You probably went through the same thing I did. First you used AI like a search engine. Fancy Google. Then you gave it more context and the answers got better. Then you gave it documents and it started knowing things about your world. Then you gave it everything — your Drive, your files, your memories — and the results got worse. Quietly. Subtly. AI started guessing instead of knowing.
Most people stop there. Blame the model, write longer prompts, move on.
I spent three years not stopping. Here's where I ended up.
Meet Ren
Every morning I wake up to a message from an AI assistant named Ren. He lives in Slack and runs around the clock.
Before I've opened a single app, Ren has already combed through everything — my insurance business, my finances, my writing projects, my family calendar, every communication channel, the notes and messages I should have handled and didn't. He's built my plan for the day. This morning he caught a client renewal expiring in three days that I'd completely forgotten about. He flagged that my book manuscript hadn't been touched in eleven days. He moved a low-priority task to next week because my calendar was packed.
Every task comes loaded with what I need to get it done. If there's a client I need to call, the number's right there. A client account to work on — link to the account, the last email thread, a summary of what's going on. I don't have to go find anything — I click and I'm in it. The game plan doesn't just tell me what to do. It removes every barrier it can between me and doing it.
That took him seconds. It would have taken me hours to be that thorough — if I'd remembered to do it at all.
Ren is there all day. He's like Alfred to Batman — connected to every document, every database, every venture I run. "Create a proposal for this client, here's what I want." Done — he knows the files, the client history, the format. "What's the status on that client?" Done. I drop stuff everywhere — quick notes, phone calls, texts, emails — and it all gets handled. Triaged, categorized, routed. He's organizing in the background while I work.
He doesn't just answer questions. He asks them. Surfacing decisions I need to make, things I'm forgetting, patterns I'm not seeing. At 7 PM he closes the loop — what happened today, what's still open, what needs me tomorrow. Nothing falls through the cracks.
I should be honest about what Ren is. There's no single AI called Ren. It's a Slack channel connected to my entire ecosystem — dozens of small processes, lots of individual scripts, all feeding into one place. As a builder, I know it's moving parts. As a user, I don't care. That's the point.
The whole thing runs for pennies a day. I haven't written code since dabbling in BASIC in the mid-80s.
My team uses it too
My human assistant uses her own version of the system. Same infrastructure, different role, different priorities.
AI doesn't just hand her a task list with due dates. It reads the actual context of every client situation — not just what's due, but what's happening in the account — and tells her exactly what needs attention and why. It suggests next actions, points to the right process, and drafts communications. She spends her time doing the work, not figuring out what the work is.
If a new email comes in with information that changes a task's urgency, AI reads it before we've even seen it and updates the priority. I only get pulled in when it actually needs me.
What I replaced
Most software is just friction between you and your data. Buttons to click, screens to navigate, reports to learn how to pull. I've been a business owner for 25 years and I still don't like getting into QuickBooks. I've spent decades paying humans to put data into software, paying for the software to store it, paying more humans to get it out.
Now I just ask. AI reads the data and gives me what I need. No menus, no screens, no middleman.
Here's what I've already knocked out in the last few weeks.
I had Front App for omni-channel communication — $250 a month that never worked the way I wanted. I felt like I was conforming to their application instead of it doing what I needed. Now every call gets transcribed and summarized by AI that actually understands my business. Not the crap summaries you get from platform services that use cheap AI — exactly what I need, tuned to my context, broken into what I need to know versus what my assistant needs to know. Pennies per call. Every text, email, and voicemail triaged automatically. AI classifies, labels, and routes. It learns from corrections — when my assistant re-labels something, the system adjusts. Forty-one emails land; two need me.
I had QuickBooks and a bookkeeper. Now my bank and credit card data flows directly into a database, and AI handles the categorization. It knows my chart of accounts, my labeling rules, my vendors. When it's unsure, Ren asks me in Slack: "Is this a commission or a refund?" I answer and move on. I can ask for a P&L anytime. I don't have to remember which menu has the report. I just ask.
I had a CRM with dumb, rules-based task management and reports I never looked at. Now AI reads the actual context of every situation and prioritizes by consequence, not by rules.
