ManaDork
Join the waitlist
Pioneer · Golgari Demon Engine · built from one player's collection
One real session, start to finish, with the moments that matter marked up. Every quote below is verbatim from the transcript — nothing rewritten, nothing tidied. The interesting part isn't just the deck list. It's what it says about the deck list.
You · 01:05 PM
let's build a deck around dreadfast demon and virtue of persistence — the plan is to have cheap and midrange creatures with powerful etb and/or ltb abilities… a nice bonus for the deck would be a way to cheat the dreadfast demon in early.
→ No card picker, no dropdowns, no correct spelling. The real card is Dreadfeast Demon — one letter off, and typed wrong twice.
ManaDork · 01:06 PM
Pick a direction — reply with the name of the one you want. I'll build it from your collection only.
1. Mono-Black Persistence Engine — cheap ETB/LTB value bodies and
self-reanimators feed Dreadfeast Demon's end-step sac for Demon copies…
2. Golgari Ramp into Demons — Green acceleration and land ramp
cheat Dreadfeast Demon and Virtue of Persistence ahead of curve…
3. Golgari Food Loop Value — Cauldron Familiar + Witch's Oven
create a low-curve death loop that drains…
→ Three genuinely different shells off one prompt, each with its own plan and key cards. It doesn't decide which deck you meant — you do.
ManaDork
Heads-up: the name Dreadfast Demon matched no card in the database or your collection, so nothing by that name is in these directions. If it matters to the deck, check the spelling and tell me.
→ This is the whole difference. A model that autocompletes would have produced a confident Dreadfast Demon with invented rules text. It said the card doesn't exist, built around the real one, and told me to check my spelling.
ManaDork · 01:11 PM
Curve note. CMC 1–2 is deliberately high for the Food/sac loop and dorks (engine density over textbook midrange curve). Top end is 3+3 of the two build-arounds (above the default 7+ cell) because they are the win condition. CMC 4–6 is light: Sheoldred is cut for copy limits/space; value sits at 2–3 and 7.
→ A 1.6 average curve with six seven-drops would fail any generic deckbuilding heuristic. It knows that, and explains the trade instead of quietly smoothing the curve into a deck that no longer does the thing I asked for.
ManaDork
Collection caps used: Familiar own 1, Oven own 3, Thoughtseize own 2, Zulaport own 2, Mosswood own 3 (playing 2), Greta own 4 (playing 1 for curve), Demon/Virtue own 4 (playing 3). Basics free.
→ One Cauldron Familiar, because that's how many I have. Note Greta: owned 4, played 1 — held back for curve reasons, not availability. It distinguishes "can't" from "shouldn't".
ManaDork
Roles. Spot removal: Fatal Push (revolt off Oven/loop) + Heartless Act. Draw: Dispute + Greta Food sac. Ramp: full dork package. Disruption: Thoughtseize. Recursion: Mosswood adventure, Virtue, Demon tokens.
→ Fatal Push is in the deck because the Oven turns on revolt — the removal was picked to work with the engine, not dropped in because it's a good black one-drop.
You · 01:27 PM
that played decent ty, but it needs more tutor or draw or search, i got one virtue very late game and no dreadfeast demon
→ No form, no rebuild button, no starting over. A sentence about how the games actually went, typed the way you'd say it to a friend across the table.
ManaDork · 01:27 PM
Fix: more dig and hard tutors so you actually see Dreadfeast Demon / Virtue of Persistence.
In: 4 Deadly Dispute (Food/Oven fodder → 2 + Treasure), 2 Diabolic Intent (sac a token/dork → any card, including Virtue), 2 Finale of Devastation (tutor Demon onto the battlefield), 1 Phyrexian Arena, 1 Woe Strider (scry on sac), 2 Castle Locthwain (late draw sink).
Out: 2 Fatal Push, 2 Heartless Act, 1 Greta, Sweettooth Scourge, 1 Vinereap Mentor, 1 Prosperous Innkeeper, 1 Zulaport Cutthroat, 1 Forest (trim to 60).
Spot removal is lighter (~5) — intentional trade for consistency.
→ It diagnosed the actual complaint — not enough ways to find the two cards the deck is built on — and paid for them by name. Then it told me what the fix cost me, instead of pretending the new list is strictly better.
Everything above is quoted from one conversation. The full transcript — both deck lists, every card, the three strategy directions and the auto-correction notes — is public:
ManaDork is chat-driven deck building for Magic: The Gathering — Constructed, Commander and Limited, and it can build from the cards you actually own. See more at manadork.ai.