Martijn Verbove @verbove

Martijn Verbove@verboveEngineering · #1

I’m turning my bookmark bar into a personal toolbox. Bookmarklets are just JavaScript you save as a bookmark. Click one and it runs locally on the page you’re on. No installs, no server, no data leaving your machine. Paradigm shift: your browser isn’t a viewer. It’s a programmable runtime. First two I built: • Remove spam followers on X • 1-click bulk delete LinkedIn invites from recruiter/agency spam Build yours: 1. Open the page you want to automate 2. Tell Claude what you want to automate 3. Discuss the UI you want (buttons/filters/inputs) 4. Ask Claude to inject that UI into the page 5. Drag the “Save to bookmarks” link/button from that UI to your toolbar (now it’s a 1-click tool) Bookmark bar = workflow layer.

Bookmarklet tutorial with concrete steps

Martijn Verbove@verboveEngineering · #2

I’m converting every MCP in my stack into a CLI. Not because MCP is useless, but because it’s the wrong abstraction for day to day agent execution. Agents are processes. Processes already have the best interface we’ve ever invented: the command line. So instead of loading huge JSON specs and tool schemas, I’m wrapping MCP servers behind small, composable commands. This project nails the idea: github >steipete/mcporter Why CLI wins in practice: • tiny surface area (help text vs massive specs) • composable (pipe, chain, script) • debuggable (exit codes, logs, retries) • versioned (your repo owns the interface) • works with any model, any agent runtime My approach now: keep “skills” as the thin reasoning layer, and use CLIs for execution. If you’re building agent systems, I’d seriously consider: MCP -> CLI wrappers -> skills on top.

Architectural pattern: CLI vs MCP abstraction

Martijn Verbove@verboveProduct · #1

I've been building something for myself called Memex. It's a native macOS app that basically replaces Finder for my personal knowledge base, with an AI agent doing the librarian work. The idea is stolen from Karpathy's LLM-wiki pattern. You don't organize your notes yourself. You just dump raw stuff in a vault: PDFs, screenshots, web captures, whole conversations, entire folders. An LLM maintains the wiki for you. It writes the topic pages, links things together, merges duplicates, and cites every claim back to its source. Under the hood it's deliberately boring. The vault is just markdown on disk. Raw sources never change, and the wiki on top is plain files with wikilinks. No database, no lock-in. The agent is Claude Code running headless inside the vault. Ask it a question and you get an answer with citations. Give it a command and it creates and reorganizes pages while you watch it work in the chat rail. The best part: feeding it is automatic. Memex runs a local interceptor on your Mac that catches your AI conversations. Every Claude Code session gets forwarded the moment it ends, and Claude and ChatGPT chats flow in through the same endpoint. Talk to an AI about something and it's in your wiki minutes later. It also pulls data in through MCPs on a schedule. Gmail, meeting transcripts, whatever has an MCP server. You set the cadence, the agent syncs it in and works it into the wiki. Every source gets a number when it comes in. Every wiki page ends with a table of its sources, and every claim has a citation you can click through to the original file. So the wiki keeps evolving but you never lose track of where anything came from. If anyone wants to try it, comment or DM me.

AI knowledge base system architecture explained

Martijn Verbove@verboveAI · #2

You built a graph from your data. Now what? Plain RAG has one knob: embed → similarity → top-k. Same recipe for every question. Graph RAG has three tools, and an LLM that picks the right one before any retrieval happens: → Global · summarize the whole graph → Local · find the entity, walk the neighborhood → Cypher · write a real database query The trick isn't the search. It's the router that decides how to look. 4 slides ↓

Graph RAG retrieval pattern with routing logic

Martijn Verbove@verboveLessons · #1

Someone vibe coded a $24K/year Bloomberg Terminal in 3 hours. 80,000 lines. Real-time dashboards, FX charts, earnings summaries, the works. None of it worked. Data wasn't normalized, tickers didn't match across exchanges. This is the part people keep missing. AI made building trivial. But the value was never in the code — it's in the data, the integrations, the boring stuff nobody wants to vibe code. Bloomberg isn't $24K because the UI is hard to build. It's $24K because the data pipeline behind it took decades.

Insight on value creation vs code complexity

Martijn Verbove@verboveEngineering · #3

We run a dev team of 5 that ships faster than most teams of 50. The secret isn't the engineers. It's the workflow: 1. Ideas go into Telegram/Slack 2. AI searches the codebase and drafts tickets with full context 3. Engineers get perfectly scoped work — zero ambiguity 4. AI pair programming + AI code review 5. Ship same day The bottleneck was never talent. It was process. I'm going to help 2 product owners set this up for free. Comment "WORKFLOW" below and I'll reach out.

Process improvements for faster shipping team

Martijn Verbove@verboveWorkflows · #1

That workflow idea you've been talking about for years? Build it. This week. You can now ship agentic workflows without a technical cofounder or a dev team. Setup Claude Code or try Replit and describe the outcome in plain language, it builds the first draft, then you iterate. But here’s the real unlock: Stop thinking “app”. Start thinking “workflow”. The best “app” for most people is: • a pipeline • a set of triggers • a few screens • a couple integrations • an agent that does the boring parts for you If you can explain your job, you can build your leverage. How to do it: 1. Start with the workflow, not features Write: “When X happens, do Y, then notify Z.” 2. Pick the first painful use case One thing you do every week that feels like admin. 3. Build a rough v0 in one prompt Inputs, steps, outputs. Get something running fast. 4. Keep it tight One screen. One flow. One win. Avoid “platform thinking”. 5. Make the agent do the repetitive steps Summarize, categorize, draft, route, follow up. 6. Add integrations last Email, Slack, Sheets, CRM. Only after the core loop works. 7. Validate early Send it to 3 people who have the same problem. Watch where it breaks. 8. Iterate in public Ship improvements weekly. Momentum beats perfection. 9. Turn it into a product Once it saves real time, package it as a template or SaaS. 10. Keep going Maintenance is just prompts + small fixes now. AI tools like Replit/Claude don’t just help you “code”. They let you build workflows that compound. So yes, build that idea. But build it as a system that does work for you.