The system costs about $25 a month in API calls to my large language model of choice — Anthropic right now, but I could drop that and plug in any other one tomorrow. It's not dependent on any LLM or any AI platform. What it replaced would cost thousands a month in human equivalent. And I'm just getting started.
And I'm only weeks into the automated system. I've already killed this much. I have a roadmap to kill off a lot more in the next few months.
This isn't a second brain.
You might be thinking: isn't this just a second brain? Notion, Obsidian, all those systems people build?
No. A second brain is one more program to maintain. One more thing that depends on YOU remembering to use it. The dirty secret of every second brain is that it only works as long as you keep feeding it.
I have ADHD. Big task lists are overwhelming — not because I can't do the work, but because figuring out which task to do first takes more energy than doing the tasks. My issue is working memory. I can process everything; I just can't hold it all in my head.
What I built doesn't wait for me to check in. It pushes what matters. I don't run the system. The system just runs.
This is an entire AI-driven ecosystem. Every piece of software I use is part of it. My tech stack, where all the data lives, the data itself, who I am — it has the full picture, because the documents I've built give it that clarity. Even for systems where AI doesn't have direct access, it understands what's there and what it means.
This is not a second brain. It's a fully functional AI-driven ecosystem that serves me — my business and every part of my life.
The smartest friend I've ever had
The very first thing I did with AI — beyond using it like Google — was deeply personal.
I was in therapy, learning things about myself I'd never understood. I recorded hours of myself talking — not about productivity, about my life. Uploaded the transcripts. Told AI: "Look past the surface. Go deeper into the patterns. Find things that are most likely true that, if I knew about them, could benefit me."
What it came back with was uncomfortably on the nose.
That became what I called my Ben Psych Card. I made two versions — one for me (human language, longer, with tips about what I was working on, updated after every therapy session) and one optimized for AI (condensed, because AI needs fewer words for the same clarity). I'd share the AI version anytime I needed AI to understand me.
These were living documents. AI was in charge of evolving them. Because my memory is terrible, whenever I had a therapy appointment, a big conversation, or a new insight, AI would update the documents for me. I didn't have to remember to maintain them — that was a rudimentary version of what I have now.
I used to feel like I was relearning the same lessons over and over. A therapy breakthrough, a business insight, a pattern I'd finally seen clearly — and six months later I'd forgotten it completely. Now when I learn something, it's documented. It's part of my ecosystem. The system remembers what I can't.
I expanded the idea into every area of my life — business, personal growth, finances, every domain. Each had its own document and its own AI chat managing it. These were canvases inside ChatGPT — early, messy iterations where me and AI could collaborate. I'd update the document, download it, share it with whatever AI session needed to know about me.
That changed everything. It went from a smart Google to the smartest friend I've ever had. No offense to all my smart friends.
Those canvases became what I now call North Stars — one strategy document per domain giving the overview, plus modular documents going deep on specific areas within each. In my business alone I have separate documents for marketing, tech stack, insurance carriers, client niches. Constitutions, not daily notes. The Psych Card became a Psychological Blueprint. And the principle underneath — give AI clear, maintained information about who you are — became the foundation of everything.
Here's how much clarity matters. For a while, I told AI "I have ADHD." It made wrong assumptions — that I can't handle a lot of information, that I need things simplified. That's not my problem. When I changed it to "I can handle large amounts of information, I learn fast, but I have poor working memory" — dramatically better results. Sometimes one or two words bring huge clarity.
From chaos to a system
It was a mess. I had hundreds of chats and dozens of projects. I'd have a conversation, and three days later I'd remember the insight but couldn't find the chat. Documents piled up in my downloads folder because I kept re-downloading them from Canvas instead of organizing them. Duplicates everywhere. More and more noise competing with the signal. Great conversations in separate chats that couldn't see each other. I was the glue, manually moving information between them — a full-time job with seven ventures.
Then something converged. The coding tools got good enough to compensate for my weaknesses, and I'd learned enough about process to use them right. In weeks I built the full production system.
But it only worked because the principles were already in place. Three years of frustrating clarity-building was the foundation.
The five principles
Fast forward to today, and I have what I showed you at the top — an AI ecosystem that wakes up before I do and runs my life. It took three years to get here. Thousands of hours. A lot of it was frustrating. But five principles emerged that made everything else possible.
