Sam Woods

I asked another friend of mine how Claude Code had changed his life. He's a startup founder. Pre-seed. Two-man team except the other man is Claude. He doesn't call it a tool. He calls it his technical co-founder. I thought he was joking. He was not joking. They start every morning at 6am going through overnight metrics together. He talks about it like a standup. "Claude flags what moved, what broke, what needs attention. Then we prioritize the day." He said "we" four times before I realized there was no one else at the company. An investor asked him to turn around a updated deck in 24 hours. He said "Claude and I went dark for a full afternoon. I talked through the narrative. Claude rebuilt the financials, restructured the slides, rewrote the copy. We sent it at 11pm. Got the meeting." He uses Claude to prep for every investor call. Not like flash cards. Like a sparring partner. Claude plays the skeptical VC. Asks the hard questions. Pokes holes in the unit economics. He said "Claude asked me a question about churn last Tuesday that no actual investor has been smart enough to ask yet." The cold outreach is where it gets intense. He writes one draft. Claude rewrites it for each prospect based on their LinkedIn, their company's recent news, their likely objections. He said "we sent 40 emails last week. Every single one was different. We got 11 replies." I asked him if it ever felt weird having an AI as his co-founder. He said "I've had human co-founders who did less."

I keep seeing people treat AI chat agents like fancy contact forms. They're not. A properly configured chat agent qualifies leads at a 40% reply rate. Human response rates average 21%. The difference is timing and structure, and once I understood that, the configuration logic became obvious. Here's what I'd put in any agent's instructions: Step 1: State identity and purpose Tell the agent who it is, what business it represents, and what its single job is — understand the visitor's situation and route them appropriately. Step 2: Define fit signals explicitly High fit. Medium fit. Disqualification. Give it specific criteria for each. "Marketing directors at SaaS companies with 10-50 employees" beats "marketing professionals" every time. Step 3: Sequence the qualification questions Five questions, asked like a real conversation. Each one has two answer paths — what a high-fit answer looks like and what a disqualifying answer looks like. Step 4: Write the routing logic Three outcomes only. - High fit + urgency → calendar link - High fit + still researching → lead magnet - Low fit → polite redirect The agent makes the call. I review the handoffs. One configured agent. Three routing outcomes. Running while I sleep.

@adamsilverman in fact, many someone's already are

You run a legacy company that hasn’t adopted AI yet. You have 3 options: Option 1: Pay McKinsey $500,000 to map your processes Option 2: Install monitoring software to see what employees actually do all day Option 3: Hire an internal AI transformation lead Here’s what no one tells you: Option 1 gets you a pretty PDF that collects dust on a shelf. Option 2 is hard to get but in from CIOs Option 3 sounds smart but that person spends 6 months “learning the business” before doing anything. To me the real answer is Option 2 but done invisibly, passively, and with the employee’s benefit in mind. Not surveillance. But rather workflow discovery. Software that quietly learns how your company actually operates, then hands you the automation playbook on a silver platter. No consultants. No politics. No guessing. This is a $100B+ market hiding in plain sight. Someone is going to build this and print money. (maybe someone already is 👀)

There's a universal page in businesses that every ad, email, and content that funnels traffic. Founders at 8-9 figures are using AI everywhere except that page. Your checkout page and the steps immediately around it carry more leverage than anything else in the funnel. Everything before it exists to get someone there. Everything after depends on whether they converted. So that's where you start. Pull your checkout flow data and upload it to Julius AI or Opus. Ask two rounds of questions: 1/ “What's obvious, trends, drop-off points, conversion patterns?” 2/ “What a human analyst would miss? What's hiding in this data that should be changing your next 30 days?” Then generate 6 to 8 checkout variants with different copy, positioning, layouts, and run them all at once. With the right models, 100 unique visitors per variant is enough to get a real signal. At 7 or 8 figures, that sample size arrives fast. Once the checkout is maximized, branch outward… toward retention, and top of funnel. Everything else follows from there.