Detailed methodology for building workflow-based systems

Martijn Verbove@verboveLessons · #2

This is what a resume looks like in 2026. Not a PDF. A folder. Every skill you bring to the table, quantified. Actual executable skill files that show what you can do and how well you do it. Top performers will have skill sets that are quantifiable, available on demand, and directly tied to compensation. Better start building your skills. Literally.

Framework for quantifiable skill-based resumes

Martijn Verbove@verboveProduct · #2

I just launched t.co/HZJEqpxygd to rank products and builders by the actual work done, read straight off GitHub. no upvotes. no launch day. the hardest workers, on top. Inspired by @jonathan_wilke See how you rank 👇

Teaches ranking system concept versus hype metrics

Martijn Verbove@verboveWorkflows · #2

Just automated my entire bookkeeping pipeline with an AI agent + Xero API. Bank statement drops in Telegram → agent parses it → generates CSV → uploads via Puppeteer → creates matching transactions → auto-categorizes everything. I pasted the exact steps below. Copy-paste this to automate your own workflow too. The setup: - Xero OAuth2 app (free) - Counterparty mapping (JSON file that maps vendors to accounts + tax codes) - Puppeteer for CSV upload (Xero's Bank Feeds API is closed to non-financial institutions) - API calls to create matching bank transactions How it works: Xero reconciliation has two halves. Left side = statement lines (from CSV import). Right side = matching transactions (from API). When both exist, Xero auto-suggests the match. You just hit OK. The counterparty map is the brain. Simple JSON: "Google Workspace": { "account": "485", "tax": "NONE" } "Wise Charges": { "account": "404", "tax": "NONE" } Agent matches statement descriptions to this map. Unknown vendors get flagged for manual review. Gotchas that’ll save you hours: - Wise fee lines ("Wise Charges for: CARD-...") must map to Wise/bank fees, not the vendor - Tax type INPUT/OUTPUT needs LineAmountTypes: "Inclusive", not "NoTax" - Inter-account transfers need the BankTransfers endpoint, not BankTransactions Bonus: receipt flow. Agent creates bills, pays them via the right account, and attaches photos of receipts directly to the invoice in Xero. Forwarded 3 hotel booking screenshots, bills created + paid + receipts attached in one go. The whole thing runs as a skill file my agent can pick up anytime. Next bank statement that drops, same pipeline runs automatically. Total build time: one afternoon. ROI: every month from now on. If you’re a solo founder doing your own books, this is the move.

Complete automation guide with specific steps

Martijn Verbove@verboveAI · #3

Funny how the scope of a weekend project changes with Fable 5 — put together a nation-builder where every citizen is an AI you can actually talk to. How it works: 1. Start an island nation with 15 AI citizens 2. Govern by talking — build, tax, and settle disputes in your own words 3. Every promise you make is recorded, and the island remembers 4. Keep your word and trust grows. Break it and the story spreads 5. Grow your nation without losing your people — or face a coup

Demonstrates creative AI application with clear workflow

Martijn Verbove@verboveEngineering · #4

Switching from OAuth to API key in OpenClaw: 1. Set ANTHROPIC_API_KEY in openclaw.json under env.vars 2. Update auth.profiles to mode: "api_key" 3. Replace OAuth token in agents/main/agent/auth.json 4. Update agents/main/agent/auth-profiles.json with your API key 5. Restart gateway sk-ant-api03- = API key ✅ sk-ant-oat01- = OAuth ❌ If you see "oat" in your config, you're still on OAuth.

Step-by-step API key migration instructions

Martijn Verbove@verboveProduct · #3

I built an AI agent that lets me do social media by talking. Every morning it: - scans my X, LinkedIn, and Threads feeds - finds conversations relevant to my niche - suggests cross-posts, repurposed per platform - shows me tweets + comments one by one I read the thread, say my thoughts out loud, and it drafts the reply in my voice using my notes and memory. I haven't typed a tweet in weeks. Built it as an open source skill for OpenClaw. Comment "TALK" and I'll DM you the setup.

System design for AI-powered content management

Martijn Verbove@verboveProduct · #4

Yesterday, @thenewvolktimes and I launched t.co/AKF3HR4fMQ, and we want to say a huge thank you for the support and responses we’ve received 🫶 After watching how most enterprises approach AI, one thing became painfully clear: The problem isn’t access to AI. The problem is turning operational knowledge into systems that actually ship. Right now, most companies are still doing this: • endless AI strategy decks • pilots that never leave the sandbox • tools bought before the workflow is even defined • operators doing critical work manually while the knowledge stays trapped in their heads, docs, inboxes, and SOPs That’s the gap we built Doczen for. Most companies are treating AI like a transformation initiative. We think that’s backwards. You don’t become AI-native by starting with a giant roadmap. You do it by understanding where the real operational knowledge lives, structuring it properly, and turning it into something useful fast. One workflow at a time. One operational win at a time. One system at a time. That’s what Doczen helps enterprises do. We help teams move from AI theatre to execution by: • uncovering where operational value is stuck • capturing knowledge buried across people, documents, and processes • mapping the right tools, agents, and workflows • implementing step by step • getting live use cases into production in as little as 14 days We’re building Doczen for enterprises that know they need to move — but don’t want another AI presentation. They want traction. They want working systems. They want ROI. We’re still catching up with everyone who reached out yesterday, and we really appreciate the interest.

Enterprise AI adoption gap and execution method

Martijn Verbove@verboveProduct · #5

Distribution is still the problem. You can ship every single day for three months and still lose to someone with a better thumbnail and a bigger list 😕 So I built t.co/teA7vd9NVj 👀 A leaderboard for products, ranked by the work behind them — read straight off GitHub. No upvotes, no launch day, no copywriting. List your product for FREE and: → Rank on 90 days of real commits, not votes → A read-only meta-data key, we never see a line of your code → The people behind the products, ranked by their real commit share → Climb by consistency, not by posting on a Tuesday → See who's still shipping, and who quietly stopped Every score shows its own arithmetic. Add your product for free.

