I Made The Perfect Magic Deck With AI (Chat GPT)

[@jacobyspurnger8488]

AI doesn't assume you have infinite money, it just assumes that you're going to proxy the whole thing. This is because in it's infinite knowledge, it knows that giving any money to WotC/Hasbro is a huge mistake.

[@ChuckleDuck]

Which is why I find channels that obsess over budget and cost kind of annoying, even when its a joke or joke deck. I will never give one red cent to WoTC or Hasbro and I have no idea why anyone would. So just give me an idea of what the BEST cards for what is trying to be done. Mentioning both the best choices and the budget choices is great. I know it is all in good fun and should not be taken too seriosuly, but when people are like, "and we're not going to discuss cards over $50.00, because that isn't budget friendly," that's really annoying for someone trying to learn that plans to proxy everything or play online. Like a sane person.

[@iaoshua]

buying cards on the secondary market doesnt benefit WotC at all, it incentives card scalping and market manipulation. proxy everything

[@Gabriel64468]

based AI

[@ahriahvanor6690]

@chuckle duck I like the budget deck options because I love actually buying the cards I play with. My dad taught me to play magic when I was just 5 years old so I buy the cards anyway. Having both would be better than just the cheap options, but not everyone proxies everything

[@electricangel3161]

How do you proxy cards?

[@theosw6961]

I corrupted chat GPT with the elf brain. For the last several days I have asked it nothing but suggestions for my elf decks. I’m glad to see the effects on full display here.

[@Gabriel64468]

I don't think chat GPT is currently updating itself, the website says the training was finished in early 2022. We have also seen multiple times that chat bots degrade really quickly if you have them continue learning from their interactions after release, because a lot of people just go to them to fuck around.

[@behemoth9543]

​@@Gabriel64468 People try their goddamn hardest to ruin them. Its impossible to make an AI able to learn from the internet and not have 4chan convince it of the benefits of racial cleansing within a week.

[@darkbodhi1416]

@@Gabriel64468 the most recent update was February 13th, 2023. It still updates as OpenAI hasn’t completed their goal/vision for this AI yet. Not arguing, just thought I’d let you know.

[@BrainwashingDetectiv]

It's him. Elfs Georg.

[@dividedbyzero11]

Knife-ear.

[@SaintHills781]

I love that the AI also recommended two orzhov partner commanders for an abzan deck

[@RakdosRabbit]

This inspired me to try to get chatgpt to design a CEDH deck, and it was pretty close but it included 2 demonic consultations which honestly is a great idea and I don't know why anyone hasn't thought about doing that yet.

[@Nightrunner116]

"The secret ingredient is crime" ~ Chat GPT

[@youenhime2770]

the second demonic consultation is actually plunge into darkness.

[@peewee0224]

Honestly the ai’s deck is actually insane because it’s not trying to build around the commander it’s trying to play a Abzan good stuff deck and use the commander to get back your big threats.

[@p.m.e.7311]

Strange. That is my deck building philosophy

[@peewee0224]

@@p.m.e.7311 yeah it’s a good deck building philosophy unless you wanna be thematic

[@steventyler3582]

Always the best philosophy. I do have a few decks that are very reliant on my commander but the second my commanders gone the pressure is off and that just sucks. Now when I build I just make sure my commander synergizes with my theme.

[@peewee0224]

@@steventyler3582 fun tip in some decks youre going to have a couple commanders that fit that theme. what you can do is buy them all then either switch them out to raise and lower the power level of your deck. or just randomly shuffle them up and take 1 out. it adds a ton of variety to just one deck

[@steventyler3582]

@@peewee0224 Thats a fun idea. I might try that with a couple of my decks. Kinda hard with something like an orvhar deck though where the deck is literally built around the commander though.

[@talonarayan]

"You can only expect so much from perfection." 8:40. Phyrexia would like to talk.

[@fafs67]

Persfection

[@epsil0gue246]

Aura Shards actually does synergize with your commander. Because of the fact that you're bringing creatures onto the battlefield with Karador and Aura Shards blows up an artifact/enchantment when the creature enters the battlefield, you basically have a passive removal option for enchantments and artifacts.

