Changelog

New Pages

• Leader leases versus quorum writes — 141 words
• The Grokster Decision - Ode To Thomas Jefferson — 4364 words
• Crash versus Byzantine faults — 256 words
• How the textile worker adapts — 573 words
• John Carmack UpperBound 2025 talk — 6 words
• UpperBound 25 talk notes — 4862 words

Edited Pages

• Cryptography and blockchain:

  • - Architecture
  • - Consensus layer
  • - Execution layer
  • - PeerDAS
  • - Malachite (Tendermint v2)
  • - Multi-Proposer Construction on top of Tendermint
  • - How does a skip ratchet work?
  • - https://vitalik.eth.limo/general/2019/04/03/collusion.html
  • - How does a skip ratchet work?
  • - https://vitalik.ca/general/2019/04/03/collusion.html

• The Path Matters:

  • Archived from: https://boz.com/articles/the-path-matters

• The World According to Rene Girard:

  • Archived from: https://eriktorenberg.substack.com/p/the-world-according-to-rene-girard

• Distributed systems:

  • - Facebook.
  • - Finding a needle in Haystack: Facebook’s photo storage
  • - Fault Tolerance (cheatsheet)
  • ## Replication
  • - quorum-based replication
  • - chain replication
  • - Fault Tolerance

• Fallacies of distributed computing:

  • https://en.wikipedia.org/wiki/Fallacies_of_distributed_computing

• Machine learning:

  • - Compression.
  • - Language modelling is compression
  • - Theory.
  • - The Platonic Representation Hypothesis
  • - Harnessing the Universal Geometry of Embeddings
  • ## Notes.
  • - RNN’s
  • - Test Time Training
  • - Test Time Training in Video
  • - The move to foundation models
  • - John Carmack UpperBound 2025 talk
  • - Heuristics
  • - Natural ML
  • In 2025, I’m going deep into ML and want a job. There is a large cache of notes here.
  • ## Sub-pages.
  • - Heuristics
  • - Natural ML
  • - Notes on RNN’s
  • - Notes on Test Time Training
  • - Notes on Test Time Training in Video
  • - Notes on the move to foundation models

• Programming:

  • - Site Reliability Engineering book (Google)
  • - See Notes on abstraction in mathematics versus computer science.
  • - Measurement, observability:
  • - Metrics
  • - Logs
  • - Traces
  • - Profiles
  • - File systems: block sizes.
  • - Databases (relational): page sizes.
  • - P2P networks: BitTorrent’s piece sizes
  • - Prometheus/Grafana
  • - the original client, in order to discover peers, would ping random IPv4 addresses. this isn’t even as hairbrained as it seems. you can query the entire ipv4 address space in 16mins
  • - The core elliptic curves of ZK-STARK’s were discovered computationally. They had to search for the parameters of two elliptic curves - MNT4 and MNT6 - which took over 610,000 compute hours
  • - Did you know, this was originally called the Albanian Generals problem (and hence Albanian Fault Tolerance) before they realised that might be a bit touchy..
  • - See Notes on abstraction in mathematics versus computer science.
  • - P2P networks: BitTorrent
  • - the original client, in order to discover peers, would ping random IPv4 addresses. the funny thing is this isn’t as hairbrained as it seems. you can query the entire ipv4 address space in 16mins
  • - The core elliptic curves of ZK-STARK’s were discovered computationally. They had to literally search for the parameters of two elliptic curves - MNT4 and MNT6 - which took over 610,000 compute hours
  • - Hahha did you know, this was originally called the Albanian Generals problem (and hence Albanian Fault Tolerance) before they realised that might be a bit touchy..

• Science:

  • - What’s inside the digits of pi?
  • - intelligence = prediction = compression (hutter)
  • - you can convert any ML model (like an LLM) which outputs probabilities to a compression algorithm by using arithmetic coding
  • - arithmetic coding converts data to a single float. it turns out, base-10 is a special case of arithmetic coding for uniform probabilities 1/10.
  • - is pi just the output of an arithmetic coding, for which we lack the ML model?