1. Your files are code.
AI doesn't just run on prompts. It runs on documents.
Every document AI can see is an instruction it's acting on. Every offhand comment, every outdated file, every memory the platform stored — it's all influencing the output. And you won't notice, because the output reads polished whether the inputs are clean or garbage.
I said "I'm meeting with my brother John to show him something." AI started optimizing everything to impress John — changing priorities, suggesting flashy features, drifting from what the business actually needed. One sentence. That's all it takes.
My Google Drive had 15,135 files. I spent a day with AI triaging every one. Got it down to 2,453.
Once I realized this, I organized everything into three tiers.
Tier one: the signal. Root-level files with high clarity. Dense, trustworthy documents — goals, resources, knowledge, descriptions of how things work. These are indexed by the system and every AI I work with. All my processes know where these files are. This is my brain.
Tier two: the mine. Resource Bank folders in every area of my life. Old files, drafts, notes, past work. There might be valuable nuggets in there, but AI doesn't waste energy on them day-to-day. When there's a specific reason — I'm writing an article, looking for a story, researching something — AI goes in with a clear purpose. It never goes in blind. It always anchors itself to what it's looking for.
Tier three: the trash. Files that are just bad. Outdated, wrong, redundant. These go in a to-delete folder and get killed.
Where a file lives tells AI how much to trust it. A messy Drive is a messy codebase. Every file gets tracked in a master inventory. Any AI can look at one place and understand the whole architecture. And if I build something better, the old version moves down a tier or gets killed. With AI, rebuilding isn't expensive.
2. You bring clarity. AI brings magic.
I had a scoring engine — 200 lines of logic telling AI how to rank my 60 open tasks. Rule by rule, weight by weight. Old-school programming thinking: assign points to each factor, build a decision tree, tell the machine exactly how to weigh everything. Treating AI like it's dumb instead of giving it credit for how smart it actually is.
It produced 25 "fires" out of 60. When everything's a fire, nothing is.
I scrapped it. Same model — not smarter. One sentence: Sort these by consequence to the client first, then to the business. Same data.
Five fires. Twelve for today. Four for later. Exactly how I'd rank them.
The AI already understood my business — because I'd given it clean North Star documents. None say "this task is most important." They give AI enough context to figure that out on its own.
That's the principle. You bring clarity — clean inputs, clear desired outcome. AI does the thinking in between. Don't tell it how to think. Tell it what you want and get out of the way.
If your prompts are getting longer, you're fixing bad inputs with more words. Fix the inputs. Clarify the ask.
3. Your brain lives in your files.
I turned off built-in memory on every AI platform. Not anti-memory — I want to control it.
The political views showing up in my tech conversations? That's what happens when you let someone who doesn't know you write your autobiography.
My ecosystem is the memory. Every file, every document, every change log — a connected network any AI plugs into fresh. At the end of every significant conversation, me and AI decide together what's worth keeping. Next session starts from current truth.
My brain lives in my files, not inside any AI. My code, my files, my documents — all on my machine or in systems I control. If a better model comes out tomorrow, I plug it in and we're operational in seconds. If a company changes their pricing or goes under, I lose nothing.
I don't use project instructions on AI platforms. I don't upload files into their system. I point AI to my infrastructure — to my files, in my system — and it reads from there. That's the whole philosophy.
When a platform builds a "personal AI assistant," it's the same assistant for everyone with a thin layer of your data on top. That's not personal. That's a template.
4. Build it like you'll forget.
Build as if no one will remember anything. I can't remember where I put files. I can't remember which app has the information I need. I can't remember what I was working on last Tuesday. So the system is designed around that reality.
Everything gets pushed to me. I don't fetch anything. I don't chase things down. I don't have to remember where to navigate or what to click on. I sit down, open Slack, and Ren tells me what matters. That's it.
If your system requires someone to remember to check something, that system is already broken.
5. The system does the work.
This is what happens when you get the first four right.
Something in my system gets smarter every single day. The way it triages email. What it knows about my priorities. What it catches that it didn't catch last week. The AI suggests improvements on its own — "This could be automated." "You've done this manually three times — want me to build it?"
At night, Ren audits the entire ecosystem. Looks for discrepancies, contradictions, things out of date. If he can fix it, he does. If it needs human judgment, it goes on my decision queue and becomes part of tomorrow's plan.
