2021-03-31
People of IOG: Dr. Lars Brunjes
- A People of IOG biographical interview with Dr. Lars Brunjes, IOG's director of education, tracing his life from a Cold War childhood in West Germany to running the Plutus Pioneers program.
- Lars recounts a mathematician's path: a mathematics degree and a PhD in algebraic geometry at Cologne, a postdoc at Cambridge working on non-standard analysis, then leaving academia for industrial optimization software.
- He explains how discovering Haskell led him to IOHK in 2016, and how his love of teaching turned him into the company's director of education running Haskell courses in Greece, Barbados, Ethiopia, and Mongolia.
- On the research side he describes designing Cardano's proof-of-stake incentive system with Elias Koutsoupias and Aggelos Kiayias, and building a Uniswap-style DEX in Plutus.
- Charles and Lars discuss the future: dynamic fee pricing, the Plutus Pioneers program at scale, and QEDA, a long-standing idea to put all of mathematics on the blockchain with proofs as incentivized assets.
118 entries
Charles introduces Dr. Lars Brunjes, IOG's director of education, joining from near Regensburg.
Lars was born in Wuppertal, known for Engels and the world's only commercial suspended monorail.
Born in 1971 and raised in Cologne, he grew up scared by the Cold War in West Germany.
His father was smuggled out of East Germany as a child under a coat on a train in the 1950s.
Visiting East German relatives required a visa and passing the death strip with watchtowers and guns.
Border guards disassembled the whole car searching for people hidden in gas tanks or under seats.
East Germany felt gray and depressing, with propaganda posters but no color, ads, or fresh paint.
An air of fear pervaded the East, with neighbors spying and visitors made to register with the police.
A childhood memory of taking two sausages illustrates the scarcity his Eastern relatives lived with.
Both his parents were mathematicians: his mother a high school teacher, his father an actuary.
His father's department got an IBM 5110 and told him about it, sparking a lifelong interest in computers.
His first home computer was a Sinclair ZX81; while waiting for it he wrote BASIC programs on a typewriter.
The ZX81 ran a Z80 CPU with 1KB of RAM, a 16KB expansion, and a tape recorder for storage.
One of his first programs drew a sine graph on the ZX81's rough 64 by 48 pixel display.
He describes Germany's three-tier high school system, choosing the nine-year university track at age ten.
Germany lagged in computer education; his whole school had one Atari while a US reservation school got 30 Commodore 64s.
He studied mathematics at the University of Cologne, taking computer science only as a minor subject.
Travel was limited to Europe by his mother's fear of flying, with Turkey the most exotic destination.
Before university he did mandatory service, declaring himself a pacifist which required proving it at a hearing.
His pacifism essay quoted the anti-war novel All Quiet on the Western Front.
Instead of the army he did fifteen months of civilian social service.
He worked as a caretaker for a parish church, mowing lawns and driving elderly women to Bible classes.
A highlight was operating the church's old printing press to produce the monthly newsletter.
During his service year he entered national mathematics and computer science competitions.
He envied East Germany's stronger math culture and its math books, which were free of propaganda.
His mother lamented the German attitude of being proud of not understanding mathematics.
In Germany all universities are roughly equal and nobody cares which one you attended, unlike US Ivy League prestige.
German university was essentially free, with only nominal tuition, which he was always proud of.
He read the German classics on the tram, catching up on Goethe, Schiller, Brecht, and Kafka.
He recounts favorite Brecht stories including Life of Galileo and their anti-capitalist themes.
As a student he played some of the first multiplayer online games with friends after lunch.
He never pursued game development, feeling he missed the timing, and instead explored neural networks and genetic algorithms.
In his undergraduate mathematics he preferred algebra over analysis.
After two years he became a tutor, his first paying job, running weekly exercise groups.
He contrasts algebraist and analyst cultures, solving a problem with a commutative diagram versus pages of inequalities.
He chose algebraic geometry, driven by the available professor and his wish to work algebraically.
He picked the hardest subject precisely because it was difficult and finally challenged him.
His Diplom thesis elaborated an obscure proof by Fields medalist Gerd Faltings of the Mordell conjecture.
After his PhD he took a postdoc at Cambridge through his advisor's connection to Tony Scholl.
At Cambridge he met Sir Peter Swinnerton-Dyer of the Birch and Swinnerton-Dyer conjecture.
Switching to daily English was eased by his South African wife, with whom he speaks English, Afrikaans, and German.
He worked at Cambridge's DPMMS, in the same ultra-modern building as Stephen Hawking.
His Cambridge research applied non-standard analysis to algebraic geometry, using infinite prime numbers to bridge characteristic p and zero.
Model theory was out of fashion and his papers were rarely read at the time.
On the Lean theorem prover forum, someone recalled and praised his non-standard analysis papers almost twenty years later.
Charles reflects on the isolated life of a mathematician and the phenomenon of write-only academic books.
German academia lacks permanent pre-tenure positions, making the tenure track risky and requiring constant moves.
At thirty and married, he decided the uncertain academic path was not worth the risk.
He sought a proper job that still used mathematics and programming, ruling out finance.
He joined a company making software for paper factories, working on paper-cutting optimization to minimize waste.
He impressed colleagues early by abstracting a task into a simple model instead of coding it directly.
A misrouted phone call handed him a wood-cutting optimization project, a harder two-dimensional problem.
His prototype won a multi-million dollar contract, and development moved to India to cut costs.
Around age forty he made his first trip to India with two colleagues, roughly eight months in total.
Landing in Mumbai and driven 200km to Pune, he found Indian traffic mind-boggling.
He describes chaotic Indian traffic with trucks parked to sleep across motorway lanes, like a video game.
He came to love India for its food and people, and even took yoga classes taught in Hindi.
