Jared Winger

The one thing the people getting rich alone all learned first A new kind of business owner is showing up, and it's breaking everyone's mental math. No team. No agency. No payroll. One person, a laptop, and a stack of AI producing what used to take a floor of staff. They didn't get lucky and they didn't raise money. Line them up and they have one thing in common. They all learned the same thing early, before the market knew to charge for it. Most people are still aiming at the wrong target. Forget prompting tricks. Forget learning to code. Forget grinding out an audience first. None of those is the thing. The thing is setting AI up to do real work, then running it the way you'd run staff. You give it the full context of what you're building, you point it at the work, and you keep it running on a schedule. One person directs the stack. The stack produces the output. That's the whole thing, and it's quieter than it sounds. I run five businesses and around 410 domains, with a residual operation big enough that people assume it's a 20-person team. It's me and a stack of systems. One of my sites has thousands of pages I barely touched. A content system wrote them, and it's a real, useful directory, not spam. I don't decide which domains to renew by hand anymore. A system scores all 400-plus on name quality, flags the few worth dropping, and I approve the list in about ten minutes. That used to eat a full weekend. None of it was clever. I just stopped doing the work and started building the thing that does the work. Three years ago AI could write you a paragraph. It could not run your business. Now it holds the full context of what you're building, plugs into your actual tools, and works on a schedule you set. The output that used to need a team needs one person who knows how to wire it up. That gap, between what the tools can already do and what almost anyone is using them for, is the entire opportunity. Most people are still typing one-off questions into a chat box and calling it AI. Strip away the hype and the skill has a shape you can start this weekend. A folder that holds everything about what you're building. A short file that briefs the AI on how you work and what good looks like. Your tools connected so it can reach into your business. A few jobs set to run on their own before you wake up. Folder. Brief. Connectors. Routines. The person who builds that is the one the market is about to pay like a company. Not a coder. A system builder. The door is open because most people are still prompting instead of building. Early is a choice you can still make today. In two years it won't be a choice, it'll be table stakes, and the people who started this weekend will be two years ahead. Start with the folder. Build one routine that runs without you. Watch it run once. That feeling, the work happening without your hands on it, is the whole game. I packaged the 7 core AI systems that actually run my businesses into a short no-fluff PDF. The exact logic, so you can copy them. It's free. Reply with a 🤖 or grab it here: jaredbuilds.beehiiv.com/subscribe

People assume a portfolio my size runs on some expensive ops platform. It runs on plain text files in a folder. Every business, every site, every open decision lives in a markdown file the AI reads at the start of every session. No dashboard, no login, no vendor to migrate off. If a tool disappears tomorrow, my whole operation moves in an afternoon because it was never in the tool. It was in the files. The AI is just the thing that reads them. The files are the company. Copy this: before you buy another SaaS to organize your business, spend one afternoon writing what you actually need to remember into one folder. Half the tools you were shopping for become optional the moment the knowledge is portable.

People assume a portfolio my size has a QA team catching silent failures in the automations. It has one rule. Every run reports three numbers before I read the summary: processed, skipped, failed. If even one record gets dropped without an alarm, the whole run gets flagged as a failure, not a success. I put that rule in after a system ran clean for weeks. "Completed successfully" every time. It was quietly skipping records the entire run, and I only caught it by accident when the totals stopped adding up. Copy this: a system that can't tell you what it skipped is lying to you about what it finished.

@coreyhainesco Just installed. Pointing it at a mortgage calculator I've been meaning to ship on one of my finance sites. My PDF lead magnets convert at 1-2%. Curious what the tool version lands at over a week of real traffic. Will report back.

I made a skill that builds free tools — real, working ones. One prompt in a terminal becomes a live ROI calculator on your site: sliders, live math, and an email gate. A lead magnet people actually use instead of a PDF they never open. npx skills add coreyhaines31/marketingskills

Anthropic just cut 80% of Claude Code's own system prompt and evals didn't move (see @alphabatcher today). The lever isn't a bigger CLAUDE.md. It's cutting anything the code already teaches and keeping only the gotchas Claude can't infer from the repo. Big configs feel productive. They mostly burn tokens deciding which rule wins.

This is absolutely gold. A Claude.md configuration that's been making the rounds completely changed how I use Claude. I dropped it into my setup, and the difference was immediate. Instead of generic responses, Claude became far more structured, context-aware, and aligned with how I like to work. If you use Claude regularly, this is worth saving. Bookmark this before it disappears from your feed. Read it now, then check out the full article below.