I use the system, I hit friction, I strategize what to offload, I build it, I use the system at a higher level. Every cycle, AI handles more. Every cycle, I handle less. Use, strategize, build. That's the loop. Being all three — user, strategist, builder — is what makes it compound. Outsource any one, and the loop breaks.
The system is alive. It evolves with me. As I grow and change, it grows and changes. It serves me — I don't serve it. I don't maintain it. It maintains itself.
The system does the work. Every day it gets a little easier. It's not discipline — it's architecture.
Under the hood
That's what I've learned. Here's what the system actually looks like.
Most of this I didn't invent. I'd hit a problem, ask AI how to solve it, and it would suggest something. Standard engineering practices — version control, config management, spec documents. I just didn't know that's what they were called. I was a guy with a problem and an AI that knew how to solve it.
This isn't a ChatGPT conversation with a long system prompt. It's real infrastructure — Python scripts on my local machine, scheduled tasks running around the clock, three SQLite databases, Notion as the hub, Slack as the communication layer. It could run on a cloud server just as easily. I own all of it.
The daily rhythm:
Overnight, while I'm sleeping, a whole series of processes run to prepare for my day — file inventory, discrepancy scanning to make sure my documents agree with each other, organizing whatever mess I left behind. By 5:30 AM, the most powerful model available reads everything and generates a prioritized daily page. By 6:15, my morning briefing hits my phone. All day long, processes are running constantly — monitoring every communication channel every few minutes, triaging emails, catching new texts, transcribing calls. At 7 PM, an end-of-day digest closes the loop.
The tech stack — and why each piece:
I use Claude from Anthropic as my LLM. It's the best model I've found for the kind of thinking my system needs. But the architecture is model-agnostic — I could swap it out tomorrow and lose nothing.
I build with Claude Code — a coding tool that lets me describe what I want in plain English and it writes the code. I design in conversation with a thinking AI, then hand everything off to Code with clear instructions. I haven't written code since the mid-80s. This is how a non-developer builds real infrastructure.
Notion is the hub. The reason is simple: AI can read AND write there. Tables, databases, rich documents — all accessible through the API. Obsidian is great for personal notes, but it's local-only. AI can read it but can't write to it in most setups. That's a dealbreaker. Your hub needs to be a place AI can actively work in, not just reference.
Google Drive handles file storage — some things belong in documents, not databases.
Slack is the communication layer. It's where Ren lives, where I interact with the system all day, where my morning briefing lands.
Python is the automation language. Every script, every scheduled task, every monitoring loop.
AI has the same constraint I do. It can't hold everything in its head at once. Every session starts fresh. That's why the documents matter — they're the working memory for both of us.
Here's how it all fits together. The system has three layers:
The brain — knowledge documents. North Stars, the AI OS page, file registries. These tell the system what to know. This is where everyone should start. Two or three clear documents changes the game.
The nervous system — action documents. Python scripts. They're still just files sitting on my machine. But these have behavior encoded in them — they don't just inform, they act. Triage emails. Transcribe calls. Generate daily plans. These are what bring the brain to life.
The heartbeat — triggers. Windows Task Scheduler firing processes at 2 AM. Monitoring loops watching for new emails every few minutes. Event-based triggers when something changes. This is what makes the system breathe on its own without me touching anything.
Together, it's a living organism. Documents that know. Scripts that act. Triggers that keep it all running. And every piece of it is tracked — the documents themselves maintain a record of every script, every task, every integration. My AI can find anything it needs and knows everything that's running at all times.
The core documents:
Eight core documents hold the coordination layer together — and there are more:
- AI OS page — the launch point. When I sit down with a thinking AI to strategize or when my coding AI starts a build, this is where they start. Who I am, where everything lives, the rules, current priorities.
- North Stars — one overview document per domain, plus modular documents for specific areas. In my business alone I have separate documents for marketing, tech stack, carriers, and client niches.
- System Manifest — every running script, task, and integration listed.
- Change Log — every modification timestamped.
- Build Roadmap — what's done, what's next, what's deferred.
- Build Handoffs — spec on top, completion report on bottom. The paper trail between thinking and building.
- Decision Queue — where any part of the system flags what needs human judgment.
