john tan

Spoke to a CRO of a hot Series B startup yesterday: “We don’t have the knowledge internally to implement AI and agents into our process.” Toast. You are toast. That is unacceptable. Everyone can learn. There is zero excuse for the above.

a small window into the opportunity of AGI

We doubled Claude usage on weekends, and outside 5–11am PT on weekdays for the next 2 weeks.

Claude Opus 4.6 with 1M context is now included in Max, Team, and Enterprise subscriptions. Select it with /model. It now counts against your normal plan limits, no extra usage required.

🎶 I've been using voice mode to write much of my CLI code this last week Can't wait to hear what you think.

McKinsey built an AI chatbot (Lilli) trained on 100 years of its work 100k documents and interviews. 70% of 45k employees use the tool, making 500k prompts a month. A research firm hacked into it with “full read and write access to production database” including “47m chat messages about strategy, M&A, client engagement, all in plain text along with 728k containing confidential client data, 57k user accounts, and 95 system prompts controlling AI’s behaviour.” Mcksinsey said it has patched up the vulnerability, which was made possible by “publicly exposed API documentation, including 22 endpoints that didn't require authentication…one of these wrote user search queries, and the agent found that the JSON keys (these are the field names) were concatenated into SQL and vulnerable to SQL injection.”

Update: this is now rolled out to 100% of users

The actual guide to agentmaxxing, since everyone’s going to misread this headline: Replit hit $240 million in revenue in 2025 with roughly 70 employees. That’s $3.4 million in revenue per head. A typical SaaS company at that revenue would have 700 people. Replit ran 10x leaner. Amjad Masad just raised $400 million at a $9 billion valuation and announced he’s hiring new grads. But the new grads he’s describing aren’t traditional CS majors grinding LeetCode. He hired an 18-year-old who never went to CS school, learned to code entirely through AI, and is outperforming classically trained engineers. Agentmaxxing is a specific workflow. You take an AI coding agent (Replit, Claude Code, Cursor), describe what you want in plain English, let the agent build it, review the output, iterate. One person running 5-10 agents simultaneously replaces a team of 4-5 junior engineers who each need onboarding, management, and code review. Masad said the quiet part out loud in an interview last year: if you’re an engineering manager at Meta, do you hire four junior engineers with all the overhead, or one senior engineer who can spin up 10 agents? Senior engineer salaries have never been higher. New grads who can’t orchestrate agents are struggling. New grads who can are getting hired at 18. The practical stack looks like this: 1. You become the architect, not the bricklayer. Your job is system design, constraint definition, and quality review. 2. You manage agents like direct reports. Break work into discrete chunks, assign each to a session, review output, course-correct. The best operators run parallel sessions. 3. Clarity of thought matters more than syntax knowledge. Masad said the highest-leverage hires right now are clear thinkers and clear communicators. He called them “consultant types.” 4. You ship 10x the surface area. Replit’s Agent 1 lasted 2 minutes before losing coherence. Agent 4 runs 3 hours doing production work. That capability is 10xing every few months. By next year, agents handle full-day tasks. The new grad who gets this builds more in month one than a traditional hire builds in a quarter. The hiring market for juniors didn’t collapse. It forked into two lanes: those who manage agents and those who compete against them.

When it comes to AI agents / AI tooling + coding, I hear an awful lot of talk about: Efficiency Iteration speed / PR output rate / lines of codes produced I hear zero mentions about: Quality Customer obsession This will bite back, and it probably already is...