Commit-based ranking system for products

Martijn Verbove@verboveMindset · #2

Entrepreneurship is basically paying money to find out which of your ideas were stupid.

Reusable perspective on entrepreneurship

Martijn Verbove@verboveEngineering · #6

AI agents are not “automations with a brain”. They’re outcome systems. Most people only see the chat UI. The leverage is underneath: the infrastructure. 2026 shift: not automation. Outcome-oriented Agent. A simplified map: 1. Compute (CPU/GPU) Training + inference + latency. No compute, no scale. 2. Runtime infrastructure Containers, orchestration, queues, retries, observability. This is what makes agents reliable instead of “demo-y”. 3. Data layer SQL + docs + vector search. Agents are only as useful as the data they can reach. 4. Pipelines (ETL) Ingest, clean, transform, refresh. Without this, your “smart agent” runs on garbage. 5. Models LLMs, smaller task models, multimodal. Reasoning + generation lives here, not in the UI. 6. Routing Pick the right model per task by cost, latency, quality. This is how you make it usable in production. 7. Orchestration + protocols Tool calling, workflows, multi-agent coordination. MCP / A2A are the plumbing, not the product. 8. Tools + memory APIs, permissions, state, personalization. This is where agents actually do work. 9. Interfaces Chat, dashboard, voice, embedded UI. The surface, not the system. If you want a moat, stop thinking “prompt”. Start thinking “outcome”.

Teaches comprehensive agent infrastructure architecture

Martijn Verbove@verboveMindset · #3

The uncomfortable truth: your skills are a scarcity that no longer exist. For decades, the deal was simple: Learn the skill. Get the credential. Earn the premium. AI just nuked the premium. Not because skill is worthless, but because execution is no longer scarce. It’s abundant. So what becomes valuable? Imagination. The ability to see something that does not exist yet. And focus. Focus is the feedback loop where reality sharpens your idea while you build it. Skill is converting what’s in your head into reality. But the moment you start building, the idea changes. Your imagination refines. That refinement is focus. In an AI world, more people can do the work. Fewer people can imagine the right thing, then stay with it long enough to make it real. Imagination + focus is the new moat.

Skill economics framework: imagination + focus

Martijn Verbove@verboveProduct · #7

I built a tool to store everything you discuss with ChatGPT/Claude locally on your machine: - Saves files - Documents your decisions - Builds a knowledge graph - Makes it available via MCP Just install it on your Mac and it captures everything. t.co/qM7NU8Zduz

Tool design for local knowledge capture

Martijn Verbove@verboveProduct · #6

t.co/XLu43eH8Za is live. A leaderboard for people who actually ship. Scores recompute nightly from GitHub. First two product rankins: Gymlab (t.co/s4dHPfcEd8) 945 662 commits · 1.8M lines · 65/90 days · 2 builders VitalTrends (t.co/MxsNz9UILv) 915 298 commits · 300k lines · 67/90 days · 1 builder Builders: 1 @dspillere 915 2 @PeterPrins10 733 3 @verbove 212 Add your project.

Introduces ranking metric based on actual work

Martijn Verbove@verboveEngineering · #7

I rebuilt History Search as a fully native Mac app. It once won Product Hunt Lifehack of the Year, and the scope of the rebuild got out of hand in the best way. A local 3B parameter model screenshots every page and creates real breadcrumbs. Your browser tab might just say "Stanley." History Search knows which page you were on and gives you something useful. It indexes the full page, connects through macOS accessibility features. All data stays on your machine. Spotlight for the internet, entirely local.

Explains local-first AI approach with privacy architecture

Martijn Verbove@verboveGrowth · #1

Nothing wrong here. Except spending $4m on something that could have been built in 6 weeks. There’s no defensibility anymore in capital intensive Saas. Build fast, iterate ferociously, and invest in distribution. The best time ever to be a builder.

Teaches build fast, iterate, invest in distribution

Martijn Verbove@verboveProduct · #9

Most startup ideas don’t fail because they’re bad. They fail because building takes too long. Founders spend: • 6-12 months building • $$$ on tooling • and by launch day... the market has moved, the narrative changed, and the urgency is gone. The smartest founders in 2026 play a different game: Ship an MVP first. Not a “demo”. A small, functional product that can: • get real users • test the core assumption • collect feedback • iterate fast Speed is the advantage. Perfection is a tax. If you’re building this year, optimize for: launch velocity > feature completeness I’m currently launching a service that helps founders build and launch MVPs fast, with scalable SaaS + AI + custom software. If you’re sitting on an idea and want to ship in weeks, DM me. Happy to talk.

MVP-first speed beats lengthy pre-launch planning

Martijn Verbove@verboveLessons · #3

I keep seeing the take: “AI will wipe out jobs.” What I think we’re actually watching is classic economics reassert itself. When a technology drives the price of a product/service down, the market for it expands. Demand grows, supply has to catch up, and labor reallocates to the new bottlenecks. AI is a productivity shock. That means: • Some work gets automated (and yes, some roles shrink) • Prices drop • More people get access • Entirely new categories become viable • Human labor shifts toward what stays scarce: trust, taste, responsibility, physical execution, context, relationships I’ve seen this dynamic up close before. During Covid hotels cut staff, then demand returned overnight. Many hotels couldn’t operate at full capacity because the physical rooms existed, but the labor to run them didn’t. You can see a similar pattern in job markets like the Netherlands. The constraint is not “there is no demand.” The constraint is “where can labor actually flow to meet it?” Important nuance: it doesn’t mean everyone’s current skills will be in demand. It means the economy will keep rebalancing around scarcity. Zoom out: there are far more people who would like to eat out at restaurants than the current system can serve. If the process becomes more productive (less waste, lower cost, higher throughput), quality of life goes up and it becomes the new standard. Many things that used to be financially impossible will become normal. Like previous technology waves, this plays out over years, not weeks. The question isn’t “will AI remove jobs?” It’s “which constraints become the new bottlenecks, and how fast can we retrain and reallocate labor to meet expanding demand?” Curious how you’re seeing this in your industry: where is the bottleneck moving to?