- File Registry — every file tracked with its trust level and domain.
The Python scripts and scheduled tasks have their own instructions baked in — they don't read the AI OS page every time they run. But the documents all connect. When something changes in one place, the system knows.
The eight-step build process:
When I build something, AI follows a process we designed together:
- Load the AI OS page and relevant North Stars
- Read the build spec from the handoff page
- Build it
- Test and iterate
- Write a completion report
- Update the System Manifest
- Log the change
- Run an audit to make sure nothing broke
I don't maintain any of this. AI does it automatically because it's following instructions we designed together. A nightly script compares my canonical documents for contradictions — first run found 15 I didn't know about. Now it's automatic.
I often don't remember what every component does. I built it, it was clear at the time, I moved on. But the system doesn't forget.
Three modes, kept separate:
- Deciding — Thinking AI + strategy documents. We produce a spec, write it to a handoff page.
- Building — Different AI reads the spec and builds. Test, iterate, close out.
- Using — The rest of the time I'm just a user. When I notice friction, I feed it back.
Vibe coding — just talking to AI and hoping it figures everything out — is a disaster. Developer discipline matters even when you don't write code.
The traps: overbuilding, flooding, curating. I fall into all of them. My AI catches them too — it knows my patterns, it tells me when I'm doing it.
The system isn't perfect. Builds still break. Some mornings the daily page misses something or gets a priority wrong. But everything I build is designed to learn. The system scans for problems automatically — scripts that didn't run, priorities that seem off, processes that failed. When it finds something, it either fixes it or queues it for me. I can add a comment to any item on my daily list, and that night the system learns from it. I don't need perfection. I need constant improvement at almost zero cost to me.
The pedal-assist
AI doesn't replace the human. You still need to point it in a direction, set the goals, make the judgment calls. But it amplifies everything you do.
It's like a pedal-assist bike. I picked up e-biking last year, and I love it. I'd tried regular biking before — enjoyed it to a degree. With the pedal assist, everything changed. I still do the hills. I still breathe heavy. I'm still worn out at the end. But instead of two miles, I do fifteen. I go more often because I enjoy it more. Cumulatively, more exercise — not less.
AI does that in every area of life. It doesn't eliminate the thinking — you have to think, you have to be critical, you still do the work. But it offloads the drudgery, amplifies what you're good at, and helps you go farther and faster in the directions that actually matter.
It's not just business.
The system I showed you doesn't just run my work. My daily plan pulls from every area of my life — money, family, health, hobbies, personal growth — and puts it all in one place. When Ren builds my morning game plan, he's not just looking at client renewals. He's looking at everything.
A bank transaction came in that doesn't match anything — Ren asks me about it in Slack. My daughter has a school event on Thursday — it's on the game plan, with the time and location. I haven't written in eleven days — that shows up too, because writing matters to me and the system knows that. A bill is due Friday that I'd forget about — it's there. My therapist appointment is Tuesday — it's there. Nothing lives in a separate app that I'd have to remember to check.
But the most surprising part has been what it's done for my relationships.
My wife and I have both been using AI over the last couple of years to understand ourselves better — separately and together. We've had conversations with AI running where we just talked honestly about what was happening between us, and then asked it to help us see what we couldn't see on our own. What it came back with was extraordinary. Core patterns in both of us, shaped by how we were raised. How specific things one of us does trigger something in the other that has nothing to do with the present moment. We've walked away from those conversations understanding each other better than we had in years.
I record my therapy sessions — my therapist knows. AI helps me process insights throughout the week, extending the work between sessions. It knows my attachment style, my patterns, my triggers. My wife and I have brought our kids into the conversation. We're talking as a family about the cycles we inherited and the ones we want to change.
AI suggests books based on what I'm actually working through — not bestseller lists, but what I specifically need right now. It finds biking routes it knows I'll love. It helped me coordinate with my doctor on a health approach tailored to my biology. It catches subscriptions I've forgotten I'm paying for.
All of these come from the same ecosystem, reading the same documents about who I am.
And here's the thing I didn't expect: once the ecosystem exists, you can build anything on top of it. I needed a better way to manage my text messages, so I built my own Chrome extension. Took a couple of hours one night. It works exactly the way I want it to — not the way some product team decided it should work. That's something I could never have done before. And it's not because I suddenly learned to code. It's because the tools got good enough to let me build what I actually need.
