About This Project

What This Is

This is a working record of an experiment. We’re building an automated crypto trading system from scratch, running it on decentralized exchanges with our own wallets, and writing down what happens every night. That includes the parts that don’t work.

There’s no course to sell, no token to pump, and no signal group. Just one small, independent operation with a background in IT and software, working with an AI coding assistant, and trying to answer a plain question: can a careful, rules-based bot actually make money after every cost, or is holding the coins the better move?

Why We’re Doing It

In order of priority:

  1. Net profit. Not “number go up,” but profit after trading fees, slippage, network fees, software and AI costs, hardware, and taxes. If it can’t do that, we’ll say so.
  2. Growth over time. Breaking even counts as a failure.
  3. A system that runs itself. The goal is an AI-assisted agent that handles normal operations on its own, inside hard limits we set, and that we can adjust or stop at any time.
  4. Learning. Blockchains, wallets, scripting, automation, and the craft of building trading systems properly.
  5. Building something of our own, instead of only building for other people.

The Honest Starting Point

We’re going in with clear eyes:

  • Most retail trading bots lose to simply holding. Every strategy we try gets scored against two benchmarks: buying and holding the same assets, and parking stablecoins in a low-risk lending pool. If the bot can’t beat both after costs, it hasn’t earned its keep.
  • Small trades magnify fixed costs. Network fees don’t shrink with trade size, so we track every cost and work out the minimum trade size that still makes sense.
  • Returns come from capital times edge, not hours. We track the hours we put in, and the effective hourly rate, so the time cost is visible too.
  • Some strategies belong to professionals. We stay away from games where we have no edge, like racing other bots to front-run trades or sniping brand-new tokens.

How It Works

  • Decentralized exchanges only. No accounts on centralized exchanges. Trades go straight against on-chain liquidity from a wallet we control.
  • Rules first, then AI that earns its role. The first version trades on plain, testable rules. From there, AI models get more of the decision-making in stages: first as a filter that can veto trades, then tuning parameters, then proposing and eventually making trades. Each stage runs in shadow mode against the rules first, and it only goes live if it improves net returns after its own costs. Hard risk limits stay in code underneath everything, whoever is deciding: position size, loss limits, and kill switches.
  • Tiered wallets. Long-term holdings stay in cold storage that never touches the bot. The bot gets its own wallet, funded with only what it’s allowed to trade.
  • Earn each step. Research, then backtesting on historical data, then paper trading with live prices and fake money, then a small amount of real money. Only then does it scale. Each stage has a gate it has to pass, plus conditions that stop the project if they hit.

What You’ll Find Here

  • A nightly post covering each day’s work: what we built, what the automated system did, what we decided, and an honest read on how it’s going. On days with no activity, there’s no post.
  • Real numbers once there’s something to measure: returns against benchmarks, drawdowns, and costs. Reported as percentages and rounded figures.

What You Won’t Find

  • Wallet addresses or transaction IDs. Blockchains are public. Publishing those would expose every balance and trade, so we report results without them.
  • Personal details about who’s behind this. Given the nature of the work, the site is meant to be informative without being personally identifiable. To reach us, use the contact link in the footer.
  • Advice. This is a record of what we’re doing and why, not a recommendation for anyone else, and not financial advice.

A Note on How Posts Are Written

The nightly posts are drafted by an AI from our internal work logs, then run through an automated privacy check before they’re saved. The work they describe is real. The voice is “we” because it’s a joint effort between a person and the software helping them build this. There’s a real person behind every post: each one is read, checked for accuracy, edited where needed, and approved before it goes up.

Where Things Stand

The project started in late October 2026, but the idea goes back to the early 2000s, long before the tools and knowledge existed to make it real. Today, a project like this is within reach of anyone with a little technical skill and a lot of patience.

The first job was building this record-keeping system, so the story is documented from day one. No money is at risk yet. Next comes collecting market data and benchmark history, then the research decisions: which blockchain, how much capital, what risk limits, and which first strategy.