John M. P. Knox

Are you stuck in #AI when you should be leading? Some Technical founders in my AI Roundtable know they should delegate the AI work. And they're still immersed in AI because they are more skilled than anyone else.

Employees shouldn’t be using their own #AI accounts in your business any more than they should be using their personal Gmail.

I built a Claude Code project that functions as an executive assistant. It pulls calendars, monitors email, manages my to-do list. Morning and evening briefings via push notification. The main function, honestly, is accountability — it guilts me into doing the things I've been putting off. Like a human EA, but one that runs on a cron job. Every founder I know has a list of follow-ups they're not sending. The fix isn't discipline. It's lowering the activation energy: have the agent detect unanswered emails, draft follow-ups in your voice, and put them in front of you. Not send them — draft them. That's usually enough. #FounderLife #AIProductivity #StartupOps

At my AI roundtable, a director described getting an unedited wall of AI text from a direct report. His reaction: "I can do that. What do I need you for then?" The report improved after feedback. But it's happening everywhere — people forwarding raw AI output without thinking. One tactic that works: reply with "What are you expecting me to take away from this?" Forces the sender to actually engage with what they sent. AI makes it trivially easy to produce text. That makes it more important than ever to think before you send. #AIStrategy #Leadership #FounderLife

Where should you deploy AI first? Two places: processes where you have excellent quality control, and processes where "taking a shot" has few drawbacks. The first is obvious — software development, content creation, anything where you can catch mistakes before they reach customers. The second is more interesting. I call them luck agents. An AI sending cold emails based on its own research. An AI monitoring search queries for patterns matching your expertise. The false positives cost you nothing. But one real discovery can change your quarter. "A luck agent is an AI tool that operates in a space where 'taking a shot' has few drawbacks — false positives and false negatives can't hurt you, but a new discovery can help tremendously." Full framework — where to start, delegation, guardrails, and when to automate: movingavg.com/essays/adoptin… #AIStrategy #FounderLife #StartupOps #B2BSales

@noampomsky @benrollert A surprising number of middle managers live here.

The lowest agency person you know: "Well, maybe everything will just work itself out without me having to confront anything that makes me uncomfortable."

Exactly. We shouldn’t look down on a low multiple based on our anchoring. There are many factors that go into that equation, including quality of life, the fact that only an infinitesimal number of businesses are passive, and having more interesting opportunities to focus on. Oh, and let’s not forget that diversification is the only free lunch in investing. Many founders have a more precarious financial position than an outsider could anticipate. Millions in revenue sounds invincible until you meet a few founders who lost their house when their industry was disrupted.

yes! 3x for saas means 3x on revenue, not profit so if you're making $1m/yr, 50% profit you get $3m in one shot with a sale that you can invest and collect truly passive income on (I'm talking index funds) also sometimes taxes are much lower on a sale - in the UK taxes on selling a business are about 20%, taxes on take-home for a business are 45% if you're taking home 500k/year so that's a VERY significant difference in the take home and saas revenue goes up and down like anything else, when you sell you've locked in that revenue

Unless you need a big chunk of cash for buying a house or something, I don't understand why anyone would sell a SaaS at a 3x multiple. Just keep it, print money, and own the IP. Am I missing something?

"I hear the same story from founders over and over again: I had to let them go because they couldn't work with AI." That's an expensive lesson. You can screen for it in the interview. Give candidates a real task and ask them to use AI tools to complete it. Watch how they prompt, whether they verify the output, and whether they blindly accept what the model gives them. "The best candidates treat AI like a sharp but unreliable colleague: they verify, they push back, they ask for tests, they iterate." The full essay covers hiring, training, overcoming resistance, and why your VP of Engineering's AI fluency sets the ceiling for the rest of the org. movingavg.com/essays/ai-peop… #AIHiring #FounderLife #StartupOps #AIStrategy

"Disclosing your trade secrets to a free AI tool is a terrible idea. These tools are free to use your content to train their AI models — essentially you are publishing your IP in a textbook for training AI." Most companies I talk to are worried about copyright when their employees use AI. That's the wrong thing to worry about first. The real shield is trade secrets — and most teams are leaking them into free-tier tools without thinking twice. The practical fix: enterprise AI accounts with training disabled, AI IP clauses in your employment agreements, and quarterly audits of every tool touching sensitive data. "If you don't draw the line, employees will draw it for you — inconsistently." Full breakdown with a five-step checklist: movingavg.com/essays/how-sho… #AIIP #TradeSecrets #StartupLaw #FounderLife

