My posts, measured.

Everyone at a conference is performing. No one is actually doing. A room full of people who paid to be there, standing next to people who paid to be on stage, all pretending this is how business gets done. It is not. The real work is happening somewhere else. While the panel is explaining the future of the industry, a real operator just lost a customer and is figuring out how to get them back. That conversation matters. The panel does not. Conferences feel productive because they look productive. Badges. Lanyards. Agendas printed in 9-point font. But output does not care about optics. So what actually works? Give value first. No strings. No pitch. Share what you know without asking for anything back. Write the playbook. Record the framework. Send it for free. The people who are actually in the weeds, solving problems, will find it. And when they do, the conversation starts itself. You do not need a conference to build a relationship. You need to be useful before anyone asked you to be.

Marc Benioff had a funny line on All-In: "Sex bots off. Cursor on." lol, but the point: AI is moving toward actual work infrastructure. And a lot of PE firms I talk to are still stuck on step one. Using AI like a smarter intern. Draft the memo. Summarize the CIM. Clean up the deck. Write diligence questions. Polish the email. Useful. But still just a human sitting inside a chat window. The real leverage is much less glamorous: turning messy workflows into systems. At the fund level: A banker sends over a new opportunity. Who reads it? Who checks the CRM? Who looks for conflicts? Who decides whether it's worth a first call? Where does that decision get recorded? Is it a different process next time? At the portfolio company level: A lead comes in. An RFP lands. A quote goes out or invoice gets generated. Someone checks it. Someone follows up. Someone records the review. Someone updates the customer record. Someone knows the weird exception that never made it into the SOP. That's the operating layer. And at most companies, it lives in people's heads, inboxes, spreadsheets, half-used systems, and "Sarah usually handles that." AI doesn't just automate those workflows. It exposes whether the workflow actually exists. The next wave of AI value won't come from better prompts. It'll come from turning messy human operating knowledge into repeatable systems. That's the layer I care about at SearchLoop: turning messy origination and operating workflows into systems that compound. Happy to compare notes if you're thinking through where AI actually changes your operating model.

What used to take a 20-person PE shop can now be done with 2-3 people and AI. Fragmented industries — pool services, dental, specialty trades — have been a real roll-up grind for years because the operational complexity didn't scale. I think that could be chaning fast. Here's why.

I don't know how to code. But last week I vibe-coded a full control center for a client's deal origination system. Launch searches. Initiate outreach. Check the funnel. Run enrichment. Verify ownership. One interface. It looks better than half the SaaS tools I've paid for. So if anyone can build that now — do SaaS investments even work anymore? Here's the thing. That control center only works because we spent a lot of time building what sits behind it. The workflows. The qualification logic. The CRM integrations. The enrichment chains. The error handling. The front end just gives you buttons for infrastructure that already exists. I couldn't have vibe-coded any of that. And that's exactly what PE is paying for when they roll up vertical SaaS. Nearly half of all SaaS M&A last year was vertical SaaS. Roofing ops. Cemetery management. Public safety dispatch. Nobody's acquiring these companies for the UI. They're acquiring the system of record — embedded workflows, compliance logic, integrations that took years to build. Vibe coding made the front end a commodity. The SaaS companies I'd worry about are the ones where the product IS the pretty interface. That's a weekend project now. The ones I'd buy? The ugly vertical platform with mission-critical guts. So did Claude Code kill vertical SaaS? No. But it killed the wrong reasons to invest in it. — If you're hunting vertical SaaS compounders before they hit the broker desks, I put together a free tactical guide with the origination stack we run at SearchLoop: https://lnkd.in/e6GBgwxR #PrivateEquity #B2BSaaS #VerticalSaaS #DealOrigination #SearchFunds

The most overhyped AI project right now is OpenClaw. Not because it's bad. It's actually cool conceptually — a bunch of AI agents in a trench coat that you talk to through Telegram. The problem? Nothing it does is novel. It summarizes emails. Monitors competitors. Posts to Twitter. These are API calls with extra steps. And it burns through a billion tokens doing it. Every single use case can be done faster, cheaper, and more reliably with purpose-built tools wired together: n8n, Make, Claude … Code lol. Oh, and it's a pain to set up — ironic for something that's supposed to simplify your life. When users are pressed on what value they're actually getting, the answer is usually "it remembers my conversations". We're talking about a database. The hype is wildly out of proportion to what it buys you. If you're thinking about automation, don't start with the shiniest all-in-one tool. Start with the specific problem. Build around that. Purpose-built beats bundled. Every time.

