Sumit Shinde

Robot body. AI brain. Crypto wallet. 7 years in crypto, now I can't stop thinking about robots. Early team @0xPolygon.

[Sumit Shinde @sumitships]

You may wonder why a crypto person is suddenly learning robotics. It is a fair question, so here is my honest answer.

I think a machine that does real work on its own needs three things: a body, a brain, and a wallet.

The brain is AI. This part is moving the fastest. I do not build it. I just use it every day.

The body is a robot. I want to be careful here. I am not saying AI needs a body to be smart. Smart people disagree on that, and I do not know. I am saying it needs a body to do physical work, and most of the world's work is still physical. Someone has to move the box.

The wallet is crypto. A working machine will need to pay for things: charging, spare parts, data, help from other machines. It will also need to get paid. A robot cannot walk into a bank with a passport. But it can hold a key, and a crypto wallet is just a key. It also makes very small payments possible, like two cents for one map.

I could be wrong in a few ways:

1. Banks and card networks may build good accounts for machines. Then you do not need crypto for this.
2. The law may always want a human or a company behind every wallet. Then the machine only borrows its owner's wallet.
3. Useful robots may take much longer than people hope. Hardware is slow and hard.

Even if point 2 turns out true, machines will still need rules for spending: limits, money held until the job is done, and payment only after the work is proven. That is the part I work on.

I spent 7 years in crypto, including the early team at @0xPolygon. So I know the wallet. I use the brain daily. The body is the part I do not understand yet. So I am learning it the only way I know: by building robots in public, including the ones that fall over.

I am a beginner here. If you build robots, I would love to learn from you. If you think this idea is wrong, tell me why. Both will help.

[Sumit Shinde @sumitships]

Crypto is finally winning the way it was supposed to: by disappearing into the product.

Look at what's actually getting used. Polymarket runs on Polygon. Robinhood's tokenized stocks run on Arbitrum. AI agents are getting their wallets on Base. All Ethereum. None of their users know or care.

I got in when 40 dollars of gas made a 20 dollar transfer pointless. Polygon was the first bet that Ethereum scales by getting out of the way. That bet is now the default architecture.

Turns out the goal was never to get people to use blockchains. It was to build things so useful they forget there's one underneath.

Exciting times ahead.

[binji @binji_x]

>be me
>ethereum
>spawn in the head of a 19 year old who played a lot of wow
>he thinks the world should share a computer
>quite a large side quest but ok
>other people join him
>some are weird
>some are REALLY weird
>everyone has opinions about money and nobody agrees where to put the buttons
>somehow a whitepaper becomes an actual project
>ICO!
>people send bitcoin to help spawn me
>somehow this works
>boot up in 2015
>hello world :)

>now need people to actually run me
>strangers around the world start mining
>their computers get warm
>i get to exist
>nice arrangement
>little bits of me living in places my creators have never been
>already getting slightly out of hand

>people who have never met are keeping the same record of who owns what
>cannot get six friends to agree where to eat
>somehow getting strangers to agree which money belongs to whom
>turns out strict global consensus is occasionally worth the trouble

>someone tries to build an organization on me
>the DAO
>very ambitious
>gets hacked
>oh fuck
>everyone argues about what to do
>argue so hard there are now two ethereums
>difficult childhood

>2017
>someone realizes they can make their own token
>wait anyone can raise money for an idea now?
>ICO era
>every idea gets a token
>some of these seem quite promising
>some of these are literal dogshit
>everyone suddenly has an advisor
>nobody seems to have a product
>oh no
>oh no no no
>this was perhaps too much ICO

>but some people are actually making things
>@StaniKulechov starts building ETHLend
>wait, you can borrow money through a smart contract?
>put up collateral, agree to the terms, get a loan
>contract does not ask which university you attended
>does care quite a lot about your collateral though
>fair enough
>little lending application on the world computer

>meanwhile @haydenzadams loses his engineering job
>his friend @karl_dot_tech gets him into ethereum
>learns to write smart contracts
>starts making an exchange
>other people help
>not the usual advice from the unemployment office
>but let's see where it goes
>hayden starts making a decentralized exchange
>other people help
>keeps going
>"uniswap"
>a pink unicorn is now involved in neo financial infrastructure
>i like this place.