He first encountered Haskell during his Cambridge years and later settled on it as his focus language.
At the paper company they used .NET and C#, and his managers repeatedly rejected using F#.
An invitation to the Haskell eXchange in London, which his manager refused to support, was the moment he decided to leave.
He joined IOHK in November 2016 after first meeting Charles in Malta.
What drew him to IOHK was less blockchain than a company genuinely built on science and Haskell.
Charles explains choosing Haskell because Cardano needed engineers who could read the Ouroboros papers, drawing on the formal methods and dependent types crowd.
Charles recounts finding Hascoin, a Haskell Bitcoin implementation, as proof Cardano could be built in Haskell despite it being brutal.
Hired as a developer, Lars quickly moved toward education, starting on the early Cardano node.
In Malta he met Duncan and Philipp, did formal methods work, then the idea arose to run a Haskell course in Athens.
Having missed teaching most after leaving academia, he jumped at the chance to run the Athens Haskell course.
The developer-training model partners with a local university, starting in Greece with NTU and Athens, then Barbados, Ethiopia, and Mongolia.
Returning to teaching he loved it, though he strongly dislikes grading.
The course runs ten weeks full-time and goes far deeper into Haskell than typical beginner textbooks.
Titled Haskell and Cryptocurrencies, the course draws examples from crypto: Bitcoin clients, peer-to-peer protocols, transaction validation.
The syllabus mixes theory like lambda calculus and monads with practical networking, parallelism, and concurrency.
The hardest transition for imperative programmers is grasping Haskell's lack of side effects.
Students value the new functional way of thinking, which reshapes how they approach problems even outside Haskell.
He teaches monads by motivating them with examples before ever naming the concept.
A popular Barbados lecture builds up features like division by zero and state, each requiring only a swap of the monad.
Plutus is essentially Haskell compiled to Plutus Core, with the extended UTXO model as the separate key concept.
Understanding Plutus means understanding the extended UTXO model: validators and the conditions to consume an input.
To build a DEX in Plutus he first translated the concept into the extended UTXO model before writing code.
He modeled a liquidity pool as a script address holding two tokens, with swaps following the Uniswap formula.
He built the Uniswap-style clone in about a week of evenings, having also written a Plutus book that is already outdated.
Teaching in Ethiopia took a full year of preparation after Charles first raised it in Barbados.
He recalls arriving in Addis Ababa, met by John and his driver, and navigating the visa queue.
He insisted on staying in the same hotel as teaching assistant Polina to plan lectures together.
Ethiopia surprised him with its 2700m elevation, leaving him cold and permanently out of breath.
He was struck by Ethiopian religiosity, and taught the first all-women class at the education minister's request.
The all-women class drew students from Ethiopia and Uganda; he notes India's strong gender balance in IT versus Germany.
He found the women's shyness rooted in an authoritarian education system rather than gender.
He praises working with Polina, a mathematician who had worked remotely for IOG a year before meeting anyone in person.
The diverse Ethiopian class included a PhD in AI, a college professor, and one who had worked with Ben Goertzel's SingularityNET.
IOG shares an Addis office with SingularityNET; the catered graduation hired about five students, with Ghana planned next.
He describes his research collaboration with Elias Koutsoupias at Oxford on game theory and incentives.
Charles asked him to design the incentive mechanism for Cardano's proof-of-stake system.
With Elias, Aggelos Kiayias, and Katarina they designed an incentive system avoiding Bitcoin's over-large pools, yielding a thousand pools.
He is proud that the deployed reward system spreads staking rewards and lets hundreds make a living from it.
He wrote Haskell simulations testing whether the mechanism's Nash equilibria are achievable in practice.
Charles sent him and Colin to a University of Oregon summer school on type theory.
The two-week Oregon school was taught by masters like Bob Harper and Edwin Brady.
He learned lambda calculus, gradual typing, and dependent typing there, writing Idris software.
Charles tells how Colin joined: he wanted to collaborate with Lawrence Paulson, inventor of Isabelle, at Cambridge.
Charles subsidized Colin's Cambridge PhD to formalize the Ouroboros forkable-string theory in Isabelle.
Colin quietly wrote tens of thousands of lines of Isabelle proofs formalizing Ouroboros.
Future game theory work includes tweaking the reward function for small pools attracting delegation and discouraging pledge splitting.
Rising ADA prices pushed up transaction fees, prompting ideas to automate fee adjustment via a stablecoin or other mechanism.
Plutus scripts add operational fees, and he wants to make the incentive models more realistic by including opportunity cost.
Charles frames dynamic fee pricing and deterministic resource modeling, citing RAML from CMU and side chains.
Plutus Pioneers, with over 1300 interested developers, will be taught at scale via a forum where students help each other.
He is excited by the birth of a new language and ecosystem, spending nights writing example Plutus contracts.
The conversation turns to computational logic and the Lean movement bringing formal mathematics into the mainstream.
IOG is subsidizing Jeremy Avigad's textbook Lean for Mathematicians at CMU, aiming to create a Center for Formal Mathematics.
He explains QEDA, a long-discussed idea to put mathematics on the blockchain with each theorem's transaction carrying its proof.
Combining a theorem blockchain with an incentive model would let mathematicians earn money by proving results.
Bounties could fund pure math research and offer income and inclusion for people in developing countries.
Proofs become NFTs in a marketplace for deduction, including certifying Plutus contracts as correct by construction.
The project is named after the QED manifesto, aiming to make mathematics machine-understandable.
His bookshelf holds Terry Pratchett novels and Hofstadter's Godel, Escher, Bach.
A favorite book is Almost History, a collection of alternative-history counterfactuals.
Closing on following your passion, Lars marvels at going from humble beginnings to IOG and teaching in Ethiopia.