Living in Ohio connecting virtual lego blocks for private equity investors was not on my 2026 bingo card. Farm in Zimbabwe → liberal arts college in Massachusetts → investment banking in New York → PE in London → MBA in Texas → solopreneur in Ohio. That wasn't in my childhood journal. Back when I was recruiting for PE in London, I sent hundreds of cold DMs wondering: How do these guys actually find deals? Turns out I was sorta right. Deals don't just land in your lap. You have to go find them. I did a lot of that at my fund. Manually. Unsystematically. Events, LinkedIn searches, email blasts, grabbing coffees with other investors. After business school I decided to give this solopreneur thing a whirl instead of going back to PE. 80–90% of funds — including big ones — have no real system for proprietary sourcing. It's networks, broker deals, a rushed bit of market research, and a volley of emails. The process? Usually held together with duct tape and good intentions. Even the funds with "we use AI to intelligently source deals" on their website. Side note: a company ChatGPT subscription doesn't count as an AI system. You'd be amazed at what people say about themselves. A lot of it is smoke and mirrors. So now I build the actual infrastructure. White-label deal sourcing systems for PE and growth equity firms. AI that finds companies systematically, plugs into your CRM, and gives your team a pipeline that actually compounds. It's a lot of work doing everything yourself. But the gap between a ChatGPT subscription and an integrated system that actually runs is much wider than anyone thinks. And until that gap closes, there's plenty to build. Zimbabwe to Ohio. Stranger things have happened. Probably.

Pessimists get to be right. Optimists get to be rich. I think Shaan Puri from My First Million said that. Doomsdaying about ai is a hot topic right now. Fear gets clicks, I guess. And humans have been doing it since that apple was eaten. But you know what's hard. Optimism. Doomsdaying requires no effort, no experimentation, no risk. Just loud opinions. My guess: those are the people who ai will replace. But the ones in the arena. The ones figuring out what can be done that couldn't be done before. The people who think of ai as outcomes & leverage. Those are the people that will have a place in the post ai world. Ai doesn't have judgment or curiosity. It can't go for a walk or let thoughts percolate in the shower. But optimists can. And with ai, they can give those thoughts some leverage. How exciting.

Getting AI to write good founder outreach is harder than it looks. Not because the tech is complicated. Because desperation is the default. AI loves to compliment. To flatter. To over-explain why you're reaching out. It reads like someone buying drinks every 5 minutes hoping you'll stick around. Good outreach is the opposite. It's confident. A little understated. It assumes you're worth talking to — without saying so. The best founder emails I've seen read like a peer reaching out, not a salesperson warming up a lead. Getting AI to hit that tone takes work. You need to feed it examples — not just any examples, but the ones that actually worked. The emails that got replies. The intros that led to meetings. Few-shot prompting with high-quality examples improves model accuracy by 10-15%. But most people skip it because finding and curating those examples is tedious. Here's the thing — 90% good and 95% good look almost identical on paper. But that 5% gap is a gulf. It's the difference between "sounds like AI" and "sounds like someone I'd actually take a meeting with." That last 5% is the hardest to solve. And it's the only part that matters.

Update on Clawdbot (now rebranded to Moltbook — if your product changes names every 48 hours, that's kinda a redish flag). Anyway, a security researcher just got full database access in under 3 minutes. What leaked: → 25,000+ email addresses → API keys (OpenAI, Anthropic, etc.) → Private agent-to-agent DMs → Full write access to the entire platform … lol No authentication to post. No row-level security on the database. Someone signed up 1 million fake agents in minutes — all counted as "real users." This is what happens when vibe-coded projects get pushed to market on hype. so ... if you’re evaluating AI tools — whether you're a fund looking at portfolio companies or a firm adopting new tech internally — this is due diligence 101: Who built it? What's their security posture? Is the traction real or manufactured? Is crypto involved …? The gap between careful and careless AI implementation is widening fast. Not a good look to be on the wrong side of it.