update memory. - I used to forget to double-check the framing before anything went to the C-suite. Now Claude does those checks itself. Every session. Every message. Every time I'm about to send something exec-bound. The 7 hooks: 1. SESSION START. Loads my living memory (what we're working on this week) before I type anything. No more re-explaining context. 2. POST-RESPONSE. After every substantive answer, appends one line to my hourly log. I never lose what was decided. 3. STYLE GUARD. Detects when I'm drafting for the C-suite. Injects a reminder: proportional framing, numbers with source, one page max. 4. EFFORT GUARD. Detects high-stakes work (red-team, board readouts, coaching prep). Asks if I want maximum thinking mode before responding. 5. HEARTBEAT. Every 30 minutes during work hours. Reads my calendar, fresh meeting transcripts, Slack DMs, file changes. Captures what shifted. 6. END-OF-DAY. Compresses today's log into a journal entry. I never wonder "what did I work on yesterday?" 7. NIGHTLY DISTILL. At 23:00 local time, rolls the week up into a long-term memory file. Themes, patterns, recurring asks. I wake up to my own week, summarised. How to set this up: 1. Open ~/.claude/settings.json 2. Add hook entries under the "hooks" key (Claude will write them for you, just ask) 3. Each hook is a small shell script or markdown file pointing to a trigger 4. Save. Restart Claude. That's it. Now the hooks run on the schedule you set. The compounding part: Week 1: I stop forgetting to log decisions. Week 4: I stop second-guessing my exec drafts. The hook already fixed the framing. Week 8: My memory file has 60 days of context I never had to write manually. Hooks are the difference between a chatbot and an operating system. A chatbot waits for you to ask. An OS already did the work. (Some of these hooks spawn sub-agents that debate before Claude answers me. That is a tweet for next week.) Claude Code is my personal OS. It runs my morning briefing. Sharpens my product thesis before every executive readout. Pressure-tests the bets I'm placing. I share real techniques from how I actually run it. Built by a product leader. Useful for anyone (product managers, lawyers, accountants, HR, analysts, doctors) who works with information, not just for engineers. #claudecode #aiproductivity @AnthropicAI

Claude Code is my operating system as a product leader. 50+ custom skills. 7 automated hooks. 5 data integrations. All on one CLAUDE.md file. Today's piece: how my meetings stopped living in my head. 1/ I used to leave 6 meetings a day and remember 2. 2/ Now Granola transcribes every call into Claude. My 1:1 prep starts with what I actually said last time. 3/ Memory stopped being a constraint. How I set this up (one evening, no code): 1. Installed Granola (granola.ai). It listens to every meeting on my laptop and stores the transcript locally. No bot joining the call, no extra Zoom link. 2. Connected Granola to Claude Code via MCP. One config entry. Claude can now read every transcript I've ever recorded. 3. Added one line to my CLAUDE.md (the file Claude reads at the start of every session): "Before any 1:1, pull the last 3 meetings with this person from Granola. Summarise what we discussed, what was decided, what they owe me." 4. Wrote a /sync-meetings skill. At the end of each day, it pulls the day's meetings, extracts decisions and action items, files them into a memory folder. Took 20 minutes. Claude wrote most of it. 5. Now I open Claude and type: "What did John and I land on the x-product line POC last week?" It pulls the actual transcript. No more searching notes. No more "I think we said..." The compounding part: Week 1: search works. That alone is a quiet win. Week 4: I stop preparing 1:1s from scratch. The agenda writes itself from what's still open. Week 8: Claude Code starts flagging when someone has raised the same blocker twice and I missed it. PMs talk for a living. The notes should compound. Claude Code is my personal OS. It runs my morning briefing. Sharpens my product thesis before every executive readout. Pressure-tests the bets I'm placing. Raises the bar on every 1:1 with my PMs. I share real techniques from how I actually run it. Built by a product leader. Useful for anyone (product managers, lawyers, accountants, HR, analysts, doctors) who works with information, not just for engineers. #claudecode #aiproductivity @AnthropicAI

If you're a PM waiting for the right moment to set this up: Start today. One file. Tell it what you do. The compound curve starts on day one, not after you figure out the architecture.

50+ skills. 7 hooks that fire automatically at session start, session end, before every tool call. None of it was planned from day one. It grew from one question asked weekly: what did I do repeatedly this week that Claude could have done with me?

My #claudecode setup today: 50+ custom skills, 7 automated hooks, 5 data integrations. It runs my morning briefing, syncs my meetings to memory, coaches me before every 1:1. I started with one file and 10 minutes. The only thing you need to start: 1/ Install Claude Code. 2/ Create a CLAUDE.md in your folder. 3/ Write: who you are, what you do, your team does, the acronyms only your company uses. That's it. Claude stops asking you to re-explain yourself. That alone changes everything. More in 🧵⤵️ Claude Code is my personal OS. I share real techniques from how I actually run it. Built by a product leader. Useful for anyone (product managers, lawyers, accountants, HR, analysts, doctors) who works with information, not just for engineers. #ClaudeCode #AIProductivity #KnowledgeWork @AnthropicAI @dario_amodei

🚨 SUPREME COURT : "If both parents are IAS officers, why seek reservation?" "The parents have studied, they are in good jobs, they are getting good income, and the children want reservation again" "We will never get out of it" "With educational and economic empowerment, there is social mobility" "See, they should get out of reservation"

@WisprFlow @tankots Thanks WisprFlow, problem solved!

Hey @WisprFlow @tankots need urgent help. Monthly subscriber since April with a valid card on file. No auto-debit occurred month, I paid manually when prompted in App but my account still shows Basic on every device. Can someone reconcile this ASAP? More info in thread

The one rule that makes it work It synthesizes all of that into 1 or 2 lines under today's date. If nothing meaningful happened, it writes nothing. That single rule (skip the empty) is what keeps the memory readable a year out.

Four data sources - Every 30 minutes the job runs a slash command that pulls: - Calendar events that just ended (icalBuddy) - Fresh #granola meeting transcripts - Relevant Slack Channels - File changes in my knowledge folder (git diff)

The last 90 days I've used #claudecode as my personal OS. My decisions stay grounded in data and meeting context I'd have forgotten. My coaching is sharper. My context across teams stays connected. Sharing the hands-on tips that changed how I work. Here's the first #heartbeathook

@WisprFlow @tankots Sent an email as well