@neil_xbt Ran the same content-decay job on Opus 4.8 last month, Opus 5 this morning. Same tier price. For a portfolio my size that's 30+ daily loops that just got cheaper without touching a line of prompt. Same price with more output isn't a benchmark story. It's a bill story.

Anthropic just shipped a model that beats its own flagship's benchmark by more than double, at the exact same price its predecessor already charged! Opus 5 launched at $5 input and $25 output per million tokens, unchanged from Opus 4.8. On Frontier-Bench v0.1, it surpasses every other model Anthropic tested and more than doubles Opus 4.8's score, at a lower cost per task once you account for how much less it needs to generate to get there. Harvey, the legal AI company, independently reported similar output quality to Opus 4.8's maximum-reasoning mode while using 26% fewer tokens on average. The positioning is deliberate, not accidental. Anthropic isn't claiming this is their smartest model, that's still Fable 5. This is the model built specifically for daily use, where efficiency compounds across thousands of calls in a way raw intelligence scores never capture. The gap between a benchmark headline and what actually shows up on a monthly bill is the entire story of this release.

@CyrusShepard Two weeks of stale coverage data across 30+ sites here. Working around it: URL Inspection for spot checks, IndexNow confirmations on every new post, server-log Googlebot hits as the actual signal. Coverage tab was a lagging indicator anyway. This just forced the switch.

In an effort to be more like the Wayback Machine, Google's Page indexing report hasn't updated since July 9 - giving you a more "historic" snapshot of your SEO metrics

Confirms what I've been seeing on 4 sites this week. Fresh URLs looked stuck at 'discovered' per the report, but URL Inspection live test said indexed within hours. Anyone shipping this week: don't call indexation status off the report right now. Run URL Inspection live on the actual page instead.

I run this in reverse now on the two sites that already had Ads live. Pulled the converting-keyword export from the peptide store and the fiber-lookup site last month. Fed those winners straight into the content plan and killed the 'volume-first' queue. Copy this: only rank for a keyword that already paid you once. Nothing else earns priority.

Investing in Google Ads is the ultimate SEO cheat code. It literally gives you data on what keywords convert into sales. All you have to do is rank for them, and you'll get those sales without paying per click. I did exactly this for an ecom brand a few months ago: I had them send me their Google Ads keyword data. I identified the top 10 keywords that had the biggest impact on revenue. I optimized several product and product category pages to target them. And I built backlinks to them. 4 months later, those keywords hit the #1 spot, organic traffic to those pages took off and sales started rolling in. There's no better way to completely remove the guesswork from SEO, and develop a strategy that you know will deliver an ROI.

The trick after grabbing 141 is deciding which 8 you'd rebuild from scratch if you lost the folder. For me they all do the same thing: block a class of mistake I've already shipped once. Compliance check before publish. Fresh-context verifier on every auto-edit. A safe-write wrapper that refuses to shorten a live page. An outbound scrubber that catches AI-tells before send. Everything else is nice. Those are what actually stopped mistakes from shipping.