[@stephenhurley1033]

You arent wrong, I think if I saw the commander or had built the commander my self I would have included a few creatures that etb destroy artifact/enchantment and work towards having them repeat entry (this takes out less but you will hit the key threats). The main advantage being you want your creature count up to synergies with the rest of your deck direction. That being said Aura shards isnt embarrassing to bring out at all 😂

[@Mostexcellant69Dude]

yeah you really pegged it, i was thinking the same thing aura shards is just value once its out. plus he has all those low cost creatures, its triggering every turn.

[@bronsoncarder2491]

@@stephenhurley1033Idk, the advantage of using aura shards here is that it's one card that can be activated by any creature. So, you don't have to spend any of your valuable creature count on artifact/enchantment control. I feel like it has really good synergy here, actually.

[@stephenhurley1033]

@@bronsoncarder2491 this might be meta depending? My group that I play with have really reliable enchantment removal so for me to include Aura shards in any deck I need to support it with slots that will bring enchantments from the grave which in my meta will end up being about the same amount of slots as it would be creatures that have the removal etb. I think I'd be better off having the creature count and including ways of abusing creatures etb's given it's the main deck focus. If my meta wasn't going to remove Aura shards reliably and quickly then my thinking would change but as a deck builder I have to build with resistance to the opponents I will face ☺️

[@Leviwulf]

Garruk was a great choice because it gives you a ton of wolves to sacrifice with your other cards to power up your planeswalker. AI was just 10 steps ahead.

[@ivandominikkresonja4116]

Whenever you aren't sure why AI did something, why don't you just ask it? Chat MTG can respond to your questions and elaborate on its previous answers

[@murpl1462]

and it will tell you that it decided to include terrastodon because it is a 5/5 for 4 that gives you strong early board control and draws cards

[@ChaoticNeutralMatt]

Chat MTG?

[@burningangel161]

ChatGPT can't elaborate on shit, because it doesn't actually know anything. What it can do is give you a slurry of regurgitated text that may or may not coincidentally be interpreted by you as correct, thereby convincing you that it is "elaborating" on its "choices".

[@GrumpyIan]

Now we need you to combine the ai that made the custom magic cards and Chat GPT to balance the generated cards.

[@ENCHANTMEN_]

I got chatgpt to make a breaking bad card game and some sample decklists, then made it write the transcript to a match between yugi and columbo

[@Zyjem]

Please tell me meth was the mana equivalent

[@judahthejwizard2734]

The reason it gave you so many sac outlets and only a garruk as a good sac targets is because your commander reanimates so you don't need as many good sac targets that don't do anything but be bodies. It's actually smart and the bot probably believed you would sac your sacers, and reanimate them one by one while annoying your opponents with removal like aura shards and get reinforcements like with that 217 dollar enchantment. It's a cool deck and if I was Elon musk, I'd build it.

[@murpl1462]

I like the idea of a sac deck built around sacrificing and reanimating ETB creatures instead of the standard token spam, so maybe Chat GPT was going for something like that.

[@DeadlyGrim]

It's not really "thinking" anymore than, say, the algorithm behind EDHREC is "thinking". ChatGPT has been fed a trawl of text from the internet. The ChatGPT neural network then uses this text to create connections to predict what the next character in a sentence is (similar to, say, predictive text except way more complicated). It uses your prompt and previous messages it's created to create context and further adjust predictions. It's not really evaluating cards or the deck as a whole. That's why it suggested illegal cards, duplicates, etc or gave the wrong description for cards. What it "knows" is that certain strings of text appear more often near other strings of text, so it reads "Karador" and then outputs cards that are mentioned in text that mentions Karador. Since all of these are predictions at the character level, sometimes it predicts wrong. As I mentioned, there's some similarity with EDHREC. EDHREC's algorithm looks at Magic decks, sees that certain cards are found more often together, and then uses that to create "synergy" suggestions and the like. (The algorithm behind EDHREC is also way less complicated but since it's so specialized it's arguably more powerful. Though, that's a whole different conversation.) Side note: notice how it writes card names in double brackets (e.g. "[[Karador]]")? That's used on Magic sub-reddits because it summons a bot which provides Gatherer, Scryfall, etc links for the card that's in double brackets. And it's doing this because places like reddit is where ChatGPT gets a lot of it's text.

