IOHK 01/11/2018

Open with a fresh shave and a selfie for Twitter, thanking a crowd of 400 people that far exceeded expectations.

The topic is third generation blockchain, striking given how young the space still is.

The first generation, roughly 2009 to 2013, solved one problem: a decentralized money with value that nobody controls.

When Bitcoin jumped from a dollar to about 250 during the Cyprus crisis, people started taking it seriously and wanting to build on it.

Bitcoin not being good enough was an invitation for augmentation, bringing altcoins, colored coins, and Mastercoin.

Knowing no one in the space, teaching a free Udemy Bitcoin course under Creative Commons brought 70,000 students and a network.

After BitShares, the recurring pain was that a deployed cryptocurrency is hard to change, so a scripting language would help.

Vitalik, stuck upgrading colored coins and Mastercoin, wanted a programming language for a blockchain the way JavaScript came to the browser.

The smart contract idea traces back to Nick Szabo in the eighties and nineties and to Sergio Lerner's turing complete scripting.

The second generation wants transactions that are eventful, stateful, rich, and programmable, not just decentralized money.

A smart contract can encode arbitrary business logic, like a board requiring three of five signatures dropping to two of three during vacations.

Smart contracts do not yet work for billions of people, and CryptoKitties congesting Ethereum proved they do not work at scale.

There are thousands of cryptocurrencies and most of them are blind, deaf, and dumb to each other, an interoperability failure.

Governance is unsolved: if two brilliant cryptographers propose rival signature schemes, there is no systematic way to decide between them.

The block size debate, Bitcoin Cash, and Ethereum Classic are the visible fractures of failed governance.

A governance crisis is a barrier to consumer and enterprise adoption, since executives cannot build on rules they have no say over.

The ICO is a Rube Goldberg machine for venture capital: issue an ERC20 token, run a crowdsale, and raise money with no strings.

ICOs are becoming regulated, with ideas from Smart Valor, Blockhaus, and Consensys Project Brooklyn on governing the funds.

Like an IPO, an ICO is a one time capital injection, which leaves a sustainability problem of how to fund a protocol long term.

The third generation is defined by three problems that come from success: scalability, interoperability, and sustainability.

After Ethereum, IOHK was founded to fix two deficits in the space: a peer review deficit and an engineering deficit.

Peer review is masochistic, with papers rejected and authors yelled at, but it gives real assurance that the claims may be right.

The plan was to build cryptocurrencies like a mathematician would, in Haskell and with formal methods.

Formal verification tackles the semantic gap between an ideal paper and the real code an engineer writes.

A formal specification plus a proof by simulation shows the code matches the spec, so it has no bugs as long as the paper is correct.

Most software skips this because it is horrendously expensive, done in tools like Coq or the Isabelle that IOHK prefers.

The seL4 verified microkernel took over five years and 180,000 lines of Isabelle to produce 7,000 lines of C.

IOHK started in 2015 with two people and now has a hundred across ten countries with nine figures of revenue, Cardano its largest project.

Scalability should mean that as you gain users you gain resources, a property BitTorrent already has.

Cryptocurrencies are harder than BitTorrent because they add transaction processing and data storage on top of moving data.

Generation one assumes everyone holds a full copy of the chain with a reliable connection, assumptions that break at scale.

Before scaling, you need bedrock and peer review, so the first question in 2015 was what a secure ledger even is.

The GKL15 paper defines a secure ledger, and proves that Nakamoto consensus forms one, so Satoshi did something magical.

Using that foundation, the team proved that proof of stake can reach a security level equivalent to proof of work under certain assumptions.

The Ouroboros proof of stake protocol took six revisions, rejected from Oakland before getting into the Crypto conference in Santa Barbara.

Beyond consensus you need a network protocol, and Cardano chose RINA, John Day's recursive internetwork architecture.

Ethereum's smart contracts were a misstep: the virtual machine was rushed and Solidity was poorly designed.

Runtime Verification and Grigore Rosu used the K framework to write formal semantics of the Ethereum VM, the KEVM.

From that they built Yella, a register based virtual machine that looks like LLVM, making it easier to build compilers.

With semantic space compilation you write a language's K semantics and get a compiler for free, already done for Java, C, JavaScript, Solidity, and Plutus.

All of that just builds a better Ethereum, but the goal was a third generation cryptocurrency.

Interoperability between crypto systems is a cryptographic problem: prove that incoming coins exist and were not double spent.

The non interactive proofs of proof of work paper builds tiny bridge proofs between chains, and is being translated to proof of stake.

