I got my first PC when I was twelve.
Since then I have ridden a few waves: the web, startups, crypto, and now AI. The sea behind this page changes with each one.
A sketch of how each wave felt to ride. Not market data, not to scale. Drag the surfer, or scroll.
At twelve sea state: calm
I got my first PC when I was twelve.
A computer does exactly what you tell it, and nothing else. If something breaks, the mistake is yours. If something works, you made it work. Every chapter below runs on that deal.
This is the calm water at the start of the story. No market, no hype. One machine.
On GitHub since May 2017 sea state: rolling swell
I opened my GitHub account in May 2017.
By 2019 it showed where my hands were: mobile and web. An Angular playground. Lottie, which plays After Effects animations on the web and on phones. The Apache Cordova file-transfer plugin. InstantClick, which makes page loads feel instant. A file-transfer repo of my own.
I also had startups. More than one.
This was a rolling swell, not a storm. The tools changed every year, but you could read the sets. A new framework is not a wave. A new kind of user is. Phones made one, and the web had to follow.
Startups teach one thing fast: the code is the easy part. Someone has to want what it does.
October 2022 to October 2024 sea state: storm
My Polkadot chapter starts in October 2022, with a Greek guide to Polkadot.js: infographics that explain the core features of Substrate relay chains.
Polkadot is a network of blockchains. A relay chain in the middle shares its security with the chains around it. Substrate is the Rust framework those chains are built with.
A month later FTX collapsed, and the whole crypto market went into a storm. I kept building.
Most of that work made Polkadot readable, in Greek and in plain words. Then I went down to the engine room.
polkadotjs-guide-gr A Greek guide to Polkadot.js, in infographics.
dotsama-tools A list of the tools you need to find your way around the Polkadot ecosystem. Around it, in 2023 and 2024: polkafeed, infosama, dotmap and pokerdot.
polkajs A multilingual how-to for the polkadot.js wallet.
dotGR Polkadot Greece: an introduction to Polkadot in Greek, and the first Greek library for the Polkadot network.
w3flib The Web3 Foundation Library: a site that collects Web3 Foundation resources.
An OpenGov tracker for Polkadot governance, in five repos: og-tracker, og-web, og-docs, og-report and gov-data. I registered ogtracker.io on 5 January 2024.
Rust. I worked through the Polkadot Blockchain Academy material: a blockchain from scratch, proof of work, a frameless runtime, benchmarking, XCM and Merkle trees.
walletConnect The last new repo of this chapter.
OpenGov is how Polkadot holders vote on referenda, including how the treasury spends. Remember the tracker. It comes back in the next wave.
April 2026 on sea state: giant swell
From October 2024 to April 2026, no new repos appeared on my GitHub. In May 2025 I registered koglabs.com, the name of my lab, KOG Labs. Then the water started to rise.
In April 2026 I paddled into the AI wave. One of my first repos, on 28 April, was polkagov: a read-only Polkadot OpenGov tracker with AI summaries, risk flags and identity lookup. Polkassembly shut down in 2026 and left a gap. My governance work from 2024 and the new models met in one repo.
In May I started ai-ops-hub, an autonomous AI ops center. It grew into the company I run today. Every employee is an AI agent: Claude Opus, Sonnet and Haiku, Codex and Grok. They write, test and ship my products. I decide what gets built and what goes live. Next to it came forge, an agent that makes content and MVPs on its own.
On 5 July 2026 I wiped out. I shut down my own agent fleet, 31 timers in one day, after a $52.78 burn in 48 hours that I had not seen happen. The company came back different: every job starts from a task, gets verified, ships and stops. A stop-loss that only downgrades the model is a discount, not a brake. I wrote it up as an essay.
The same month I built triad, a live console where Claude orchestrates Codex and Grok, and HookOS, a local lead-gen SaaS.
In October came eta, a track record of AI agent estimates that my harness verifies, nerfIndex, which tracks AI model quality every day, and this site.
In the last 30 days my agents logged 2,576 runs and 805.5 busy hours. A 30-day month has 720 hours. They work in parallel. In the last 90 days: 12,566 commits across 32 repos. In the last year GitHub counted 63,211 contributions, 99% of them private.
Live numbers from my own harness, git and GitHub, measured . More on Pulse.
This is the biggest wave I have ridden so far. The only way I know to ride it is to stay on the board and keep shipping.
Today sea state: settled
Today I am a solo founder and the only human in my company. The agents build. I decide.
What the company ships is on Shipped. How it works is on Machine. What I measure about it is on Research.
The sea behind this chapter settles to your real time of day.
Calm water does not mean the wave is over. It means I am out past the break, where you can see the next set coming.
Calm water at twelve. A rolling swell on the web. A storm in crypto. A giant swell with AI. Each one taught me something about timing. This is how I read a wave now. It is my view from my own rides, not a forecast.
A wave starts far from the beach. Tech waves form the same way. First the tools appear, and only builders can use them. Then the builders ship things. Then the headlines come, and then the crowd. ChatGPT came out on 30 November 2022, weeks after my first Polkadot repo. While I rode one wave, the next one was already forming behind me. So I keep one eye on the horizon, even mid-ride.
Noise is price and headlines. A swell is what builders can do this year that they could not do last year. My test is simple: can I build something this month that I could not build last year? If yes, it is a swell. In the crypto storm the prices were loud and the tooling was quiet. I worked on the quiet part. Two years later it came back inside an AI product.
By the time a wave is on the news, it is breaking and the good spots are taken. Paddle in while it is still a swell, and start small. My Polkadot chapter began with a guide, a month before the market crashed. One of my first AI repos was a small tracker built on what I already knew.
Everyone who rides gets thrown. What counts is how fast you come up and what you change. One wipeout cost me $52.78. I shut down 31 timers the same day. Now every loop that spends money has a hard off switch I can reach in one tap.
From The giant swell
A wave is easy to talk about. I keep a ledger instead. Every agent run, every estimate and every verdict gets recorded by my own harness, and the numbers sit on this site with their windows. If the ride is real, it shows up in the numbers.
From The giant swell
Between sets the sea goes flat, and many people paddle back to the beach. The ones who stay are in position when the next set arrives. On my GitHub the flat water ran from October 2024 to April 2026. The OpenGov work from before it came back after it, with AI on top.
I am still in the water. If you ride the same wave, interview me or find me on X.