Every week I sit in a room with founders, CTOs, and board members who are actually deploying AI — not theorizing about it. One founder's team sends 2,000 cold emails a day through an AI agent that handles the replies and books meetings — humans only step in when the conversation gets complex. An engineering director built and deployed an internal app in 5-minute chunks between meetings using Copilot. And the IP conversation keeps coming up: work-made-for-hire doctrine doesn't cover AI-generated output, which means most companies have a gap in their agreements they haven't noticed yet. Chatham House Rule. Small group, vetted seats, no vendors, no slides. We publish the takeaways as essays afterward. If you're a founder or C-level leader making real AI decisions and want a room where people talk about what's actually working — request an invite: movingavg.com/ai-roundtable.… #AIStrategy #FounderLife #ExecutiveAI #AILeadership

If your team is using AI to create content, code, or product designs, you should know: work-made-for-hire doesn't cover AI-generated output. U.S. copyright requires human authorship. Patent law requires a human inventor. If AI did the work alone, you can't protect it under either regime. The one exception: trade secrets. AI-generated output kept confidential can still qualify — no human authorship required. Most companies don't have an AI IP policy yet. They probably need one. Full breakdown of copyright, patent, and trade secret law as it applies to AI 👇 movingavg.com/essays/ai-gene… #AI #intellectualproperty #founders #startup

Your employees are generating content with AI on company time, using company tools. You probably assume you own it — the same way you own everything else they produce at work. You don't. Work-made-for-hire doctrine requires an employee or a contract. AI is neither. The D.C. Circuit confirmed in March 2025: AI cannot be an author, employee, or party to a contract under U.S. law. If nobody human contributed enough creative input, the output has no copyright at all — it's public domain. Most companies I talk to don't have an AI IP policy. They need one. The fix isn't complicated: document human contributions, treat sensitive AI output as trade secrets, and update your employee agreements. movingavg.com/essays/ai-gene… #AIIP #StartupLaw #FounderLife #AIStrategy

@einarvollset Death just got its ass kicked, no doubt.

Oh no Chuck Norris died. RIP king.

I’m hosting a roundtable at @MicroConf: 🦾Managing Cybernetic Organisms: Hiring, whether AI- or carbon-based, too often leads to procrastination (or micromanagement) rather than productivity. If you want to discuss how to hire and manage intelligent beings to help grow your organization, this table is for you. Join us? microconf.com/us-flagship

There are a crazy number of cool open source projects launching. Here's a list of some of the most popular ones. I know a lot of these lists are garbage, but this is a genuinely interesting list of projects that are going viral. And for good reason:

"If I have to tell the agent twice, I put it in the rules. That works 99% of the time." — a principal engineer at our AI roundtable. That one line captures the gap between people who find AI useful and people who find it indispensable. The difference isn't prompting skill — it's whether you close the feedback loop. Capture corrections. Build instructions files that grow over time. Make every session teach the next one. The same roundtable surfaced an organizational problem: people copy-pasting raw AI output into team messages without thinking. One director's response — "I can do that. What do I need you for then?" Full summary from roundtable session 2, shared under Chatham House rules: movingavg.com/essays/close-t… #AIStrategy #ClaudeCode #FounderLife #AIProductivity

I'm going to call this right now. We are going to have a large population with absolutely no critical thinking skills if they blindly trust AI for everything. We have all already seen it. They don't validate outputs. They don't really understand anything. They just ask questions, it looks good, and they go with it. There are going to be huge issues in every company as this continues over the years. The amount of technical debt and knowledge gaps are going to be insane. So much opportunity if you actually know what you're doing.

The more AI does for your medtech company, the less IP you actually own. Copyright requires human authorship. Patents require human inventorship. If AI generates material with insufficient human contribution, that output has no owner under copyright law. Meanwhile — an engineer pasted error messages into ChatGPT last week. A quality manager used it to draft a CAPA response. Trade secret protection didn't degrade. It evaporated. Most medtech boards haven't asked the four questions that matter. We wrote about what they are. medtechconnect.net/essays/medtech… #MedTech #AIGovernance #IntellectualProperty #HealthTech

I watch founders spend weeks building elaborate AI automations. Most of those setups are obsolete within a month. OpenAI and Anthropic are burning billions improving their $20/month tools — don't compete with that. My rule: if my skills and perspective don't have a unique-to-me impact on the work, I let an AI try it. Bad output gets thrown out. Good output earns less oversight over time. It's exactly like managing employees. The full framework — delegation, guardrails, when to automate, and deploying AI to your team: movingavg.com/essays/adoptin… #AIAdoption #FounderLife #StartupOps #ClaudeCode