SVRGN Weekly Digest #34 💫
Cryptography, extinction events and massive heat pumps
🤼 People
Jens Ernstberger - ? at various ideas
Jens did his PhD in EE and cryptography at TUM and was sucked into the world of crypto. He’s recently left a16z to go back to Munich and is now exploring a few different ideas. We chatted about applications vs infra in crypto, building stuff the world needs and trustless usage of the internet.
💼 Portfolio jobs board
This weeks selection of opportunities from the portfolio:
🚀 Companies
Nustom - The ultimate dev agency
Status: 6mo in
Source: CAVI
Founders: Jonas Templestein, Oliver Beattie
Why it’s cool:
Nustom is doing the whole code-gen-by-agents spiel, which we love of course, it’s a logical conclusion of LLMs scaling etc. But it’s not super interesting after we’ve seen the (failed?) Devin demo, work with Cursor, draft in v0 and use lovable’s GPTEngineer for every side project. What Nustom seems to be doing that makes them interesting is actually instead of selling the coding tool, to use it to build products themselves. The next logical conclusion of “we can scale code-making at high quality infinitely” is “let’s keep it for ourselves and build everything”. Also, the co-founder founded nubank Monzo before, so we can assume they know how to build products and companies.
Karman Industries - Electrifying industrial heat
Status: Pre-Seed raised
Source: CAVI
Founders: David Tearse, Chiranjeev Kalra
Why it’s cool:
Electrifying everything is possible, but it’s not easy. Imagine a BASF cooking chemicals, or Alcoa corporation smelting Aluminum; they can’t do that with any old power source, no they need immense amounts of stable power, or CHGs onsite in BASF’s case. Karman is building industrial level heat pumps, a necessary step in the road to full electrification.
💡 Ideas & Science
Stem cells reverse woman’s diabetes — a world first
Imagine Novo Nordisk reading this. Just kidding, I’m quite certain they’re acting for the better of humanity. Finally being able to engineer stem cells with such precision and certainty that nothing will go wrong that we can use it to cure diabetes, permanently is maybe the most trust-inducing development in humanities ability not to go extinct this week.
Protein design meets biosecurity
On the other hand, we should maybe be a bit careful when giving these design tools to everyone, says boring academic. And we agree! It’s not the AI agents (bots) we should be worried about in the first case, it’s the other people out there who wish others maximal harm, and the amount of intelligence and capital required to orchestrate a mass extinction event decreases rapidly with each new foundation model.
The proposal is interesting (using cryptography to screen for known pathogens and other bad combinations of peptides), but this seems like opt-in only. So pretty easy for bad actors to circumvent.
Speaking of extinction events
We seem to be heading to one quite clearly. However, this one we may not want to solve by technology only, but rather societal structures.