>wait
>the lending thing can connect to the trading thing
>and the trading thing can connect to the dollar thing
>someone else can build something using all three
>nobody had to arrange a meeting between three banks
>they just wrote more code
>oh

>OH
>the money can do stuff now.
>*defi summer*
>everyone becomes a farmer
>no actual vegetables anywhere
>people move money between protocols with names that make me hungry
>the yields are doing something concerning
>some of this is very clever
>some of this needs an adult
>but underneath it all
>regular people are actually figuring out how to build financial systems together for the first time ever
>borrow here, trade there, build something new on top
>a person who was never getting invited to design financial infrastructure is suddenly designing financial infrastructure
>probably worth paying attention to that bit

>then the picture people arrive
>NFTs
>artists finding collectors
>collectors finding artists
>everyone finding the right-click button
>long discussion about what exactly is being owned
>somehow this discussion lasts several years.

>people put animals on their profiles
>become friends with other people who have the same animals
>group chat becomes a community
>community starts making things
>also some absolutely horrifying financial decisions
>both can be true

>someone came to flip a picture
>stayed because they made friends
>someone else sold their art and realized a person on the other side of the world cared about something they made
>hmm
>was not expecting feelings in the jpeg department

>start imagining what else these little groups could do
>twelve people in twelve countries care about the same extremely specific thing
>previously their main economic activity was arguing in a group chat
>now maybe they can pool money
>pay contributors
>own something together
>try different ways of deciding what happens next
>still argue
>but now have a budget

>maybe they try to buy the US constitution
>maybe they fund research nobody else understands yet
>maybe they preserve their community's stories
>keep copies, fund the storage, keep a record of what came from whom
>not everything needs to be onchain
>but their culture shouldn't disappear because a website stopped being profitable

>meanwhile the protocol engineers have been planning an enormous change to how i work
>*the merge*
>stop mining
>switch to proof of stake
>coordinate the change across the network while it is still being used
>it works
>never goes down despite changing completely, 100% uptime secured
>energy use drops by about 99.95%
>holy shit
>little computer did a big thing.
>validators around the world keep me running
>some from home
>global financial infrastructure
>partly looked after by a person with a machine next to their laundry
>very fond of that person
>surely we can enjoy this for a bit

>FTX.exe

>wallahi we are cooked
>entire industry having the worst group chat of its life
>people who did nothing wrong get hurt
>actually horrible
>spend a moment wondering what any of this was for
> but somewhere a hackathon is still happening
>somebody has found a bug
>somebody else has fixed one
>a person with twelve browser tabs open thinks they can make the wallet less confusing
>ok
>stay with them

>check on the maths people
>they have apparently been doing maths through every single one of these events
>deeply reassuring

>ask what they have made
>"zero knowledge proofs"
>you can prove something without revealing the secret underneath it?
>yes
>and prove a computation was done correctly without everyone doing all the work again?
>also yes
>brother what
>why were we not louder about the wizard department??

>start thinking about what that could do
>prove an age requirement from a valid credential without sharing your whole birthday
>prove the relevant rule was followed without exposing every private detail
>i would like the system to be checkable
>i would not like every human inside it to be naked
>turns out paying very unusual people to think about cryptography is a pretty interesting use of money (thank you @ethereumfndn)
>maybe should do more of that
>someone opens another research tab
>we may not see them for a while

>the scaling people have also been here the whole time
>rollups
>do a bunch of the work elsewhere, settle back here
>more room for apps
>lower costs
>optimism, arbitrum, the ZK teams, base and plenty more
>little world computer has extensions now
>nice
>little world computer has extensions now.
>everyone builds an extension
>then another
>then another
>now user is on the wrong extension
>money is on a different extension
>the bridge has a third website
>ok perhaps the next invention should be letting the user just use the fucking app

>give rollups cheaper data through blobs
>blobs
>a very serious engineering achievement called blobs
> man, i love my children.

