I Skipped the Practice Subnet. Nobody Told Me To, Which Is the Problem.

 Three posts ago, I described the moment I ran btcli subnet burn_cost --netuid 41, watched a number blink back at me in TAO, and sent it. Real TAO. Real subnet. No undo button. I wrote about it like a rite of passage, because it felt like one — my hands were actually shaking a little when I hit enter.

It took me until this week, going back through the Co-Learning Camp curriculum to write something new, to notice the thing sitting in plain sight that I'd completely walked past: there's an entire subnet built for exactly the moment I was in, and I never touched it.

It's called NetUID 1. It runs on testnet. The TAO is fake. You can register, mess up, get deregistered, and register again, and the only thing you'll have lost is a few minutes. It exists specifically so that the first time you ever type btcli subnet register doesn't have to be the time it actually matters.

I went straight to mainnet instead. Nobody stopped me. The chain doesn't check whether you've practiced first — it just takes your TAO and gives you a UID either way. And that's the thing I want to write about, because I don't think I'm the only one who did this, and I think the reason says something real about how this ecosystem teaches people, not just about my own impatience.

The step I didn't know I'd skipped

Here's roughly how the camp curriculum is actually sequenced, which I only properly absorbed in hindsight: you install the dependencies, set up the Bittensor SDK in a clean virtual environment, create a wallet — a coldkey and a hotkey, with a 24-word mnemonic you're told, repeatedly and emphatically, to write on physical paper and nowhere digital — and then, before any of that wallet ever touches real money, you register on testnet NetUID 1 using testnet TAO you got for free from a faucet.

That registration is functionally identical to a mainnet one. Same command shape, same confirmation prompt showing you the cost, same UID assigned afterward, same metagraph you check to confirm it worked. The only difference is the word test in one flag, and the fact that the TAO behind it doesn't exist anywhere except a sandbox. You get a real UID. You get a real immunity period — the window where your slot can't be taken even if your miner is doing nothing at all, which exists purely so you have breathing room to figure things out without instantly losing your spot. You get to make every beginner mistake there is to make, on a system that behaves exactly like the real one, with zero financial consequence.

I read through that page after the fact and felt something close to embarrassment. Not because testnet is hard to understand — it's almost insultingly simple once you see it laid out — but because I'd had three separate chances to notice it existed before I burned real TAO, and I didn't take any of them.

Why I skipped it without ever deciding to

I want to be honest about the actual mechanism here, because "I was impatient" is true but incomplete.

The camp materials don't hide testnet. It's right there, a full step before mainnet registration, with its own page, its own troubleshooting table, its own explicit callout that POW registration is disabled and burn is the only path — exactly the kind of detail you'd want to know before, not after. But I came into this with a specific subnet already in my head. I'd heard about Sportstensor before I'd properly read the curriculum in order. I wanted to mine on SN41, specifically, and "specifically" is the word that did the damage. I wasn't moving through the camp's sequence; I was moving toward a destination I'd already picked, and I skimmed past whatever didn't look like it was on the direct path to that destination.

Testnet didn't look like progress toward SN41. It looked like a detour. NetUID 1 isn't Sportstensor — it's not even a "real" subnet in the sense of doing anything economically meaningful, it's explicitly a learning environment. So my brain filed it under "optional tutorial," skipped to the part that looked like the actual goal, and only the cost of that decision — a non-refundable burn fee, gone the second the transaction finalized — made the skip visible at all.

That's the part I think is worth naming plainly: the danger wasn't that testnet was hard to find. It's that it's easy to convince yourself a practice step is optional right up until the moment you've already paid for skipping it.

What the immunity period actually would have taught me

Here's the specific thing I think I lost by skipping straight to mainnet, and it's not abstract — it's a concrete mechanic I didn't understand until I read about it properly this week.

When you register a miner, you get an immunity period. On mainnet it's roughly 24 hours; on testnet it's shorter, but the idea is the same either way: for that window, your UID can't be deregistered no matter how badly your miner is performing, even a flat zero score. It's a grace period, deliberately built into the protocol, so a brand-new miner has time to actually get its software running, start responding to validator queries, and produce something before the network's natural churn — new miners constantly registering and pushing out the worst performers — has a chance to bump it off the subnet.

