SVRGN Weekly Digest #12 š«
ZKSummit should change their name to verifiable compute summit, or VCSummit, change my mind
This week was ZKSummit in Athens, the highest signal to noise-ratio crypto conference running since 2018 or so. If you missed it, hereās the tl;dr:
Zero knowledge proving is getting more and more practical - you no longer need a PhD to build useful stuff
Hardware acceleration was front and center, both on the main day and the second āside-eventā ZKAccelerate hosted by Ingonyama
Two years ago we saw everyone doing ZK-EVMs, last year (and arguably this year too) itās ZKVMs but with the strategy being to grab remaining parts of the proof value chain as well
Very few privacy-related things, most likely related to Tornado Cash
Athens is a cool city with amazing food and nice people
Taxi driver: āThis is the best job in the world. Iām paid to learn from my passengers. I ask questions and listen. People tell you everything because they think Iām just the cab driver.ā
š¤¼Ā People
Daniel DorƩ - Cryptography Engineer at Lita
Daniel is a number theorist turned cryptographer turned cryptography engineer. They gave two talks about Lita and ValidaVM at the ZKSummit and we chatted about math as art, strangeness of academia and working in remote teams. Fun fact: itās normal for number theorists to publish their first paper in their post doc.
Alin Tomescu - Chief cryptographer and founding team at Aptos Labs
Alin rides motorcycles for long distances. We didnāt talk much about crypto, but a little bit about the Aptos ecosystem, which I am woefully uninformed about. One cool thing theyāve been working on is āKeylessā - a way to effectively sign in with google to the aptos blockchain, using Google Auth credentials instead of having to deal with a private key. Ease of use here we come?
Raphael Volpert - Explorer/ex resident at the residency/founder of fixkey
Raphael hosts cool exploration deep-dives about frontier tech topics. Just tonight he hosted one in Munich on in-space manufacturing. We chatted about that, but also about the idiocy of typing to interact with LLMs, there must be higher bandwidth ways, no?
šĀ Companies
Astro Mechanica - Hypersonic flight and space launch
Status: Seed
Source: X
Founders: Ian Brooke
Why itās cool:
Jet engines stopped developing a long time ago, aerospace primes stopped being innovative and agile. Why shouldnāt we fly 3x faster than today? Why shouldnāt we launch rockets from airplanes, from any runway?
Powdr Labs - Generalizing the prover toolchain for writing VMs
Status: Ethereum Foundation funded
Source: Friends
Founders: Leo Alt
Why itās cool:
They make it sound more complicated than it is. You write your code in rust, and the powdr toolchain compiles a custom ISA for you, on the go. Most approaches are:
fix a general-enough purpose VM circuit with some ISA (e.g., RISC-V)
let people compile programs into opcodes in ISA
instead, powdr allows you to:
write program you want to prove
compile program into a custom ISA and generate the circuit for the VM that runs it
Many provers are supported out of the box, you can use co-processors for especially heavy lifting and aggregate proofs automatically.
Lobster Robotics - Visual seabed mapping through autonomous UW robots
Status: Seed
Source: CAVI
Founders: Stephan Rutten, Daan de Groot
Why itās cool:
More companies should be called after opportunistic omnivores. Make it clear you understand how the game works. But besides that, itās ridiculous how little we know about the ocean and the topology of the seabed, what lives there and how itās evolving.
š”Ā Ideas & Science
Building Jolt: A fast, easy-to-use zkVM
For full verifiable compute to become reality, we need very fast zkVMs. a16z is showing their dedication to contributing to research by releasing Jolt. It outperforms RiscZeroās VM by 5x and Succint labs SP1 by 2x on the fibonacci benchmark š
Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models
Boaz Barak et al. show that itās impossible to strongly watermark generative models, by introducing an efficient general attack. Cryptography for data provenance could still be used.
Swarming Proxima Centauri: Optical Communications Over Interstellar Distances
How do we communicate once we send probes to Proxima Centauri? Of course with a mesh network, or:
launch a long string of 100s of gram-scale interstellar probes at 0.2c in a firing campaign up to a year long, maintain continuous contact with them (directly amongst each other and via Earth utilizing the launch laser), and gradually, during the 20-year cruise, dynamically coalesce the long string into a lens-shaped mesh network ā¼100,000 km across centered on the target planet Proxima b at the time of fly-by.