[@dr.limeade3035]

@@murpl1462 If thats a deck idea you like, Alesha, Who Smiles at Death is a really good commander for that

[@judahthejwizard2734]

The reason it gave you so many sac outlets and no sac targets is because your commander reanimates so you don't need as many good sac targets that don't do anything but be bodies. It's actually smart

[@collinbeal]

Yeah and it still gave a couple easily recursive fodder in case your commander isn't out. Brain size immense.

[@KodamaoftheWeastTree]

Fr, the bit actually made a good deck.

[@amberhernandez]

ChatGPT clearly overlooked the mighty Mycoloth when going over sacrifice outlets! You can sacrifice any number of creatures upon summoning it, and then generate TWICE that many Saproling tokens on _each_ of your turns!!

[@amberhernandez]

It DID generate Fathom Mage, however, proving that ChatGPT is just 100% based

[@murpl1462]

Maybe Chat GPT doesn't really like mycoloth because if you sac a bunch of stuff to it and then it dies before your next turn then you just feel really bad.

[@amberhernandez]

@@murpl1462 Thankfully, Abzan is a trio with two of the _best_ protection colours in the game in green and white! If you can spare the mana for a Heroic Intervention, Tamiyo's Safekeeping (budget lol), Blossoming Defense, or even a Teferi's Protection, you've got a beautiful value machine starting! Mycoloth is my secret commander in my Prossh deck, hahaha!

[@BlackfeatherMain]

@@murpl1462 that never happens

[@kagesora007]

This actually gave me a lot of good ideas, so really, the AI was using you to help multiple other people. Net gain. Big brain play. I don't see any problems here.

[@TherosonDragonstoungue]

The "sac" in this deck is literally sacrificing your M&M's to the dark abyss of your graveyard during combat phases.

[@javierl.2377]

i came for the 50 rats. Where are the 50 rats.

[@boneytonez]

Also he questions every etb trigger card in a reanimator deck🤣

[@collinbeal]

I know why he rated Abzan so low now. He understands all of Magic the Gathering, but is slightly off in his understanding of Abzan.

[@boneytonez]

@@collinbeal absolutely. If I’m playing Ghave, my group shivers

[@Alfred_Benson]

In practice, losing your free will to compleation actually means having all your decisions be taken by open source ai

[@Zwork101]

Painful Truths is actually quite good in three or more color commander decks

[@kathleenbirrell3222]

Honestly, still better than stuff I’ve seen people build

[@sb0ng3]

CONGRATS ON 100K NOAH, I HAVE BEEN FOLLOWING THE CHANNEL FOR A WHILE AND AM SO HAPPY TO SEE THIS MILESTONE HIT FOR YOU!!!. lots of love from canada

[@israeldavila27]

I want someone to please build this deck in a cockatrice and test it out. There has to be footage of the deck. Clearly it is superior. Just decide this was funny as heck.

[@Rammkard]

Things I learned from this video chat gpt is better deck builder than most people that I help build

[@strikermodel]

"I'm not reading because it's upside down" You almost made me spit my drink

[@aragonaut]