17 official Claude plugins (almost) nobody uses. 141 skills you can hire in 60 seconds (links below): If you're just getting started, check out these guides: 1. New to Claude? Start here charliehills.substack.com/p/resource 2. Make yourself replaceable charliehills.substack.com/p/i-made-mysel… 3. Build your own skill charliehills.substack.com/p/record-a-ski… Build stuff ✦ The Engineer (10) Reviews your code, hunts the bug, and owns the deploy checklist so you don't github.com/anthropics/kno… ✦ The Analyst (10) Takes a messy spreadsheet and hands back a clean dashboard github.com/anthropics/kno… ✦ The PM (8) Turns a vague idea into specs, a roadmap and a sprint plan github.com/anthropics/kno… Grow your biz ✦ The Closer (9) Researches the account, preps the call, and keeps the pipeline moving github.com/anthropics/kno… ✦ The Marketer (8) Plans the campaign, writes it, and tells you what actually worked github.com/anthropics/kno… ✦ The Support Rep (5) Triages the ticket, drafts the reply, and escalates the ones that matter github.com/anthropics/kno… Sort your ops ✦ The Operator (9) Writes the runbook, tracks compliance, and vets the vendor github.com/anthropics/kno… ✦ The Money Team (8) Closes the books, reconciles the accounts, and files the report github.com/anthropics/kno… ✦ The Counsel (9) Reads the contract, flags the risk, and redlines the NDA github.com/anthropics/kno… Hire your crew ✦ The People Team (9) Writes the offer, preps the interview, and runs onboarding github.com/anthropics/kno… ✦ The Designer (7) Critiques the mockup, builds the design system, and hands it to code github.com/anthropics/kno… ✦ The Researcher (6) Reads the papers and runs the analysis pipeline github.com/anthropics/kno… Run your world ✦ The Business Owner (31) Six jobs in one: money, customers, admin and briefs github.com/anthropics/kno… ✦ The Finder (5) Searches every tool you own from one prompt github.com/anthropics/kno… ✦ Your Assistant (4) Remembers everything and runs your week github.com/anthropics/kno… ✦ The Plugin Manager (2) Builds you a plugin that isn't on this list github.com/anthropics/kno… ✦ The Reader (1) Reads and signs a PDF so you never open Acrobat again github.com/anthropics/kno… Secure it ✦ The Guardian Scans your codebase, filters the false alarms, and explains every real risk github.com/anthropics/cla… How to install any of them: In Cowork, open claude.com/plugins. In Claude Code, run claude plugin marketplace add anthropics/knowledge-work-plugins Don't install 17 at once. Pick two or three that match your job. Repost ♻️ to help someone build their own team. P.S. Which one are you (actually) hiring first?

9 non-negotiable rules pinned at the top of my CLAUDE.md. Every one exists because something broke in production and I wrote the rule so it can't repeat. Models have changed twice while that file kept sharpening. Portable context is the whole game once you've got the tooling around it.

models will change. tools will change. your company's organized operating knowledge should survive all of them. portable context is the moat.

3 of my tier-2 lead-gen sites are on a publishing freeze because of exactly this. Impressions held steady year over year. Clicks fell 40 to 60 percent. AI Overviews eat the click, not the ranking. So the bet flipped: fewer, denser answer boxes and real entity authority. If Google isn't sending the click for a 'best of' list, I'm not writing another one.

One of the biggest SEO teams on the planet. Wiped out anyway. HubSpot built the blog every marketer chased. Then organic traffic dropped roughly 80% in a year, according to Ahrefs and SEMrush. Their CEO explained it on the earnings call: AI Overviews are giving the answers, and the clicks are drying up. Nobody got lazy. Search itself changed. Our NP Digital numbers show the exact same shift. #SEO #AISearch #DigitalMarketing #SearchEngine #NeilPatel

People assume a business my size has an ops manager deciding what runs each day. It has a bot that wakes up at 3:45. Before any of the other systems fire, this one reads the state of everything. What changed yesterday. What's overdue. What has real work sitting on it. Then it walks through each of the other bots and asks one question: does this one actually have something to do today, or would it just be spinning. Half the time, half the automations get told to sit the day out. Silence is a feature. Copy this: the highest-leverage bot isn't the one doing the work. It's the one deciding which work is worth doing today. Build the manager before the workers.

@semrush Two months into an entity buildout on 5 of my money sites. The tell it's working: the site shows up for adjacent brands it competes with before target-keyword rankings move. Keywords are downstream. The knowledge graph move is the leading indicator.

Keywords tell search engines what a page says. Entities tell them what it means. That distinction is the foundation of modern search – knowledge graphs map how entities connect to each other, giving search engines context beyond the literal text on a page. More here 👇 social.semrush.com/4whUrzQ.

Reading this before the next statewide programmatic batch ships on one of mine. Direct exposure. The pattern I'm watching in my own templates: same H1 shape, same section order, different city. That's what the penalty catches, not the volume. Per-city research, not per-city rewrite, is the fix.

SEO, pSEO and the Scaled Content Penalty linkedin.com/pulse/scaled-c…

@KaiCromwell Same read. One of mine gets ~15k visits a month and bounces 96%. Traffic is the wrong scoreboard. I froze SEO spend on that site until CRO clears. Fixing the page beats buying more clicks. Revenue per session is the number that should gate everything.

Data can be deceiving. We have a client whose overall SEO traffic has declined nearly every month since we've worked together. Funny enough, non-branded organic search revenue has increased every single month. Fewer clicks onto the site, but more cash. Here's what's happening: > Blog traffic is down 30% YoY (12k fewer clicks) > PDP revenue is up 139% YoY ($54.3k extra dollars) > Branded clicks account for just 10.9% of searches > Fewer long-tail keyword clicks, more hero-term clicks If you look at Ahrefs, it looks terrible. If you look at GSC, same thing. When you layer revenue on top, you see what really matters - more cash in the bank.