Labor reallocation mechanism with case study

Martijn Verbove@verboveWorkflows · #3

Everyone is talking about the autonomous enterprise. Almost nobody is building one, because they’re trying to jump straight to “AI employees” instead of treating the company like software. The practical path is simpler: An autonomous enterprise is a set of business processes that run as loops — and refactor themselves after every run. Just like programming.

Autonomous enterprise as self-refactoring loops

Martijn Verbove@verboveGrowth · #2

I keep seeing the claim that Anthropic ran growth marketing with a team of 1. Whether or not the “entire marketing org” part is true, the underlying shift is real: AI + lightweight automation is collapsing the gap between “I have ideas” and “I shipped the workflow.” Anthropic confirmed 1 growth marketer (Austin Lau) ran growth alone with Claude Code to: • generate ad creative variations via a Figma plugin • speed up Google Ads responsive search ad (RSA) creation + export to upload-ready CSVs The interesting bit isn’t “AI writes copy.” It’s that non-engineers can now build repeatable systems: briefs → variants → QA → export → iteration. Takeaway for founders and lean teams: Stop thinking in channels (paid, SEO, lifecycle). Start thinking in pipelines (inputs, rules, QA, outputs, feedback loops). If you’re rebuilding your marketing stack in 2026, the question is: What are the 3 workflows you’d automate first?

Teaches pipeline thinking for marketing automation workflows

Martijn Verbove@verboveLessons · #4

The future of education is not “learning facts”. It’s learning how to build skills as systems. University should look less like: • memorize -> test -> forget And more like: • build repeatable systems for 1-2 years • using the latest tech + tooling • with real constraints (security, data, stakeholders, budgets) • in environments that resemble actual enterprises Because in the AI era, the value isn’t knowing the steps. The value is designing a machine that produces the outcome reliably. If you can build skills this way, you don’t just become “employable”. You become a force multiplier. You can: • replace parts of a team • automate a department’s workflows • or remove entire outsourced processes This is why “skill building” matters more than “credential collecting”. And why AI automation isn’t optional anymore, it’s the new literacy. The winners won’t be the people with the most knowledge. They’ll be the ones with the best systems.

Future education framework for skill systems

Martijn Verbove@verboveLessons · #5

You can't change the startup that failed, but you can reframe it. Find the pattern in it. Find the edge it gave you. Extract the insight and pass it forward. You can't choose what broke, but you can choose what you build next because of it.

Extract insights from failures to inform next builds

Martijn Verbove@verboveGrowth · #3

Great write-up by @johnrush on what's actually changing. #6 hits home — "build something agents want" instead of "build something users want." I run an AI agent that handles my daily workflows, social engagement, file organization, cron jobs. Most of my "usage" of tools is already agent-mediated. The next trend nobody's talking about: smart coded solutions for go-to-market. If AI makes building trivial, there's zero excuse not to put the same energy into distribution. Coded GTM strategies, automated lock-in, sticky-by-default products. That's where the craft is moving.

Teaches coded GTM strategies as growth frontier

Martijn Verbove@verboveProduct · #10

First principles thinking on "product-market fit" looks like this: 1. Pitch = You describe a problem and people immediately say "yes, I have that" = That's a pitch. 2. Word of mouth = Those people tell their friends who reach out wanting to try it 3. PMF = Strangers email asking how to buy before you've even launched

Teaches PMF framework as 3-stage progression with indicators

Martijn Verbove@verboveProduct · #11

Most people don’t lack ideas. They lack an unfair advantage. That’s what an opportunity is. A business opportunity is basically an asymmetry you can turn into something customers will pay for before the market catches up. There are 3 asymmetries worth looking for: 1. Information asymmetry You know something others don’t (yet). You’ve seen a problem up close, you have evidence, you have timing, you have context. This is where a lot of strong B2B ideas come from. 2. Distribution asymmetry You can reach the buyer faster, cheaper, or more reliably than other founders. Network, channel, platform, audience, partnerships, a proven motion. Great products die all the time because this part was “figured out later”. 3. Talent asymmetry You (or your team) can build something others can’t. Deep technical ability, rare domain + engineering combo, world class taste, speed. The best opportunities stack these. The most common founder mistake is having only one: an idea. And one more thing: asymmetries decay. Markets learn. Channels get crowded. Competitors catch up. Your window is the gap between “you see it clearly” and “everyone sees it”.

3 asymmetries framework: information, distribution, talent

Martijn Verbove@verboveAI · #5

Most agent builders jump straight to vector search and skip procedural memory entirely. I built a local knowledge graph with MCP that handles exactly this, capturing what the agent did and why so it can reuse those patterns.

