Blending AI insights with Product Management wisdom 🤖🚀. Exploring the intersection of tech and strategy. Here to share, learn, and evolve. #AI#ProductMgmt
Dmitry, CEO of https://t.co/vx9x21Ijxg, runs three AI agents across three different computers that manage his relationships, his inbox, and his fundraise, and he showed us the entire stack live.
00:00 The messy lab: how Dmitry treats his whole setup as constant experimentation
01:17 Meet the team: Watson, Sherlock, and Harry, and which machines they each run on
08:09 G Brain: the memory system behind 20,000 connected contacts
09:10 Semantic search in action: finding the fastest path to anyone in his network
15:24 What this would cost on Opus versus what he actually pays
17:20 How Linear manages communication between three separate agents
18:03 Live demo: a task gets created, triaged, and routed with zero human input
20:20 Building a 1,200-person enriched VC list for his current raise
27:59 The 7am scan that tells him exactly who's worth engaging on Twitter
30:17 The bridge people strategy for reaching someone who doesn't know you exist
33:31 What happens every night while he sleeps: the "dreams" cleanup process
34:42 His bartender and barback framework for where humans still belong
54:10 Why two decades of experience alone won't get you hired anymore
56:13 The mental model for deciding what to automate first
1:00:06 What Jensen Huang meant by "the self-driving computer"
Claude Code gets useful when you stop pasting context and connect one real tool. A first MCP server turns Claude from a smart prompt box into part of the workflow. Start with one source you keep re-pasting, verify it in /mcp, then ask a real work question. #ClaudeCode#MCP
Stop asking which model is best. That's the wrong question.
The better question: What stack are you building?
AI advantage is shifting from model choice system design.
The future AI "brain" isn't one model. It's a layered system that reasons, researches, executes, and remembers.
Claude Code + Obsidian has completely changed how I operate daily.
I recently built an AI second brain that runs my entire life.
The best part is, it only takes a few minutes to build.
This video will teach you EXACTLY how to build one yourself - inspired by Andrej Karpathy:
You're picking your AI model based on a benchmark with a 57% error rate in Virology.
Same test, 3 implementations, 3 different winners.
The leaderboard isn't lying — it's just not measuring what you think it is.
Lazy evals are dead. Build your own.
🚨 Anthropic just restricted Claude subscriptions for third-party agent tools like OpenClaw.
AI subscriptions are priced like gym memberships — assuming most won't max out.
But AI agents don't sleep. They maximize every token. The economics break.
Platform risk is real.
Tomorrow’s a huge day for Pakistan. Your vote is important. As soon as you’re able to, please post a photo or video saying “I voted PTI” with a hashtag #votePTI to show your support!
Pakistan Zindabad! 🇵🇰
The former prime minister of Pakistan writes from prison that his party is being unfairly muzzled, in a guest essay for The Economist https://t.co/mKnvCidPm6
Since my colleagues and I raised our concerns about human rights in Pakistan last November, things have only gotten worse. There can't be free and fair elections when one of the opposition parties has been criminalized.
https://t.co/ZG8giXTjVM