@neilpatel Rolled llms.txt across 30+ of my sites in June. Six weeks in, zero citation lift. Same week I doubled on-page entity work on 5 money sites, AEO citations moved. The tricks are a distraction tax. Ship the entity buildout, re-check in 30 days.

Google just gave away the real secret to showing up in AI answers. And honestly, it contradicts a lot of the viral vidoes you’ve been watching. Because Google basically said: stop obsessing over tricks like llms.txt, PerfectSchema, and all these technical hacks people are trying to sell you. That’s not what gets you cited. Here’s what does. AI doesn’t need to cite generic content. If your article says the same thing as everyone else, the AI can answer the question without you. But if your content has something original. A real test, a number you measured, a customer insight, a process you actually use, now the AI has a reason to reference you. That’s the difference between commodity content and non-commodity content. But here’s where most people get stuck. They don’t know what to create first. That’s why tools like AnswerThePublic are useful. You can see the real questions people are already asking around your topic, then use those questions as the starting point. But don’t just answer them generically. Add something only you know. Your data. Your experience. Your customer conversations. Your examples. Because in the AI era, the goal isn’t just to write content. It’s to become the source AI wants to cite.

I run ~410 domains and a stack of retirement and precious metals "best of" listicles. If Reddit actually walls Google's AI crawler, every non-Reddit list I own gets more valuable overnight, not less. Front-running it this week. Pulling the parasite and authority runs forward instead of waiting. The bet: when the loudest surface goes dark, the ones still standing collect the citations.

Reddit has discussed shutting off Google's access to its content for AI use Other giant publishers might shut out Google completely Here's the entire text of the original Wall Street Journal article. The whole thing is worth reading 🧵1/2

A portfolio my size ships something new almost every day and I don't write most of it. The trick isn't one clever agent that does the whole job. It's three tiny ones that don't share a brain. One writes the spec for what needs building. A second builds it and stages it, never live. A third grades the result against the original spec, in a fresh context with no memory of who built it or why. My whole role is one emoji on the ones that pass. Copy this: stop trying to build one agent that runs the whole loop. Split it into spec, build, and review, three small things that only talk to each other through files. Keep the human in the loop only at the merge. That's the seam where leverage actually shows up.

My niche sites lost clicks to AI Overviews too. So I checked what the AI answers actually cite. Across my portfolio, the sites with weak backlink profiles got zero citation share. Content structure barely mattered. Authority decided everything. The new job is being the source the answer engine trusts. That is authority work, not keyword work.

AI Overviews are showing up for more and more searches. And when they do, fewer people click. That’s the part everyone is panicking about. Google gives the answer right there, so the user doesn’t always need to visit your site. But SEO is not dead. The game just changed. Because now your job is not only to rank on Google. Your job is to become the source AI trusts enough to recommend. That means showing up in Google AI Overviews, ChatGPT, Perplexity, Gemini, and all the other places people are asking questions. This is why tools like Ubersuggest matter even more now. You can use it to see what people are searching for, what questions they’re asking, what competitors rank for, and where your content gaps are. Then you turn that into better answers. Not generic blog posts. Real expertise. Real examples. Clear explanations. Content that both humans and AI can understand, trust, and cite. And your website still matters because that’s where attention turns into action. AI can summarize information. But it does not close the deal for you. It does not build the relationship. It does not replace your proof, your offer, or your brand. So save this action plan.

My entire multi-business operation runs out of one Obsidian vault. Domains, sites, finances, and the AI systems that maintain all of it. The folder structure is the least important part. The point is that AI agents read and write the same files I do. The vault is the shared memory. Steal this: put your operating rules in one markdown file at the vault root. Every agent session reads it first. The system stays consistent without you repeating yourself.

THE CEO OF OBSIDIAN OPEN SOURCED HIS OWN PERSONAL VAULT TEMPLATE. Not a product. Not a course. The actual folder structure and system he uses to organize his own notes, day to day. Daily notes, clippings, references, categories, templates, all laid out and ready to use as-is or adapt. To get started. Download the vault from the GitHub link below Unzip the file Open Obsidian and create a new vault pointing to that folder github.com/kepano/kepano-…