Sparse Notes

Dynamic programming over subsets for a three-way join of EMP, DEPT and JOB: best single-relation access paths, then best two-relation plans per interesting order, skipping the DEPT–JOB Cartesian product, then the best three-relation plan.

Access Path Selection in a Relational Database Management System (Selinger et al., 1979)

Weekly Paper Notes — the Seminal Paper of the Week for the 2026-11-14 digest. Area: Databases. Authors: P. Griffiths Selinger, M. M. Astrahan, D. D. Chamberlin, R. A. Lorie, T. G. Price (IBM Research, San Jose) Published: Proceedings of the 1979 ACM SIGMOD International Conference on Management of Data, Boston, pp. 23–34. DOI: 10.1145/582095.582099 Why the paper still matters Codd’s relational model (our seminal pick in August) let users say what data they wanted and leave how to the system....

Raven architecture: a client engine talks gRPC to the catalog service (ERQL transpiler), which reads and writes catalog metadata and edge logs in the metadata store; a separate garbage collector with a graph store and in-memory graph engine consumes the edge logs.

Raven: Towards Unified File Management for Data Lakehouse Systems

Weekly Paper Notes — one of the top picks from the 2026-11-14 CS paper digest. Area: Databases. Authors: Keonwoo Oh (University of Maryland), Akshay Mehra (Dolby Labs), Joshua Kimball (Dolby Labs), Amol Deshpande (University of Maryland) arXiv: 2611.15975 · PDF · Code · Submitted to ICDE 2027 TL;DR Lakehouse metadata sits at one of two extremes. Open table formats (Iceberg, Delta Lake, Hudi) write everything as files next to the data....

Slide: steering moves to the model, approving to auto mode, saying no to deny rules and the environment, checking to tests, hooks and CI, starting sessions to Claude

Rethinking Claude Code Best Practices (Lydia Hallie, Anthropic)

Lydia Hallie works on the Claude Code team at Anthropic. Her AI Engineer talk (re-recorded at home after the live demo had technical trouble) is a 17-minute audit of the Claude Code advice Anthropic itself published earlier in the year. Her claim is that most of that advice was written for models that stopped early and needed correcting, and that with Opus 5.5 and Fable 5.2 a lot of it now gets in the way....

Slide: the Transformer encoder-decoder, with self-attention and position-wise feed-forward layers in the encoder and causal self-attention, encoder-decoder attention and feed-forward layers in the decoder

Self-Attention for Generative Models (Ashish Vaswani & Anna Huang, Stanford CS224N, 2019)

In early 2019 the Transformer was two years old. BERT and GPT existed, but most NLP courses still taught LSTMs as the default. Ashish Vaswani, first author of Attention Is All You Need, came to Stanford’s CS224N as its first guest lecturer that term and spent half an hour explaining why self-attention works, what it costs, and which inductive biases it brings. Anna Huang, lead author of the Music Transformer, gave the second half, on using relative attention to generate piano music....

Bar chart: PUT_OBJECT throughput per core; aws-sdk-cpp 1.7, aws-sdk-go-v2 1.9, anyblob 5.3, TigerBeetle 25.2 GB/core

The Fastest Object Storage Client in the World (Georg Kreuzmayr, TigerBeetle)

TigerBeetle is a database for financial transactions. It exposes debits and credits in place of SQL, supports only strict serializability, and runs a replicated state machine on top of an LSM tree. To grow from terabytes to petabytes it needed to put cold data in object storage, and the cloud vendors’ SDKs didn’t fit how TigerBeetle does I/O, memory or testing. In this 24-minute talk Georg Kreuzmayr walks through the client they wrote instead, one layer at a time....

Three panels. Pretrained versus scratch encoder: separate pretraining is optional. Encoder-free versus scratch encoder: dedicated vision capacity is key. Pre-fusion versus in-fusion capacity: vision capacity can move in-fusion.

Unbundling Multimodal Vision Encoding

Weekly Paper Notes — one of the top picks from the 2026-11-14 CS paper digest. Area: AI / ML. Authors: Ellis Brown, Chris Hoang (equal contribution), Shengyi Qian, Shang-Wen Li, Mike Lewis, Marjan Ghazvininejad, Luke Zettlemoyer, Lucas Beyer, Mengye Ren, Jakob Verbeek, Saining Xie, Hu Xu (Meta FAIR, New York University) arXiv: 2611.17334 · PDF · Project page TL;DR Most vision-language models bolt a separately pretrained image encoder (CLIP, SigLIP and relatives) onto a language model....

Slide: a two-by-two of complex versus simple and engineered versus emergent, with constructed, revolutionary, accreted and rebellious systems

Code Matters: Reflections on Rust in the LLM Age (Bryan Cantrill)

Bryan Cantrill, CTO of Oxide Computer, spent the 1990s and 2000s writing operating-system kernel code in C, mostly at Sun, while the rest of the industry built languages to make developers’ lives easier. In this Rust NYC talk he argues that much of that era’s convenience was an unacknowledged trade: easier for the developer, harder for whoever ran the software in production. Rust reversed the trade. The second half applies the same lens to LLMs, which he thinks are excellent at Rust for the same reasons, and dangerous for a reason most people haven’t named....

Slide: concurrency is about dealing with lots of things at once; parallelism is about doing lots of things at once

Concurrency is not Parallelism (Rob Pike, 2012)

When Go was announced in 2009, many programmers read “concurrent language” as “parallel language”, ran their programs on more cores, watched them get slower and decided Go was broken. Rob Pike gave this 31-minute talk at Heroku’s Waza conference in January 2012 to correct the mistake. He thought the problem was the programmers’ mental model, not the language. The talk is a classic because its central distinction still gets confused in code reviews, and because the explanation is a cartoon about gophers burning manuals that’s hard to forget....

Three SMR architectures. Conventional: clients talk to servers that run consensus and hold materialized state. Disaggregated: servers hold state but the log is on external storage. Serverless (S2MR): clients talk directly to external storage holding both the log and a single copy of state.

Eliminating Servers in State Machine Replication

Weekly Paper Notes — one of the top picks from the 2026-11-07 CS paper digest. Area: Distributed Computing / Databases. Authors: Dohyun Park (UIUC), Wenjie Hu (UW–Madison), Xiangyao Yu (UW–Madison), Mahesh Balakrishnan (Meta; work done prior), Yongjoo Park (UIUC) arXiv: 2611.01363 · PDF TL;DR State machine replication (SMR) is how etcd, ZooKeeper, Spanner, CockroachDB and their relatives keep state consistent: agree on a total order of commands, then have servers apply them....

Top: a benchmark turned into easier versions by truncation, masking, or logit mixing at hint level H. Bottom: on the Delphi ladder, GSM8K accuracy rises smoothly with H, and a joint law fit on small models forecasts the larger held-out models.

Hinted Scaling Laws

Weekly Paper Notes — one of the top picks from the 2026-11-07 CS paper digest. Area: AI / ML. Authors: Suze van Adrichem, Luke Bailey, Rohith Kuditipudi (equal contribution), Christine Ye, Neil Band, Ahmed Ahmed, John Duchi, Percy Liang, Tatsunori Hashimoto (Stanford University) arXiv: 2611.02270 · PDF · Code TL;DR Scaling laws on pretraining loss work. Scaling laws on benchmark accuracy mostly don’t, because the small models you can afford to train score zero on hard benchmarks, and a row of zeros gives you no trend to extrapolate....