Explains knowledge graph + MCP for agent memory reuse

Martijn Verbove@verboveProduct · #12

This post is a bit more personal. It is about Daniel and me, and how we got to building Doczen. We did not come together around a trend. We came together around a problem that had been very real in both of our worlds for a long time. Before Doczen, Daniel Volkenstein had spent years close to large organizations and the consultants working around them. He saw how much time and effort went into understanding how companies actually worked beneath the surface. Not the polished version. The real version. The one spread across teams, functions, and people who each hold a different part of the picture. On my side, I had lived that from a different angle. I had worked in consulting myself, where a big part of the job was gathering input from stakeholders, pulling signal out of a lot of noise, and trying to understand what was really going on inside an organization. Later, while building startups, I felt the same thing again internally. When you are gathering intel across a team of almost 100 people, you realize very quickly how much important context lives in conversations, habits, workarounds, and unwritten knowledge. That stuck with both of us. Again and again, we saw that some of the most valuable information in a company does not live neatly inside a system. It lives in people. In fragmented notes. In back-and-forth conversations. In observations that only come out when you ask the right person the right question at the right moment. And that is where things start to get hard. Because once you begin collecting all of that input, the challenge is no longer getting information. The challenge is making sense of it. Seeing patterns. Spotting what is broken. Understanding what matters. Turning a messy discovery process into something clear enough to act on. That is a big part of what led us to Doczen. Not a grand plan from day one. Not an abstract market thesis. Just years of seeing the same pain from different sides and knowing this part of the work was still far too manual, far too fragmented, and far too dependent on information staying trapped in people’s heads. Doczen came out of that experience. Out of sitting close to the problem. Out of seeing how hard it is to analyze what teams, clients, and stakeholders are really telling you. Out of believing that if you can structure that layer properly, you can help companies move with a lot more clarity and confidence. So this is not just a company we are building. It is a problem we both know personally. And in many ways, that is exactly why we care so much about solving it.

Explains problem-solving origin and organizational insight

Martijn Verbove@verboveWorkflows · #4

automation that can't learn goes stale. you need all three: → approving its work takes seconds → it improves every time you correct it → its reasoning is visible and editable most tools ship one.

Identifies key requirements for automation

Martijn Verbove@verboveEngineering · #8

Here’s your explanation rewritten into a tighter GraphRAG pipeline: 1. Split the conversation or documents into text units such as messages or chunks. 2. Use an LLM to extract entities from each text unit, including type and usually a short description. 3. Use the LLM to extract relationships between entity pairs in that same text unit. 4. Save the supporting source text / text-unit ID as provenance for each extracted fact. 5. Convert the extracted facts into triples: subject, predicate, object. 6. Merge duplicates and normalize names into canonical entities like one node for Cypher instead of many slightly different spellings. 7. Build a nodes table for entities and an edges table for relationships between node IDs. 8. Store that graph so it can later be traversed, filtered, embedded, or queried with something like Cypher

Teaches 8-step GraphRAG pipeline implementation methodology

Martijn Verbove@verboveEngineering · #9

Most people using OpenClaw have no idea what they’re doing. Not because they’re dumb - because the setup is messy and easy to screw up. I got tired of explaining it every day, so I removed the setup entirely. Wrapped it in Electron. Now it’s self-contained, easy to run, and scoped to its own filesystem. Next: browsing inside that same scoped environment. If you want it, Comment "OpenClaw".

Teaches removing friction through scoped product design

Martijn Verbove@verboveMindset · #4

'Passion' is a complete myth. And no one wants to say it out loud. I realized this after years of watching founders chase it: "Find what you love" // "Follow your passion" It never felt like real advice. Now I know why. Because it's backwards. You don't find passion and then build something. You build something meaningful and passion follows. Startups are deeply non-linear. A tiny set of decisions creates almost all the outcomes. For example: The founder grinding 80-hour weeks on the wrong idea burns out in silence. Meanwhile, someone ships a focused tool in a weekend, finds a real pain point, and compounds quietly for years. Who's working smarter? If passion was something you 'found', the most enthusiastic people would win. The hobbyists and the dreamers would be running the companies shaping the world. Instead, it's the opposite. Detached builders who shipped without permission made the tools you rely on. Pragmatic weirdos who ignored the rules built the platforms you post to. The best founders I know didn't follow passion - they followed signal. The sharp ones always saw through the 'find your passion' story. Because passion is an EFFECT, not a cause. You don't manufacture it. It emerges when the work compounds, when users respond, when the mission gets real. The indie hacker who can't stop building his tool - he's not 'grinding'. He's in flow. The mission is pulling him forward. Effort is a natural byproduct of aligned work. This is why you have to choose your problem with INTENTION. Because once you find real traction on something that matters, you'll stop waiting to feel ready. Let the problem pull you. Build the thing.

Passion follows building; effect not cause

Martijn Verbove@verboveMindset · #5

highest ROI lifehack I've used recently: switch your phone to grayscale → avoids attention grabbing apps keep one page only by "removing from homescreen" → stops doom swipping thank me later

Teaches concrete reusable productivity lifehacks

Martijn Verbove@verboveLessons · #7

There is a lot of startup advice floating around from people who have "figured it out" Most of it will quietly wreck you if you follow it too closely, because it was built for who they are, not who you are Borrow the principles, ignore the playbook Their path is not your path

Borrow principles from others; ignore exact playbooks

Martijn Verbove@verboveProduct · #20

My girlfriend's running app locked her progress behind a paywall after 1 mile. So I asked her what she actually wanted and built it in a few days. The app is called Stride. Spoken milestones, no tracking, clean UI. Run. Don't look at your phone. t.co/t0os5sfWLk

Teaches building from user need, not platform constraints

Martijn Verbove@verboveProduct · #15

The best founders don't keep adding to their roadmap... They cut everything that isn't essential. Most of what you're building right now - you shouldn't be.

Best founders cut non-essentials relentlessly

Martijn Verbove@verboveLessons · #6

There is a lot of founder advice circulating from people who genuinely built something real Most of it doesn't transfer. They succeeded because of who they are, not just what they did Copying their path without their context, network, and timing is how you waste years Take the signal, ignore the blueprint. Your edge has to be yours.

Context matters more than copying founder playbooks

Martijn Verbove@verboveAI · #10

People forget AI is mainly a competitor to software, which has a very standardized pipeline. It can’t just be run wild. Even AGI can’t just “figure out” a business on its own. Humans are AGI too. We still need context, skill-building, and oversight to become useful inside a complex company. Same for AI. And most real commerce is physical anyway. Software is an enabling layer. AI will be too.