On testnet, that grace period costs you nothing extra to use badly. You can let your miner sit there doing absolutely nothing for the entire immunity window, watch it get pushed out, and the only thing you've spent is time. You can register, deregister, and register again as many times as it takes to get the actual deployment — the part where your miner process runs continuously and answers validator queries correctly — right.

On mainnet, the immunity period is the same mechanic, but every minute of it is minutes you paid real TAO to get, on a slot you can't casually throw away to try again. I didn't get to practice "what does it feel like to watch the immunity countdown while your miner isn't doing anything yet" in a setting where that feeling was free. I got to feel it for the first time with money already spent, which sharpens the anxiety without teaching you anything you couldn't have learned for nothing.

The boring detail that would have saved me real stress

There's a smaller, more mechanical lesson in the testnet flow that I think is genuinely underrated, and it's this: the registration troubleshooting table for NetUID 1 lists, plainly, the exact errors you're likely to hit — insufficient balance, a hotkey that's already registered, a wrong NetUID, a subtensor connection timeout, a UID that takes a few minutes to show up in the metagraph because the chain hasn't synced yet. Every one of those is something I now know is normal, not a sign that I'd broken something irreversibly.

When I hit roughly the mainnet equivalent of some of these — a UID that didn't immediately appear, a command that failed because I'd used a flag with a dot instead of a hyphen — I didn't have a troubleshooting table in my head. I had a transaction that had already gone through and a creeping worry that I'd done something to actual money that I couldn't undo. The error was completely recoverable. My emotional response to it was not proportional to that fact, because I had no calibration for what "normal mainnet friction" looks like versus "you've made an expensive mistake."

That calibration — knowing which errors are Tuesday and which errors are actually bad — is exactly the kind of thing testnet exists to build cheaply, before you need it to hold up under real stakes. I built it the expensive way instead, one anxious Google search at a time, with real TAO sitting in the balance the whole time I was searching.

What I'd actually tell a beginner, if I'm being honest about my own order of operations

I don't think the lesson here is "never take risks" or "testnet is mandatory before you're allowed to have an opinion." Plenty of people probably go straight to mainnet and are fine. But if I'm being straight about what I'd tell someone starting today, knowing exactly how my own sequence went, it's this:

Notice when you're skipping a step because it's "in the way" of your actual goal. That's precisely the kind of skip that doesn't feel like a decision while you're making it. If a step in the curriculum looks like a detour from the subnet you already want to mine, that's worth a second look, not a faster skim.

Testnet isn't a worse version of the real thing — it's the same thing with the stakes temporarily removed. Same commands, same prompts, same metagraph, same immunity mechanic. The only thing missing is consequence, which is exactly the thing you want missing while you're still making first-timer mistakes.

Calibration is worth paying for in time, not money. Every error message you've seen before on testnet is one less moment of real panic on mainnet. That trade — minutes now for composure later — is a genuinely good one, and I made it the wrong way around.

Being further along doesn't mean you've outgrown going back. I'm not writing this from before my SN41 experience — I'm writing it after, having gone back through the curriculum and noticed the gap in my own sequence. There's no rule that says testnet only counts if you do it first. Going back to register on NetUID 1 now, after already understanding what mainnet actually feels like, is still a real and useful thing to do, and it's genuinely next on my list — not as something I'm claiming to have already finished, but as the step I'm going back to close before I touch another mainnet registration.

The bigger thing I keep relearning in this camp

If there's a thread connecting this post to the other things I've written about Bittensor, it's that almost every uncomfortable lesson I've had so far was one the ecosystem had already tried to teach me gently, in advance, in a form I skipped past because it didn't look like the urgent part.

The curriculum put a free, consequence-free practice subnet directly between "you have a wallet" and "you're registering for real," and I walked past it because I already knew where I wanted to end up. I don't think that makes me unusual — I think it makes me exactly the kind of learner this specific step was built to slow down. I just didn't let it.

If you're somewhere earlier in this than I was — wallet created, mnemonic written down somewhere safer than a screenshot, itching to register on the subnet you've already decided is "the one" — this is me asking you, slightly sheepishly, to go to NetUID 1 first. Not because mainnet is dangerous in some dramatic sense. Just because the practice round is sitting right there, it's free, and the only cost of skipping it is finding out later, the way I did, exactly how much you didn't know you didn't know.