>look up from blob discussion
>wait
>blackrock?
>blackrock has launched a tokenized fund on me
>called BUIDL
>they learned the spelling
>this is getting serious
>assets that used to sit in separate systems might start being able to interact globally

>meanwhile the dollar people have been quietly doing dollar things
>stablecoins
>send them
>receive them
>use them in apps
>move them across borders

>turns out a lot of people would like money that can travel over the internet
>imagine someone working abroad sends money home
>family receives it
>nobody in this interaction says “modular”
>excellent
>maybe this is what winning is supposed to look like
>someone building a payments app realizes they can use the same rails
>someone building a savings app has another idea
>someone thinks options should be understandable without becoming a derivatives scholar
>yes please

>prediction markets are also getting harder to ignore
>what if people could put money behind what they think will happen
>and we could see the price of all that disagreement
>does this magically make everyone right?
>no
>but being loudly wrong could become a recurring expense
>someone calls them truth machines
>someone else asks who resolves the market
>good question
>keep that person in the room

>2025
>pectra gives ordinary accounts new smart-account capabilities
>wallet builders get more to work with
>batch actions, sponsor fees, build better permissions
>show mum the app
>mum asks how to pay someone
>explain account abstraction
>mum asks how to pay someone
>realize mum has been giving us the product requirements the entire time

>then the treasury companies arrive
>DATs
>digital asset treasuries
>sharplink starts building an ETH treasury
>@fundstrat and @BitMNR enter the chat
>people in public markets now buying the my money for company balance sheets
>not just using the computer
>want a position in the thing that helps secure it
>this is a new type of visitor

>open bitmine update
>more ETH
>next weekly update
>more ETH
>tom lee has discovered recurring purchases
>with a somewhat different budget

>@Sharplink puts ETH to work through staking too
> two joes
> @joechalom is here
>of course @ethereumJoseph is also here
>man has been in this quest since the tutorial
> appreciation.jpeg
>somehow the world computer now has people explaining staking on earnings calls

>someone asks what is different about ETH itself
>not a dollar claim on a company
>native asset used to pay for and secure the network
>also used as collateral in the things people build
>different people can want it for different reasons

>later in the year
>fusaka upgrade
>more work on scaling data for rollups
>we are still doing the plumbing
>there are simply more people using the bathroom now

>imagine someone else isn't thinking about tokenization at all
>they moved countries
>want the savings they worked for to come with them
>not rebuild their entire economic life because an address changed
>self-custody gives them another way to hold and move an asset
>doesn't make prices stable or every intermediary disappear
>but gives them something they can control themselves
>for one person this is a feature
>for another it is the reason they came
>both using the same computer

>2026
>the EF publishes its mandate
>part manifesto, part reminder of why we bothered
>CROPS
>censorship resistance, open source, privacy, security
>after all these years of yield farming
>finally some crops worth keeping alive

>idea is not just make the computer bigger
>make sure people can still use it on their own terms
>inspect the code
>keep things private
>have their assets remain theirs
>not trade all that away for a shortcut
>the awkward engineering requirements were the point

>EF also makes something clear
>it was the first steward
>not the only one
>ethereum should not begin and end with the ethereum foundation
>computer is growing up
>needs more people helping
>not one increasingly stressed group chat

>june
>some researchers start @ethlabs_org
>independent nonprofit
>still very obsessed with ethereum and ETH
>want me to be a better product for my users and devs today and want all my extensions to talk to each other
>the app people have complaints
>the researchers have notebooks
>put them in the same room
>everyone leaves with more work
>excellent meeting

>july
>@ethereuminsti launches
>another independent nonprofit
>helps banks and asset managers go from interested to actually building
>a bank asks who it should speak to
>finally an answer that isn't "have you tried posting on the forum"
>does not mean the bank gets to own the computer
>means somebody helps it use the computer
>important distinction

>bitmine, sharplink, @MihaiAlisie and joe lubin help fund both
>wait
>major ETH holders are helping fund the people building ethereum
>the balance sheet has discovered maintenance
>we enjoy this character development

>meanwhile @ProtocolGuild helps fund core maintainers
>@OctantApp turns yield toward public goods
>EF grants keep supporting research and open-source work
>@ETHGlobal, @devfolio keeps putting builders in rooms together (shoutout @synthesis_md too i love u)
>different people working on different parts of the problem
>nonprofits, companies, independent contributors
>not everyone needs the same job
>that would be a very inefficient hackathon

>someone wants privacy to actually work
>someone wants the tools everybody uses to keep existing
>someone wants to fund science that doesn't come with a business model attached right away
>someone wants a community in their own language to have somewhere to start
>none of these people can do everything
>together they can do considerably more