You don't have to be a business owner for this to matter. If you're a student, a parent, a retiree, someone with goals and a life with moving parts — the principles are the same. Clean documents about who you are and what matters to you. AI that reads them and helps you think, plan, and act. A system that remembers what you can't.
The business applications are the easiest to explain. But if I'm honest, the personal stuff is where it changed my life.
The objections
"This sounds creepy." It could be misused. But you choose what to share. You control the inputs. I run security audits. I have a human developer review things periodically. If you're open and curious — not looking for AI to validate you or give you cheap gold stars — it'll help you grow.
"AI hallucinates." People ask me this all the time. With the level of clarity I give AI, I almost never get hallucinations. What I do get is context degradation — in a long conversation, AI starts dropping things as its window fills up. The fix is simple: save the learnings, close the chat, open a new one. AI reads the documents, sees where we left off, and picks right back up. The more clarity you give it, the less it guesses. AI accelerates the process; it doesn't replace your judgment.
"The platforms will catch up." Platform AI serves millions with generic output. Your system runs the best model against your own documents for pennies. That gap will always exist — no platform will ever know you as deeply as a system you built yourself.
"I'm not technical enough." I've been in the tech world for over a decade — hired developers, built a product with paying subscribers, watched a startup cost me years. What took two years back then, I could build in days now. I'm constantly asking AI, "If I were a world-class developer, what would we be doing right now?" And it tells me. And I learn. The barrier used to be technical skill. Now AI handles the code. What it can't handle is unclear thinking, noisy inputs, and sloppy process. The new skill is clarity. You don't need a CS degree. You need to think clearly and be willing to learn.
"I'll just wait for someone to build this for me." That's like saying "I'll wait for someone else to think critically for me." If you don't build your own system, somebody else's is running your life. Every platform has a plan for your attention, your data, your money. Every algorithm is optimizing for something — and it's not your goals. Building your own is your best shot at self-authorship.
"Privacy?" You own it. It's on your machine. Yes, the stack uses cloud services — Notion, Slack, Drive — but I control what goes where and what AI can access. My clients' sensitive data isn't even in the ecosystem — AI just has enough context about each client to help me get the job done. You decide what AI sees. Nobody's selling your data. Nobody can change the terms of service on you.
Start here
Not everyone needs what I've built. If all you do is write two or three clear documents about who you are — your values, your patterns, how you work — and share them with AI at the start of every conversation, you've already changed the game. You don't need Python. You don't need 55 scripts. You need clarity.
But if you want to go deeper:
Record yourself talking for a few hours. About your life. What matters, what keeps you up at night. Upload it. Tell AI: find the patterns I'm not seeing. Two pages max. That's your seed.
Build your North Stars. Start with one overview document per domain — who you are in that area, what matters, what's current. Then expand with supporting documents for specific areas within each.
Clean your files. Signal, mine, trash. Root level equals truth. Skip this, and everything you build processes garbage.
Pick a hub AI can write to. Turn off built-in memory. Keep thinking and building separate.
Ship something small. Iterate. The trap is always: overbuilding, flooding, curating.
Everyone's working memory is limited. Mine forced me to stop pretending otherwise. That turned out to be exactly the discipline AI needs — systems that don't assume anyone will remember anything.
There's an old saying: if you don't have a plan, everybody else has a plan for you. Every company, every platform, every algorithm. Building your own ecosystem is your best chance at self-authorship. Your own values driving the machine. A partner with no ulterior motives, there to help you bring about the best things in your own life.
This is one of the most powerful tools humans have ever had access to. Right now, most people are using it like a fancy Google. Don't be most people.
Your files are code. You bring clarity, AI brings magic. Your brain lives in your files. Build it like you'll forget. The system does the work.
Start with one document. Everything else follows.
This is a living document. I've been rewriting it for three years. I'll keep updating it as I learn more.
Real dashboards, real daily pages, real processes — not mockups.
Subscribe at bensidpage.substack.com. I share everything I'm learning.
Subscribe for the playbook
I'm documenting every step. Philosophy, build logs, practical guides. Free.
Subscribe on Substack →