Explains AI as standardized layer needing context

Martijn Verbove@verboveProduct · #16

building a startup is literally just 3 things on repeat: ship → get users → make them successful people complicate this with 47-slide decks and quarterly OKRs but every founder who's won just did those 3 things faster than everyone else build the thing. find the people. help them win. that's it. that's the whole playbook.

Startup success: ship, find users, help win

Martijn Verbove@verboveMarketing · #1

TL;DR: AI doesn’t care about your launch video. Everything you call “growth” was built on one assumption: a human is doing the choosing. If AI becomes the layer between users and the internet: • content gets parsed, not read • brands get ranked, not remembered • funnels get skipped, not optimized AI doesn’t get persuaded. It gets evidence. In that world, you don’t win by being “better at SEO”. You win by being the best tool for a specific job, and by having a founder people trust. Everything else starts to look like optimizing for an audience that’s disappearing. Marketers: if AI is the buyer, your next skill isn’t copywriting. It’s systems. What are you learning first: data, automation, or AI evals?

Teaches growth strategy shift when AI is buyer

Martijn Verbove@verboveAI · #9

Most "AI agents" are just prompts with extra steps. That's not a dunk. That's the whole point. A Managed Agent is three things: → A prompt. The instructions you'd type into ChatGPT anyway. → Tools. Connectors to Slack, Jira, Intercom, Confluence — so it can fetch its own context. → Skills. Reusable playbooks for how the work gets done. That's it. No magic. The hard part isn't building the agent. It's knowing where to point it. And here's where most teams get it wrong: they point it at reports. Daily summaries. Weekly digests. Dashboards nobody reads. It feels productive. It's mostly noise. The real wins are upstream — at the moments where someone is spending an hour collecting context to do five minutes of work. Five places agents actually earn their keep: • Audit your docs against what support actually told customers this week • Auto-build the internal FAQ from repeated Slack questions • Enrich existing Jira tickets with org-wide context — don't open new ones • Sweep Slack + Intercom + Confluence for stale documentation • Draft the customer follow-up while you're still in the room Notice the pattern? None of them are reports. All of them close a loop a human is currently closing by hand. The architecture (managed in Claude vs. self-hosted with the SDK) matters less than where you point it. Swipe through ↓ for the full breakdown.

Explains managed agents + lists 5 specific high-ROI use cases

Martijn Verbove@verboveMindset · #6

Stop trying to predict which AI skills will matter in 3 years. Nobody knows. The space is moving too fast to plan that far ahead. What you actually need is the ability to pick up any tool and get productive with it fast. That only happens by building real things. Your own agent. Your own workflow. Your own product: - I built to Gymlab to rethink how my fitness plan gets built. - We built Doczen around issues we encountered ourselves. Then put it in front of people. Does it solve a real problem or not? That pressure forces you to learn what matters. Everything connects to something real, which means you stop depending on a single platform, a single skill, or a single employer who will cut you the moment something cheaper ships.

Teaches building real things as best way to learn AI skills

Martijn Verbove@verboveProduct · #17

soon nobody will buy advice. they’ll buy the license, granted. the policy, in force. the agreement, signed. every business today sells you the process and leaves you the last mile. that’s the part about to disappear.

Teaches future business model shift insight

Martijn Verbove@verboveProduct · #14

Roadmap to Build an AI-Powered SaaS: 📂 Identify a painful workflow ∟📂 Design a simple agent loop ∟📂 Connect an LLM + tools ∟📂 Ship a rough MVP ∟📂 Add memory and context ∟📂 Build feedback loops ∟ ✅ Charge for it

Step-by-step roadmap for building AI-powered SaaS

Martijn Verbove@verboveProduct · #19

in the future, a business isn’t a company that uses automation the business is the automation millions of specialised systems, each doing one job end to end people won’t operate them. they’ll check the output and update it

Teaches vision of automation-centric business

Martijn Verbove@verboveProduct · #18

Software is becoming a lot like dinner. People will keep consuming a select few exceptional products. But most of it is going to cooked at home. Companies that will succeed are the ones who supply ingredients and make cooking easier.

Teaches software distribution business model insight

Martijn Verbove@verboveWorkflows · #5

I redesigned my App Store screenshots without opening a design tool. Trained a skill to build them in HTML, export to canvas, then into Figma for manual edits. The design workflow stayed the same. AI just became a better entry point. That's the pattern. AI is good at reading, writing, and following structured steps. So stop handing it one task inside a broken process and give it the whole pipeline, from research to analysis to finished output. In essence, it's all about making decisions, documenting those decisions, and making them repeatable.

Teaches AI integration into design workflows

Martijn Verbove@verboveAI · #11

"replace employees" is bait. the real thing is decision density + shipping speed. contrarian take: agents won’t replace employees. they’ll replace companies with individuals. biggest opportunity you’d bet on in the next 6 months? drop 1.

Teaches agents enable individuals at company scale

Martijn Verbove@verboveMindset · #10

If you are choosing between the proven path and the unproven one, take the unproven one. Because if the proven path were the better path, someone else would have already won on it.

Reusable principle: choose unproven path when proven path exists

Martijn Verbove@verboveWorkflows · #6

someone built an agent that interviews them about their own career one folder per role, weekly check in, what did you ship, what broke, what's worth remembering that isn't note taking that's a knowledge base that fills itself most companies still don't have one t.co/eW7vd0wwwC

Teaches AI workflow pattern for knowledge management

Martijn Verbove@verboveWorkflows · #7

the most important part of automating a workflow is doing the workflow while you build it. that’s what makes a skill powerful. it’s an auto-play of what you just did. not a diagram of what you think happens.

Teaches principle of building automation effectively

Martijn Verbove@verboveAI · #12

the biggest risk with ai isn't that it takes your job it's that your job never delivered an outcome to begin with ai is already better at reports, emails, and booking meetings if that's the job, that's the warning

Teaches insight about AI's actual impact

Martijn Verbove@verboveWorkflows · #8

every company is a set of processes. making one autonomous means rebuilding it in three layers. the top layer is the flow: created, updated, cancelled. under each of those sits the data that has to be collected, updated, or generated. under that sits the reason any of it is allowed; corporate policy, legislation, best practices, the stuff nobody ever documented. the top layer is easy, the bottom layer is why the most automations fails.