I just straight up asked ChatGPT for a decklist for an EDH deck and this was what it gave me: Commander: 1x Tasigur, the Golden Fang Lands: 1x Command Tower 1x Mana Confluence 1x Exotic Orchard 1x City of Brass 1x Overgrown Tomb 1x Breeding Pool 1x Watery Grave 1x Hinterland Harbor 1x Woodland Cemetery 1x Drowned Catacomb 1x Sunken Hollow 1x Polluted Delta 1x Verdant Catacombs 1x Misty Rainforest 1x Marsh Flats 1x Flooded Strand 1x Bloodstained Mire 1x Wooded Foothills 1x Windswept Heath 1x Temple of the False God 1x Bojuka Bog 1x Reliquary Tower 1x Evolving Wilds 1x Terramorphic Expanse 4x Swamp 4x Island 3x Forest Creatures: 1x Eternal Witness 1x Mulldrifter 1x Ravenous Chupacabra 1x Shriekmaw 1x Acidic Slime 1x Fleshbag Marauder 1x Sakura-Tribe Elder 1x Farhaven Elf 1x Coiling Oracle 1x Snapcaster Mage 1x Torrential Gearhulk 1x Sheoldred, Whispering One 1x Jin-Gitaxias, Core Augur Artifacts: 1x Sol Ring 1x Dimir Signet 1x Golgari Signet 1x Simic Signet 1x Commander's Sphere 1x Chromatic Lantern 1x Crucible of Worlds Enchantments: 1x Sylvan Library 1x Pernicious Deed 1x Grave Pact 1x Dictate of Erebos 1x Rhystic Study 1x Necropotence Instants: 1x Brainstorm 1x Ponder 1x Preordain 1x Mystical Tutor 1x Vampiric Tutor 1x Demonic Tutor 1x Diabolic Intent 1x Go for the Throat 1x Hero's Downfall 1x Cyclonic Rift 1x Counterspell 1x Swan Song 1x Negate 1x Reality Shift 1x Beast Within 1x Putrefy 1x Sultai Charm 1x Abrupt Decay Sorceries: 1x Buried Alive 1x Entomb 1x Life from the Loam 1x Toxic Deluge 1x Damnation 1x Villainous Wealth 1x Exsanguinate 1x Torment of Hailfire 1x Praetor's Grasp 1x Mind Twist 1x Army of the Damned

[@littlemisspipebomb4723]

I put this list together and it's surprisingly coherent? Obviously pulling a lot of cards purely because they're popular, but it's honestly not the worst I've brewed

[@mitchelllayzell2553]

i just found your channel a few hours ago and subbed because of the last video, holy shit not even a minute and I get new content

[@themadgamer4571]

Ever since Karador came out he's always been my favorite commander. This video is very satisfying for me.

[@Astraljinx]

The chat bot gave a really good abzan deck that had a lot of synergy with other cards in the deck with the commander being a buffer, personally i think he is just being satirical because a player of his dedication would likely have seen the trend

[@P3rsonL]

Noah, what was the final cost of all the cards it picked?

[@eugenecaudill5361]

Idea for a video. Have Chat GPT build a 5 Color Frog Tribal deck, and name it Alex Jones Gay Frogs. It doesn’t have to be good. It just has to be fabulous.

[@israeldavila27]

F R O G S 🐸

[@teamcosmicseeds]

1ST...if the AI has spoken...where the hell is the joiner adept?

[@teamcosmicseeds]

I didn't see a mox diamond either. You are playing more than 1 color correct?

[@bronsoncarder2491]

Several of these that you say "have no synergy" have fairly clear synergy... Best example is Aura Shards. In a deck like this, creatures will be entering the battlefield under your control almost constantly, and sometimes at instant speed. So, in this situation, Aura Shards is basically just a free ban on the opponent using artifacts or enchantments. lol I actually think (with a bit of tweaking, of course) this is an outline for a pretty awesome deck. Sure, you only have two cards that are built for saccing, but I feel like the point here is that anything can really be used as sac material. Depends on what kind of advantage you need at the moment, relying on your commander to get anything good back out into play by the time you need it again. In a deck like this, having good cards in the graveyard can be even more valuable than having them in your hand. So, it ends up being less about direct synergy with the commander, and more about cycling through good cards to gain moment-to-moment advantages. Granted, I didn't read but maybe a dozen of the cards it put in here, it's possible it needs a lot more tweaking than I'm thinking. lol But, it's at least a decent *idea* for a deck. lol

[@Russian_engineer_bmstu]

Yavimaya granger honestly isn't that bad, seeing as it sacks itself to be replayed later if needed.

[@SpacePirateFromSpace]

I actually really like abzan but the commander options aren’t what you’d expect. You’d really think there’d be a keywords-in-graves matters commander in abzan but no, it’s in sultai because of course it is

[@danielkemp6249]

Well, don't forget about Kathril, Aspect Warper. He's exactly what you want for an Abzan keywords-in-graves matters commander. Dump your keyword soup in the graveyard and have an almost untouchable commander that one-shots each opponent.