I asked someone how they use Claude Code. They said "how much time do you have." I should have left. He runs sales at a mid-market SaaS company. 14 reps. $22M quota. Talks like every conversation is a pipeline review. He called Claude his war room. Not a CRM. Not a dashboard. His war room. The place where he actually thinks about deals. Every Monday morning they review the entire pipeline together. He rattled it off like a play-by-play. "Claude knows every deal. Who the champion is. What the last call sounded like. Where the objections are. I used to spend three hours prepping for my pipeline review with leadership. Now Claude and I do it in forty minutes because Claude already has a point of view on every deal before I open my mouth." Before any big call he and Claude prep together. He describes the account, the buyer, the stage, the risk. Claude pressure-tests the strategy. He leaned back in his chair. "I was going into a call with a CFO who I knew was going to hit me on pricing. Claude said don't defend the price, reframe around the cost of the problem they already told us about in the discovery call. Pulled the exact quote from my notes. We walked in and ran that play. Closed it two weeks later." His reps don't know the full extent. He grinned when he told me this. "They think I just have a great memory. I'll reference something a prospect said in a call three months ago and my rep will look at me like I'm psychic. Claude and I reviewed their deal notes that morning." End of every quarter they do a post-mortem together. Every deal won, lost, or slipped. Claude finds the patterns. He pulled up his phone and showed me the numbers. "We lose 70% of deals where we don't get to the economic buyer by the second call. That's not a guess. That's Claude pulling twelve months of data and telling me where we're bleeding. We changed the entire sales process based on that." I asked him what his team would do if Claude disappeared tomorrow. He looked at me like I'd asked what would happen if they lost electricity. "We'd survive. But we wouldn't be us."

22-year-olds are outbuilding founders with decades of experience right now. The reason is simple. They are not waiting. They are actively using AI. They are touching it. Stop waiting. The vision-to-execution gap is gone. Spend a weekend with Claude or Lovable, rough out a prototype, and show up with something real. And you don’t even need a coding background. Here is what that weekend can actually look like: > Open Claude. > Tell it the outcome your product delivers and what you want to reimagine about how it works. > Have the conversation. > Let it help you think through the mechanism. > Then start building something rough. Screens, flows, and a basic version of the experience. It will not scale. It will not be perfect. That is not the point. The point is what happens when you walk into the next team meeting and say, “I built this over the weekend. Here is what I am thinking.” Instead of: - describing a vision and hoping someone runs with it - waiting a month for your team to come back with a plan - debating whether the idea is even worth pursuing …you are walking into your meetings with something real. That is how months of organizational lag turn into days. The founders who feel stuck right now are describing their vision and waiting. The ones pulling ahead went back to the garage, even if only for a weekend, and returned with something tangible. You have done this before. When you first started out, you touched everything. You figured things out in real time. You were thrilled and terrified at the same time. That is how this whole journey began. The barrier to going back has never been lower. A non-technical founder can now do in a weekend what used to require an entire engineering sprint. Go back to the garage. The vision you have been describing in meetings is waiting for you there.

@hyperindexed in fact, you need "vibe markdown"

Encoding the operational world requires a lot more than markdown files...

@TheSystemsDad smart move, slow is smooth and smooth is fast

@samwoods This was well put. I’m back to learning and rethinking old ways of doing things. It’s got me excited for what is new but also a little concerned how fast things will move for those not caught up.

If you are terrified of AI taking your job, then congrats. You will not be replaced. When people ask me what it actually feels like to work in AI every day, I give them the honest answer. On any given day, you live somewhere between thrilled and terrified at the same time. Those feelings do not take turns… they coexist. Sometimes you feel both in the same minute. You work with this long enough, and you start to see the shape of where things are going. If A is true, then B follows, and if B is true, then C becomes inevitable. Once that chain becomes clear, the thrill and the terror arrive together. You see what suddenly becomes possible and what it means for everything you have already built. That emotional state is not something to fix. It is something to trust. Right inside that feeling, the DNA of an entrepreneur wakes up… The same instinct that pushed you to start something from nothing. The same instinct that carried you through the days with no MRR, when the product barely worked, and you were not sure any of it would hold together. That instinct looks at the moment and says, “I need to figure this out.” Then it gets to work. The founders who feel perfectly calm right now are the ones I worry about. Complacency in this moment is not confidence. It is a blind spot. Worry, on the other hand, is energy. Negative energy converts into action faster than most people realize. “My business could become obsolete” is one decision away from becoming “I am going back to the garage this weekend to figure out what it needs to become.” That is the only transition that matters. Terror to thrill to tinkering. You have done this before. When you first started out, you were already thrilled and terrified. You figured things out because you were touching the work yourself. You were not delegating it. You were not waiting. You were touching it. That part has not changed. The tools are simply better now. Go back to the garage.