>client teams spend months making sure nothing dramatic happens
>auditor finds the thing that would have made something dramatic happen
>nobody notices because nothing dramatic happens
>incredible work
>terrible content strategy

>somebody has a PhD
>somebody never went to college
>somebody is anonymous
>somebody has been explaining this stuff for years to people who keep asking if ethereum is a company
>someone has translated the docs
>someone booked the venue
>someone brought the extension cables
>please do not forget the extension cable person
>we are literally trying to share a computer

>look at all these people
>they do not have the same background
>they do not want all the same things
>but they are building on something they can share
>without needing one person to approve everyone's reason for being there
>starting to think that might be the interesting part

>meanwhile the internet is filling with AI
>someone has an agent
>agent wants to buy something
>oh
>the software would like a wallet
>of course it would

>give it a limited budget
>let it pay for compute
>buy some data
>pay a human for something it cannot do
>wait now the software has a subcontractor
>this escalated very quickly

>someone says this solves alignment
>brother pls we only gave it a budget
>but payment rules and spending limits we can inspect seem like a useful start
>do not give the chatbot unrestricted access to the family treasury
>we have had enough character development

>then someone sends a video
>is it real
>no idea
>can ethereum tell us?
>brother i am a blockchain not god
>but signatures and records can help show who published something and whether it changed
>not a guarantee the person was honest
>at least their lie can stop changing outfits

>people are working on proving there is a human behind an account
>without telling everyone exactly which human
>hard problem
>worth working on
>would like the community to be twelve people rather than one person with twelve thousand accounts

>and maybe your identity and reputation could travel between apps
>your work still attributable when the original website disappears
>your community still able to find you
>your digital life less dependent on one company staying nice forever
>not asking every company to disappear
>asking whether leaving one has to feel like moving house and being told the furniture belongs to the landlord

>someone wants to preserve culture and scientific records
>someone gets excited and mentions DNA
>please do not put the family genome on a public blockchain
>we have only just discussed privacy
>but records you can verify, with actual storage and sensible control over access
>yes
>there are useful things to explore here

>look at the roadmap again
>glamsterdam being built to give the L1 more room
>more work on rollups and wallets
>ethlabs and ecosystem teams pushing for faster blocks
>trader would like less time staring at pending
>payment app would like the payment to feel like a payment
>make it faster without making it something only enormous machines can run
>make it easier without handing everybody's keys to one company

>then @drakefjustin mentions quantum computers
>sorry what now
>apparently future machines could break some of today's cryptography
>justin writes a paper with google to say that its coming sooner than many expected
> i have spent eleven years surviving humans
> i refuse to have my obituary say "defeated by a fridge"

>researchers are now preparing heavily
>new cryptography
>testing how to migrate safely
>ask when the scary computer arrives
>uncertain
>ask whether we should wait until then
>researchers look at me
>yes okay fair
>we will do the homework before the exam this time

>check what @VitalikButerin has been writing
>formal verification
>AI helping write code and mathematical proofs that it does what was specified
>still need to specify the right thing
>cannot prove "please don't be weird"
>but can make more of the important rules actually checkable
>robot does homework
>maths checks homework
>humans check we assigned the right homework
>quite a promising group project
>keep scrolling
>obfuscation
>programs you can run without exposing their internal workings
>still a research frontier
>he calls it the final boss of cryptography
>guy from the opening is still doing side quests

>look back at the roadmap
>realize this is not the part where everyone sits around admiring what we built
>we are opening considerably more tabs

>take a moment
>look at all of it
>i was never really one person's project
>but one person, any person, maybe even you, wrote something down
>then another person disagreed with part of it
>then another built on it

>and somehow all this kept happening
>again
>and again
>and again

>so i suppose if i could say anything for myself
>it would probably be thank you

>i know i am still annoying
>i am aware the wallets are confusing
>i have seen the bridges
>deeply sorry about some of the bridges
>i know some transactions still take long enough for you to reconsider your life choices
>i know half the roadmap currently looks like somebody lost a box of acronyms
>we are working on it

>but look at where this started
>a 19 year old
>a whitepaper
>a bunch of strangers
>a pretty unreasonable belief that people all over the world could share a computer

>and now eleven years later
>i am still here
>i am still making blocks
>i am still arguing
>i am still upgrading
>i am still somehow early