[@SpacePirateFromSpace]

@@danielkemp6249 holy shit how did I not know about this

[@4lukeification]

I can help you understand. You have given it multiple different goals. First make the best deck. Second make it abzan. Third make it karador. It's trying to meet all three goals when you really only care about the first one. The best abzan decks are Protean hulk combo decks. It's basically making a Protean hulk deck, but by the time protean hulk was unbanned karador was sorta on its way out of the meta. It's training data reflects this, so when you ask for the best deck and also tell it to make karador it's just sorta combining both those goals. You see this with pattern of rebirth and karmic guide, cards used almost exclusively with protean hulk combo. A lot of early karador decks ran the high cost creatures like avenger or zendikar and would often loop things like to replay with karador. Sacrificing things like terrastodon and replaying them with grave pact really destroys opponents board states. Painful truths and read the bones are meta choices for card draw in abzan. As for your Planeswalkers it's getting it confused with the garruk that flips into survival of the fittest.

[@Mcgoolio]

Painful truths is actually a very efficient card draw spell because you’re in 3 colors

[@BlackfeatherMain]

you know what I actually approve of the Painful Truths. It's a great draw spell in tricolor because 3 cards for 3 mana is a hell of a value.

[@jimbo3]

the first creature it recommends is an enchantment that costs $217, i would recommend. lmao

[@chadwalston28]

Okay but I'd like to see a game from the all knowing AI deck 😂

[@NamelessFacelessWhoa]

“And a card I’m not reading cuz it’s upside down” fucking sent me

[@colinmcvicar2403]

As someone who made a Karador deck I can say chat GPT is not wrong it is the best deck I own

[@meanauntrena4667]

That is a fascinating deck, a bit loopy but you can sac those lanowar elves to give your opponent a bad day.. But you would have a hard time ensuring you can build that machine...

[@m4dd0ck80]

Strangely enough, there were some weird card choices that actually had some synergies with the commander. Though, a lot of the cards made no sense with the commander in all fairness.

[@Tharkon]

30-35 creatures, 5-10 instants and sorceries, 5-10 enchantments and artifacts, 1-2 planeswalkers and 35-38 lands. Is it recommending you make a deck that has 76-95 cards?

[@tomhanks4899]

Great choice for graveyard plays

[@LoveisLight-n9r]

chatgpt said my standard deck is well thought out and balanced.

[@thatguyintherain3168]

Look at that thiccc 100k! Congrats to you man! All hail Phyrexia!

[@randymotter51]

The Aura Shards is actually brilliant because any recursion or mass recursion can result in a ton of hostile artifacts and enchantments being removed and that's on top of normal casts. This AI has a few misses but honestly I've seen (and myself have made) worse decks than this.

[@iclimbeverything2990]

50 rats? Sounds like a perfect option for gishath

[@Aventrixmusic]

I like child of alara because you can choose whether or not to nuke the entire board every time it leaves the field

[@Garl_Vinland]

The Ai was being honest with you. Expensive foodstuff decks are the best way to build decks.

[@quintessenceoflife7870]

Congratz on 100K subs Noah

[@atopax]

ChatGPT's knowledge ends in early 2022 so it's also likely out of date in terms of what it thinks is a good commander deck, including its price knowledge which is likely either older or just made up. I will say the Bing AI gave me some pretty fire suggestions but it can cheat and actually search the internet.

[@LonePC]

its all based on playing the creatures from the grave yard and the battlefield. the commander is key here. it all makes sense, so go ahead and sac the creatures because you can play them again.. at least that's what i got from it

[@DurdleDers]

Congratulations on 100k Noah!!!

[@LucasDenhof]

Idea: I'd love it if a future video has like a live slide of going to ChatGPT and asking for a card

[@notabene9804]

See, idk why ppl think AI will overtake. Out of all the colors, it picked Abzan. Maybe we're the problem for going Izzet.

[@T.BG822]

So there's a step you missed. Every good deckbuilder plays a deck then revised based on how it played: you need to provide the decklist to the AI again, alongside your estimation of strengths and weaknesses, telling it to iterate and edit the deck for reliability and strength. In general, it'd be good not to stagger prompts as much, as well. It'll have an easier time retaining info within its own response than comparing to previous responses. Frankly, it gave you a good deck with a bit of a wonky mana curve and some backwards synergy. It's not bad by any means for casual commander.