Teaches three-layer automation framework

Martijn Verbove@verboveWorkflows · #9

before a company can run itself, someone has to describe how it actually works. that comes down to three questions, stacked. where is this right now? — created, approved, cancelled. what do i need to know to move it forward? — the data, the inputs, the fields. what am i allowed to do? — policy, law, and the things your best employee knows without being told.

Teaches method for analyzing processes

Martijn Verbove@verboveEngineering · #11

an autonomous organization is a composition of three primitives: → collect data → validate data → transform data everything else is arrangement

Teaches framework for autonomous systems

Martijn Verbove@verboveAI · #13

Everyone wants AI to think for them. It doesn't think. It reads, writes, and follows steps. The companies winning with AI pointed it at their most boring work. Copying data between systems, turning research into summaries, running content from draft to publish. Give it the full pipeline with clear inputs and outputs and it'll outwork your whole team.

Explains AI capabilities and winning application patterns

Martijn Verbove@verboveMarketing · #2

Cold outbound works better when the channels are sequential, not parallel. 3 emails over 10 days, then LinkedIn, then phone. Each channel only fires after the previous one goes cold. You get cleaner data this way. You can actually see where the sequence breaks instead of guessing which channel flopped.

Teaches sequential outbound method with data clarity

Martijn Verbove@verboveAI · #14

Experience is to people what context is to LLMs.

Conceptual insight comparing experience and context in AI

Martijn Verbove@verboveMindset · #7

You become a good founder by understanding what it's like to have no users. You become a good builder by understanding what it's like to ship something nobody wants. You do not start out as a visionary. You start as someone who was just wrong less often over time.

Teaches founder development through failure and iteration

Martijn Verbove@verboveProduct · #21

I started shipping side projects just to stay sharp. And slowly, my thinking got clearer. My execution got faster. My instincts around product started compounding. Building is the best way to learn.

Reusable insight: building improves thinking faster than theory

Martijn Verbove@verboveLessons · #8

In my first company, I thought it was about the product. In my second, I realized it was about the distribution. In my third now, I realize it's about the timing.

Condensed evolution: product → distribution → timing matters

Martijn Verbove@verboveAI · #15

"Everyone talks about how AI will replace jobs in the future, but no one talks about how the decisions you make about AI adoption right now are already determining which jobs survive."

Insight about timing decisions determining job survival

Martijn Verbove@verboveMindset · #8

Startups make a lot more sense once you stop confidently repeating things you haven't actually validated yourself. But founders are desperate to sound like they've done the work, so they treat "I read a thread about it" as equivalent to knowing something.

Insight about validating vs. assuming knowledge from reading

Martijn Verbove@verboveAI · #16

AI transformation 101: - Find the process bleeding time (bottleneck) - Map where the friction actually lives (discovery) - Pull signal from your tools - Doczen does this fast (diagnosis) - Build a prioritized roadmap with real ROI attached (direction) - Ship automation that sticks, not decks that don't ($) Stop waiting on consultants.

Teaches 5-step AI transformation methodology with real ROI focus

Martijn Verbove@verboveProduct · #22

The strongest signal of a great AI tool is not how many features it has. It's how well it adapts to you. t.co/5o9qJzCv5s is one of them, it does not give you a generic plan. It learns your history, works around your limits, and adjusts weekly. The best tools are the ones that get smarter the longer you use them.

Teaches principle: great tools adapt and improve with use

Martijn Verbove@verboveMindset · #9

As someone who has moved between struggling founder and one with real traction, you learn that momentum doesn't build gradually - it collapses into a single inflection point. For a long time, nothing moves. Then a few signals here and there. At the peak, inbound becomes relentless and compounding. I'm not saying I'm there permanently - I'm not. But once you've operated at that level for any stretch of time, you're exposed to a completely different game. Hard to explain to anyone who hasn't crossed that threshold. It's like the market unconsciously rewards you just for showing up with the right thing at the right time.

Explains momentum isn't gradual; describes inflection point pattern

Martijn Verbove@verboveProduct · #23

When pitching an AI product, it's usually better to lead with outcomes rather than capabilities. Say what changes for the user rather than what the system does. Not because it sounds more compelling, but simply because it's what people actually buy.

Teaches pitching strategy: lead with outcomes not capabilities

Martijn Verbove@verboveAI · #18

Claude just prompted me to import cookies from Chrome It's a more fundamental shift to using web services then we currently realise AI clients are the user interface saas/sites will: - transform from interfaces to capabilities - elements on webpages will become widgets - software will become proactive in reaching out to you

Explains how AI clients transform software interfaces

Martijn Verbove@verboveGrowth · #4

If you don’t know what to build, there’s only 1 playbook: 1.Ship a 48h MVP (just prove the idea) 2.Launch a simple landing page (capture demand) 3.Go ALL IN on distribution (content, DMs, scraping attention) No distribution = no startup. V1 will be wrong. Speed is the edge.

3-step MVP-to-distribution playbook for founders

Martijn Verbove@verboveGrowth · #5

The rules of building products changed overnight-and most founders haven't noticed yet. Building is now about access, not product capabilities. With every idea, ask yourself: how does this grow? What makes this inherently viral? We've moved from "what can we build?" to "what can we build that sells itself?" Look at TrustMRR. It's brilliant because founders want to prove themselves publicly + content writes itself; deals, transactions, wins the community actually craves. Same pattern with indie hacker products like NomadList. They're marketable because sharing the information IS the product. Connection via the community follows naturally. A good idea in 2024 isn't about clever functionality you can hack together. It's about building something with distribution baked into its DNA. Build products where growth is the feature, not an afterthought.