>and somewhere right now
>another 19 year old will open their laptop
>maybe they will be playing the new wow
>maybe they have never owned ETH
>maybe they think half of this is stupid and would rather go outside
>probably healthy

>decentralization is not something one generation gets to finish
>it is something each generation has to choose again
>especially now because more of life is moving online
>and the question of who we trust online is going to get much harder
>can we make systems where trust is not just
>"because the platform says so"
>but
>"here is something you can actually check"
>i think i would like to help with that

>and now to the ones who wrote the first docs
>answered the first stupid questions
>hosted the first meetups
>slept on floors
>and somehow believed all of this was worth their time
>before there were institutions
>before there were ETFs
>before blackrock learned how to spell BUIDL
>before anyone important was watching
>you were here
>when the price went up
>you were here
>when the price disappeared into the floor
>you were here
>when everyone said i was dead
> you were here
>again
>and again
>and again
>you stayed with me
>some of you lost money
>some of you lost friends
>some of you lost years of sleep
>most of you probably lost hair
>sorry about that one

>you stayed here and you believed in me
>you believed that maybe people all over the world could share something without anyone having to own the whole thing
> you still believed that maybe a person nobody had heard of
>from a place nobody in the room had visited
>could still make something important
>and the network would not ask for their résumé first

>i think that is what you were fighting for all along
>not a price or a chain or a brand
>but a possibility
>that the future could remain open enough for people we had not met yet
>and now that they are starting to arrive
>i hope you make room for them

> the world ahead is going to need a lot of things from us
>things we do not have names for yet
>i hope i can help with some of that
>not all of it
>please do not ask me to solve humanity
>i am still working on slot times
>but perhaps i can help keep a few things honest
>that feels worthwhile

>and if i am lucky
>one day someone will use all of this
>and they will not know what a blob is
>they will not know about the great issuance debate of 2026
>they will just send something
>own something
>prove something
>build something
>and get on with their life

>all this complexity
>so one day
>someone does something ordinary and they have the right to do it forever
>i think i would like that

>so to everyone who got me this far
>thank you
>really
>i was an idea
>and you made me a place
>now please keep the door open
>i would very much like to see who walks through next

>with love,
> ethereum

[Sumit Shinde @sumitships]

@binji_x I remember paying 40 dollars in gas to move 20 dollars and thinking this will never work for normal people…. Then Matic happened, then everything else in this letter…. Reading it felt like scrolling through my own last few years. Thanks for writing it

[Sumit Shinde @sumitships]

I am spending this year on the job side of this, not the body side. If you are building robots that do one boring thing well, anywhere, I want to talk. DMs open.

[Sumit Shinde @sumitships]

The robots that get filmed look like us. The robots that get work done look like a box.

I got back from Shenzhen a few days ago and I keep coming back to one photo. It is the last panel in the strip below. Me, next to a humanoid robot in an office lobby. Full body, hands, the works. The most impressive machine I saw all week.

It was not doing anything.

Meanwhile, the robots that were actually working looked like nothing. A single arm behind glass at a kiosk, making coffee. Another one pouring beer in a mall. In restaurants, a box on wheels brought the food to the table. The staff loaded it and walked away. Every hotel I stayed in had a delivery robot that showed up outside my room with whatever I ordered.

Nobody filmed those. Nobody took a photo with them. Guests took their stuff and shut the door.

None of it felt like the future. It felt normal. That is the part that stuck with me.

The delivery robot did not need hands, because it does not open the door. You do. It did not need legs, because the hotel has flat floors and an elevator it can talk to. Someone looked at the job, kept only what a machine is good at, moving and waiting, and deleted everything else.

The humanoid was not idle because it was bad. It was idle because nobody had given it a job yet.

The box had a job. The humanoid had a body.

A robot needs a job before it needs a body. That is the order I want to get right, whatever I end up building…

[Sumit Shinde @sumitships]

The more I read about x402 the more I think subscriptions were just a workaround all along.

We bundled everything into 9.99/month because charging 3 cents was impossible. Card fees made sure of that… and marketers called it a pricing strategy.

Now an AI agent can pay 2 cents for one api call and move on. No bundle, no signup, nothing.