[@riverlodes3129]

I'll subscribe when you give white the appreciation it deserves

[@CrabTribe]

The abrupt endings never cease to surprise me! “Please subscri…” is a classic cliffhanger!

[@EDoyl]

if you have enough sac outlets, you don't need to add anything specifically to sacrifice. Just sacrifice your excess sac outlets.

[@Terrakinetic]

The sarcasm is too heavy handed

[@speedysloth2868]

congrats on 100K

[@coder1852]

You can get better results by saying to build a commander deck with the specifications

[@alexrobinet7576]

This could easily be a big video fixing the gpt deck.

[@maclainwright3338]

unironically? not an awful deck, i approve

[@dynesaurus8264]

Beezy knows what he is talking about

[@vincentbarwell5943]

you can just ask it to make a commander deck then it will ask you a bunch of questions then just say I want you to decide that and it somewhat synergises

[@anonymousaxolotl]

next ask it to make custom magic cards lol, also CONGRATS ON 100K

[@DAZEINNIT]

Dude, congrats on reaching 100k! This is sick even though you don’t like white

[@AparnaModou]

Jund fan here! Tried to generate an alternate art for my Korvold deck on Bluewillow. Will be using it for as my commander soon!

[@zeo1987]

The end got me. 😂

[@mr.mammuthusafricanavus8299]

LOL Everyones favorite Tear-ass-odon ;p

[@faztznya5207]

Noah I never played commander, when are you going to show us some high quality tutorial?

[@ms.aelanwyr.ilaicos]

Karador is the perfect commander because it was the first Commander ever spoiled.

[@thisHertz]

Not 0%. I must be the only one subscribed. Your videos are hilarious. How can I not be!?

[@chy03001]

I would love to see you play this deck

[@capeciucc5268]

when an AI will be able to build a good commander deck, the human kind will probably be olready slave of the machines

[@Venomtankmod]

I mean it has decent synergy for mid game. Garuk spawns wolves with death benefits. You sack the wolves. You profit

[@KumaGorath]

karador commanders rise up.

[@ultratog1028]

You should fight the chat gpt deck vs the Multani KZbin comment deck

[@firesean_]

I asked ChatGpt about a yugioh deck, and it gave me the entire banlist as a deck.

[@johncox8835]

Spore Frog plus the commander is a wombo combo.

[@alwayrember193]

i dont even play magic but this is entertaining

[@facepalmjesus1608]

i will share my genius deck with you... Encroaching Mycosynth + Mishra Prodigy Artificer in play ------> Cast a Rat Colony thank me later

[@thebrentster2563]

I saw your intern over at The Commander's Quarters make a video just like this. The guy just can't help copying your ideas.

[@dragonboyjgh]

Your commander returns things. ANYTHING you don't need that turn can be sacced. Hell your commander can be sacced, as long as you got at least 2 more creatures in the graveyard since the last time it died it's not any more expensive.

[@Ris-vulpis]

My brother has played mtg for 15 years and has taught me to play about 6 times. I have no idea how to build a deck. This is making me want to build a killer deck and destroy him next time I see him.

[@samanthaclark2863]

This deck seems p sweet honestly

[@deusexmachina3047]

Made it here quite early. Good video.

[@jean-francoisseguin2569]

I pissed in my pants laughing lool!

[@sebastianhogan4875]

I really want to see you play with this deck

[@faxmachine1162]

I will say, you think it doesnt have anything to sac, but I think this deck is about ETB effects, using your commander to get multiples of those triggers.

[@malmasterson3890]

This deck just from the template is scarily close to my Thalrog deck I pre-brewed on Moxfield. Except I have slightly less creatures & more sorceries.

[@benjamincarter6095]

Prompt engineering required

[@hallaloth3112]

LOL, as someone who has a (very sad) Abzan deck. . .seeing the first it suggested . . .I have used both of them for that deck. Currently sticking with Anafenza. . .-sob- why are there no Abzan commanders that play into +1/+1 counters? Black has some lovely cards for it that I just can't go solid G/W

[@elosyyy]

dude, i reach the end and i was expecting some gameplay. feelsbadman

[@kassandra_sae4563]

Hey Noah! I've just started playing mtg, is this a good first deck to start playing with. Too bad, I've already bought all these cards anyways sooo...