Six months of Spellweave

Six months ago Spellweave was one release old and had a handful of people using it. This week the two-thousandth Spellweave user signed up.

A lot of what follows started with someone telling me it was missing. So before anything else, thank you.

Where things stand

Since the first release on 2 February, more than 2,000 people have signed up, built or imported over 4,300 decks, and catalogued more than two million cards into binders.

Deck import launched at the start of April. By June, more decks were coming in from other sites than were being built from scratch, and it has stayed that way since. That surprised me, and a lot of what followed was built around it.

Importing is rarely the end of it. Nearly two thirds of all Deck Tunes have been run on an imported deck, and imported decks account for most AI Analysis runs too. People are bringing a list they already play and asking what it is actually doing, which is a different question from "build me something".

You asked, I shipped

Thirty feature requests have come in through the board and Discord. Twenty-one of them have shipped, including:

• Lock specific cards into a build and have the engine work around them
• See why each card was included, rather than trusting a list
• Partner commanders, including Backgrounds, Friends Forever and Doctor's Companion

Five more are open, four I said no to with a reason, and twenty-one bug reports found edge cases that only appear in a collection shaped differently from mine.

Sixty-one releases in six months. Most are small on their own: a mechanic the engine could not read, a card type it scored wrong, a screen that took one click too many. Together they are the difference between a list of good cards and a deck that does something.

If you have not opened it since April

Here is one loop worth half an hour.

Import a deck you already play. Paste the list straight in, or drop in a Moxfield or Archidekt link. File exports work too.

Run Deck Tune. Point it at a card pool first: your own binders, or All Legal Cards, which is what most people pick. You can cap the price per card, and anything above the cap is simply never considered. Mark the cards you refuse to cut, so nothing suggests cutting your pet card.

Then pick how hard to push. There are two modes now: Balanced keeps the deck in its lane, and Power focus trades solid on-plan cards for top staples to raise the ceiling. Either way, read the Bold suggestions rather than the Safe upgrades: each says which guardrail it trades off, and if a swap would break one of your combos it names the combo and makes you confirm.

Run an AI Analysis when you want the read on the whole deck rather than a list of swaps. It opens with Mana Reliability, your exact list simulated a thousand times for hands kept, commander on curve, tapped-land drag and color trouble. The Rule Zero Brief comes next: an honest pre-game pitch for your pod covering the combos, fast mana and game changers worth flagging before anyone shuffles. From there it works through mulligans, sequencing, combo lines, resilience, what opponents can exploit, and upgrades from your own binder scored against the whole deck.

Open the Deck Lab. The Goldfish Test runs the deck about a thousand times and gives you a keep rate and a speed timeline: the median turn your commander lands, the median turn your engine comes online, and the median turn you present something an opponent has to answer. It is not a win rate. It answers what goldfishing is actually for: once you keep a hand, how fast does the deck do what it is supposed to do?

That whole loop existed in pieces in April. It did not work like that.

Building from your collection rather than importing? Open Collection Potential on the deck. It scans your binder for that commander and tells you the highest bracket your own cards can reach, which combos you already own the pieces for, and which game changers you are holding. This is the part I would not try to do by eye. Combos hide across old cards, reprints and odd wording, and when it finds one it gives you the prerequisites and the steps, not just the card names.

Also new since then: Spanish and Japanese across the site, builds, card and token names, and now email.

The Hobbit is live

Prerelease is running now and the set releases on 14 August. It is legal for Commander now, and Spellweave already builds with it, including storied and recruit.

That last part matters more than it sounds. When a deck builder cannot read a brand new mechanic, it falls back on generic signals and you get a deck that misses most of what the commander is asking for. Storied counts artifacts, legendaries and Sagas, so build around Balin or Dáin today and the deck comes back stocked with them.

You can start before you own a single card from the set. Pick a Hobbit commander you do not own and Spellweave builds around it from the cards already in your binder, the commander being the one exception. The Recommendations tab ranks up to fifteen cards you do not own that would improve it, and Deck Tune does the harder job of proposing the actual swaps. Crack your prerelease packs, add them, build it again.

Build or tune a deck

What is next

The engine keeps getting corrected when real decks expose edge cases, and most of those fixes start with someone telling me. Deck Tune has more coming, and Standard support is in progress.

If something is missing, the feature request board is the fastest way to reach me.

Spread the word

Spellweave gets better mostly because people tell me what is wrong with it, so more players genuinely does mean a better tool. If you have somewhere you talk about Commander, a pod chat, a Discord, a subreddit, mentioning it there does more than anything else I could do.

There is also a referral link on your account page. Whoever uses it gets 5 bonus mana on their first purchase, and you get 3.

Thanks for six months. The next update will not take another six.