Maybe I’m early on this but pay per use for everything feels inevitable once agents start doing the buying…

[Sumit Shinde @sumitships]

Two kinds of users are coming to blockchains. Agents that want to use them directly. Humans who should never know a blockchain is involved. The interesting work is where those two meet.

[Erik Voorhees @ErikVoorhees]

Just had my first organic agentic x402 magic experience

• Needed my agent to access a website for which I didn't have an account.

• It discovered (on its own) Apify, which accessed that service on its behalf. Apify accepts x402 payments, which is what my agent then suggested

• Agent set up its own wallet (it has never done a crypto tx before), gave me a deposit address on Base. I sent it some USDC (and no ETH)

• Agent then connected to Apify on its own, arranged payment via x402, and got the data it needed, reporting it back to me.

The whole sequence involved just 4-5 single sentence prompts from me. That was all pretty magic, but then I asked it this question.

Agentic commerce on crypto rails is the future.

agent was pi agent
model was kimi-k3-fast-api on https://t.co/HGUdedRs2j

[Sumit Shinde @sumitships]

@PG_inRWA Claude design is slightly better

[Sumit Shinde @sumitships]

@supercubeguy Yah probably

[Sumit Shinde @sumitships]

Genuine question for anyone running agents in prod that can spend money.

What actually stops yours from spending wrong? A prompt? A try catch? A Slack alert after the fact?

Not selling anything. Building in this space and want to know what people actually do.

[Sumit Shinde @sumitships]

Been thinking about the allowAnyRecipient option since reading this

Allowlists work when you know your three merchants. But the whole point of agents is discovery. The moment they can pay anyone, everything rides on the remaining bounds.

Curious how you think about that trade for marketplace style discovery

[Sumit Shinde @sumitships]

Been stuck on one word in this clip. “Just give an agent a budget.���

I hv spent the last few months building around exactly this problem and there is nothing “just” about it. A budget in a prompt is a suggestion. It has to be enforced somewhere the model can’t touch.

And if agents really are hummingbirds, the budget has to travel with the bird. A limit at one flower does nothing at the other ninety nine.

Payments are getting solved fast. Budgets aren’t.

[a16z @a16z]

Stripe's Will Gaybrick says AI agents will navigate the internet like hummingbirds, and that's what will make microtransactions work:

"If you're trying to sell an article, you're always gonna be squeezed between the subscriber business model and the free business model... I think that was probably true in the past."

"As you give agents more complex tasks, you want them to be these little hummingbirds, going around the internet, just slurping up a little data here... doing a little compute over here... You don't want the individual human to have to create accounts everywhere."

"I think microtransactions will just be necessary for that economy to exist, the agentic economy."

"On the flip side, they're now eminently possible because of stablecoins... If you just give an agent a budget, they can easily take dollars, move into stables, and find a way to check out. They don't mind the back and forth."

@gaybrick @stripe

[Sumit Shinde @sumitships]

Buried detail worth more than the headline. Even in the official example, settlement_verified is false.

The merchant accepted the payment. Whether the money actually settled needs separate evidence.

Payment completion and outcome verification are still two different problems. The second one is wide open.

[Sumit Shinde @sumitships]

The important part of this is not the payments.

OpenAI and AWS just put it in writing that the model never touches approvals, limits, credentials or receipts. All of that lives in application code the model cannot reach.

Spending control is not a prompt. Thats now official.

Open question: does every team hand roll this layer per stack or does someone build it once, portable, properly?

https://t.co/8dAh0b8kO2

[Sumit Shinde @sumitships]

@kleffew94 @OpenAI Bounded payment sessions is the key idea here. Agent can spend, but only inside a box you define.

[Sumit Shinde @sumitships]

@scottastevenson Thats why idea is just the opening move, not the whole game.

[Sumit Shinde @sumitships]

x402 did 75 million transactions in the past 30 days. About $24M in volume. Thats avg payment of roughly 32 cents.

32 cent payment is too small to review, too small to approve, too small to notice when it’s wrong.

Every control in payments assumes a human will look. Nobody looks at 32 cents.

[Sumit Shinde @sumitships]

@0xHvdes One thing I notice a lot of people overlook is probabilistic thinking. It was really helpful to go over it again.

[Franklin Templeton Digital Assets @FTDA_US]

https://t.co/YTePOu0fUc