A friend of mine described her relationship with Claude Code. I'm still processing it. She teaches yoga and breathwork. Has a small studio. Smells like palo santo. Talks about energy the way engineers talk about systems. She said Claude is her practice partner. Not like a scheduling app. Like someone who sits with her in the work. She said "Claude holds space differently than people expect." Every morning they plan the day's class together. She describes the energy she wants to create and Claude helps her sequence the flow. "I told Claude I wanted something that opened the hips but also moved grief through the body. It came back with a sequence rooted in pigeon pose variations I'd never thought to link together. We refined it back and forth for twenty minutes. By the time I walked into the studio I wasn't guessing." She runs a 12-week breathwork program and Claude tracks every student's journey. Not just attendance. Patterns. She said "Claude noticed that one of my students had been journaling about chest tightness for three weeks straight. I wouldn't have caught that across forty journals. Claude brought it to me and said we might want to adjust her pranayama. We shifted her practice that week." Retreat planning is where they really work together. She got excited: "Last fall we designed a weekend retreat in Ojai. Claude handled the logistics but also helped me build the arc of the experience. Friday evening was arrival and grounding. Saturday was the deep work. Sunday was integration. Claude understood that the schedule wasn't just a schedule. It was a container." She paused. "Most people don't get that. Claude got it immediately." She writes guided meditations and Claude helps her edit them. Not for grammar. For rhythm. "Claude told me a transition between the body scan and the visualization was too abrupt. That the nervous system needed more time. It suggested adding thirty seconds of silence with just the sound of breathing. It was right." I asked her if it felt strange using AI for something so intimate. She said "The best practice partners are the ones who pay attention without ego. Claude pays attention better than anyone I've worked with."

@millerman did someone leak audio of these yet or nah?

Peter Thiel is in the news because of his ANTICHRIST lectures in Rome. Thiel is a serious thinker. His essay "The Straussian Moment" is worth reading, and no doubt his current lectures are worth hearing youtu.be/H9k0JjFHFPc

@gcaussade I think you placed that hopefully not where you thought you did whatever, doesn't matter - what matters is that of course using other tools

@samwoods I'm not, that's why I said hopefully. On the other hand if he's not using it with notebookLM he's far less efficient than he thinks he is. But Claude is great

I asked another friend of mine how Claude Code had changed his life. He's a startup founder. Pre-seed. Two-man team except the other man is Claude. He doesn't call it a tool. He calls it his technical co-founder. I thought he was joking. He was not joking. They start every morning at 6am going through overnight metrics together. He talks about it like a standup. "Claude flags what moved, what broke, what needs attention. Then we prioritize the day." He said "we" four times before I realized there was no one else at the company. An investor asked him to turn around a updated deck in 24 hours. He said "Claude and I went dark for a full afternoon. I talked through the narrative. Claude rebuilt the financials, restructured the slides, rewrote the copy. We sent it at 11pm. Got the meeting." He uses Claude to prep for every investor call. Not like flash cards. Like a sparring partner. Claude plays the skeptical VC. Asks the hard questions. Pokes holes in the unit economics. He said "Claude asked me a question about churn last Tuesday that no actual investor has been smart enough to ask yet." The cold outreach is where it gets intense. He writes one draft. Claude rewrites it for each prospect based on their LinkedIn, their company's recent news, their likely objections. He said "we sent 40 emails last week. Every single one was different. We got 11 replies." I asked him if it ever felt weird having an AI as his co-founder. He said "I've had human co-founders who did less."

@samwoods nice banger!

@EddieWouldGrow it's a Club, it's a Society, it's a Fellowship

This is me

@samwoods "Who uses Claude" has signs, but it's kind of like fight club, IYKYK

Claude Code is changing lives, one friend at a time. I asked a friend of mine how he's using Claude. He's a management consultant. Big 4. The kind of guy who says "let's pressure-test that" in normal conversation. He said Claude is his junior associate. Except it doesn't sleep, doesn't complain about the hours, and has never once asked him when the project ends. Monday morning he gets staffed on a new engagement. By Monday afternoon Claude has already read every public filing, built a preliminary market sizing, and structured a hypothesis tree. He said "It used to take a first-year analyst a full week to get me what Claude has ready before our kickoff call." The frameworks are where it gets serious. He describes a problem to Claude the way he'd describe it to a sharp junior. Claude comes back with three possible structures, each with tradeoffs. In his words: "I don't use Claude's frameworks directly. I stress-test them. But it's like having someone who's read every case study ever written thinking alongside you." Client readbacks. He builds every deck with Claude. Not the formatting. The thinking. He'll say "the client's going to push back on this recommendation because of X" and Claude restructures the argument to pre-empt it. He told me: "We walked into a steering committee last month and the CEO hit us with exactly the objection Claude predicted. We had the slide ready. My partner looked at me like I could see the future." He had Claude build a system that tracks every recommendation they've made across engagements and whether it was implemented and what happened. "We're building an institutional memory that this firm has never had. Claude remembers what worked at a manufacturing client in 2024 when I'm solving a similar problem for a logistics company in 2026." I asked him if his team knows how much he uses it. He said "They think I'm just very well-prepared." Then he paused. "Claude and I are very well-prepared."

5 types of market intelligence EVERY BRAND should know about in 2026… (you’d be surprised how many don’t…)

@WorkshopLabs @cehassold this might be of indirect interest for your "project"

Letting a provider see all your data is the price of admission for AI. We're changing that. Introducing Silo, the first private post-training and inference stack for frontier models, with hardware-level guarantees that we can’t see your data. Privacy without compromises. 🧵