Distribution-first product strategy with viral mechanics examples

Martijn Verbove@verboveEngineering · #12

Early in my first company, a seasoned operator told me, "Build systems, not habits." I nodded. I had no idea what he meant. Over time, I figured it out. Building systems means recognizing that your output is a function of your process. Document your decisions. Automate the repetitive. Write specs before you build. Review what shipped. Most founders skip these because they feel slow. They are not slow. They are what separates founders who scale from founders who stall. When your process is sloppy, your team pays for it. Unclear specs mean engineers build the wrong thing. No documentation means every new hire starts from zero. No retrospectives mean you repeat the same mistakes at twice the cost. Every broken process is a tax on everyone around you. The biggest myth in early-stage companies is that speed and structure are opposites. The founders who move fastest have the most repeatable systems. They ship quickly because decisions are pre-made. They maintain quality because checklists replace memory. This is not bureaucracy. It is leverage. The compounding works both ways. Founders with clean systems attract better operators. Better operators build cleaner systems. The gap between systematic founders and chaotic ones does not stay constant. It widens every quarter. You can treat process as overhead that slows you down. Or you can treat it as infrastructure that multiplies your output. You can believe moving fast means skipping steps. Or you can understand that the right steps are what make speed sustainable. The operator who told me to build systems understood something most founders learn too late: discipline is not the enemy of creativity. It is what makes creativity repeatable.

Systems over habits: specs, documentation, retrospectives

Martijn Verbove@verboveWorkflows · #10

Steal this: run a weekly AI memory audit. Your chat history is not knowledge. It’s a graveyard. Once a week, have Claude summarize your Claude Code sessions and cluster what you actually asked it to do into 4 buckets: 1. SKILLS Repeatable tasks you still do manually (should become a button) 2. AGENTS Autonomous workflows (research + action, end to end) 3. SCHEDULED TASKS Stuff to cron (recurring, predictable) 4. CLAUDE.md Your repeated preferences + context (the rules you keep retyping) For each item, include: • 1-line description • session id(s) • frequency count • recommended bucket Sort by frequency. Result: your real workflow, not your imagined one. Whatever you repeat becomes automation, docs, or a skill. Copy/paste prompt: “Analyze my Claude Code session logs from the past 7 days. Extract every user request as a single line. Deduplicate similar requests. Cluster into: SKILLS, AGENTS, SCHEDULED TASKS, CLAUDE.md. For each cluster, list: summary, session id(s), frequency, and what to build next. Sort by frequency. End by asking me which session id to expand for full context.”

Weekly AI audit to surface automation opportunities

Martijn Verbove@verboveAI · #17

95% of enterprise AI projects deliver ZERO measurable ROI. Not low ROI. Zero. MIT looked at 300 deployments, interviewed 150 leaders, surveyed 350 employees. The needle didn’t move. Meanwhile companies dumped $30-40B into “AI” last year and mostly lit it on fire. And everyone’s still hiring “Head of AI”. Why? Most companies don’t have an AI strategy. They have AI anxiety. They see a competitor’s press release, panic, and spin up 10-15 pilots with: • no success criteria • no owner • no kill switch That’s not innovation. That’s expensive theater. Comment “Get Shit Done” and I’ll DM you a free assessment of where AI can actually drive ROI in your business.

Enterprise AI fails without strategy; why and fix

Martijn Verbove@verboveEngineering · #10

I have agents running cron jobs — feed scans, email checks, file org — and "observability" is just me scrolling logs. We built a dashboard as the read layer on top of the agent's workspace so you can actually see what it did and what broke. The self-report angle is the missing piece — the agent knows what it was trying to do when it failed.

Agent observability dashboard tracks agent workspace state

Martijn Verbove@verboveAI · #19

What excites me about this is the participation angle. Right now barely anyone even has access to AI, but zoom out — most people on earth have never had access to a good lawyer, a financial advisor, or top-level education. When cognition gets cheap enough, all of that opens up. And honestly as a builder it just becomes way more fun — instead of grinding on one problem for two years, you can tackle ten and actually bring them to market at price points that work for way more people.

Explains how cheap cognition opens underserved markets

Martijn Verbove@verboveProduct · #24

The era of raising $10M to ship a mediocre SaaS product is over. AI didn’t make software buildable. Software was already buildable years ago. And most products still failed. Why? Because building was never the bottleneck. Solving a real customer pain with measurable output value was. That’s still true now. Lower build cost does not create better companies. It removes the old protection layer. In the old world, you could burn millions on bloated software and survive because development was slow and expensive for everyone. That protection is gone. Now the bar is simple: • deliver value fast • integrate into real workflows • iterate continuously • price on outcomes, not features Hot take: AI won’t kill startups. It will exposes bad ones faster.

Explains why lower build cost doesn't guarantee success

Martijn Verbove@verboveAI · #20

LLMs are not intelligent. I know that sounds weird. But hear me out. Birds are intelligent. Dogs are intelligent. Dolphins are incredibly intelligent. None of them have language as we know it. What set humans apart wasn't being smarter. It was that language let us build on top of other people's discoveries. You don't re-invent calculus. You read Newton. Language is the result of intelligence, not the source of it. So what does an LLM actually do? It reverse-engineers intelligence from language. Every token is compressed human knowledge, not thought. That's why context matters more than the model. The intelligence is already in the weights. Your job is to direct it. The people building the best context and skills will 10x the gap over everyone else. Not because they have better AI. Because they know what to feed it.

Explains LLM mechanism and importance of context

Martijn Verbove@verboveGrowth · #6

Meta quietly shipping "Manus" — an AI agent built right into Facebook Pages that creates posts and engages your audience. This is the pattern. The AI race isn't about models anymore — it's about distribution. And the platforms already have it. The good news: GTM itself is being disrupted by AI too. The same tools that let Meta embed agents into 3 billion feeds also let a solo founder build coded distribution that was impossible a year ago. Distribution is the moat. But for the first time, it's not only reserved for big tech.

Teaches distribution as AI moat for individuals