Sustainability comes down to voting, and a bad voting system gives you poor choices by the time you get to vote.

Cardano leans toward liquid democracy, delegating votes per situation with additive homomorphic encryption and thresholds.

A Lancaster University team under Bingsheng is writing papers on a voting system and treasury abstracted as a reference model for any blockchain.

Voting and treasury systems need management, because if blockchain communities cannot govern themselves the governments will govern them and they lose.

All the work is open source and peer reviewed, with research centers at Edinburgh, Tokyo Tech, and Athens, and one coming in the US.

Competitors specialize: IOTA and EOS on scalability, Dash and Tezos on governance, Ripple and Aeternity on interoperability, while Cardano does a bit of each.

On backward compatibility, Phil Wadler warned that accepting Solidity means accepting all its sins, like the Parity wallet exploit.

Cardano splits into a settlement layer for accounting and a computation layer, which can run both an EVM clone and Yella.

Developers choose between safety and familiar tools, and the strong security guarantees live only on the encapsulated Yella side.

Fed controlled instruments will not run on a permissionless ledger, so an enterprise Cardano will come for coordination among trusted actors.

Once a blockchain lets institutions share a common state, it raises whether central banks are still needed and how decentralized they could become.

Institutional blockchain adoption will start not in America but in small central banks like Barbados and Jamaica that want bilateral swaps instead of trilateral settlement.

Institutional blockchains bring proof of solvency and real time auditing, but also easier capital controls, which interests the Bank of China.

Regulated money carries metadata and attribution that crypto transactions lack, and users have every right to resist that encroachment.

To move money into a regulated ledger you must play by their rules, so the next big conversation is what rights you lose crossing enclaves.

Asked why IOHK still supports Ethereum Classic: it is like asking which of your children you love more.

Ethereum's code is law social contract meant a better Bitcoin, while Vitalik wanted a world computer, two different visions.

After the DAO fork, being in the minority meant hiring a Scala team and building the Mantis client as a donation.

There is room for digital commodities: Bitcoin as digital gold, Ethereum Classic as digital silver with more utility.

Building bottom up in a federation, coordinating upgrades with people who dislike your ideas, is the real test of a decentralized system.

On history, Cardano surprised North America because the initial token sale focused on Japan and Asia.

Asia is the future of finance with four billion customers on old systems, like Japan's cash economy charging 300 dollars cash for a train ticket.

Innovate where there is real need and willing listeners, since American and Western European regulators protect incumbents.

Africa is next in 2018 with Ethiopia, Ghana, and Kenya, and a Haskell class is already running at the University of the West Indies in Barbados.

Cardano has been in Argentina since 2016, where a friend showed off decoy buried silver, since those people understand monetary collapse.

You cannot ask Chase and Goldman to change the rigged system, so you build a system for three billion people that they must connect to.

The peak TPS is unknown, and the harder question is propagating messages through the gossip protocol, studied with Delta Q and PNSol.

Ouroboros runs in epochs, electing slot leaders via cryptographic sortition using a source of randomness and follow the Satoshi.

The next generation of the protocol runs epochs in parallel for n times the throughput, adjusting like mining difficulty.

Moving and storing data stays the hard part, needing RINA and sharding as chains grow to exabyte scale.

For financial contracts, a domain specific language may beat a turing complete one, as Project ACTUS shows with 30 composable primitive contracts.

Marlowe, built at Kent with Simon Thompson, is a turing incomplete paint by numbers language for financial contracts.

For robust smart contracts there is Plutus, built with Phil Wadler and Darryl McAdams as the revenge of Haskell.

Mobile stays a walled garden controlled by Apple and Google, and if they get too draconian the answer is for the industry to build its own hardware.

The Mr. Burns Simpsons scene captures crypto: elaborate security to reach the core, then a rickety unlocked back door.

At the end of the tunnels sit six people running the system in the shadows through mining pools, exchanges, or the media.

Where there is power and money there will be attempts to centralize it, which is why Satoshi had to disappear.

Ripple putting Ben Lawsky on its board felt sickening, a necessary step for their IPO but a betrayal of the philosophy.

The first lesson is that no one is above corruption, so favor process over people and trust the peer review process over any individual.

The external auditor FP Complete will start publishing reports in February on the good, bad, and ugly of the code.

A system needs a constitution and rights baked in before everyone gets rich and powerful and tempted to change things.

Opposing the DAO fork came from the view that admitting a person must make that decision is admitting your system has a CEO.

The closing lessons: peer review scientific claims, have credible third parties review code, and bake a constitution and voting into the protocol.