Jacob retweeted
Distributed systems are hard to test due to inherent non-determinism in their execution traces. Unmesh Joshi describes how deterministic simulation allows reproducible execution.
martinfowler.com/articles/de…
Introducing dbu6. A new kind of personal finance tool, smarter than any expense tracker or accounting software in the world today. MIT licensed.
The tool is mostly air; it is a just shim over a SQLite database and double-entry bookkeeping.
When you upload a new bank statement, it runs to the coding agent. "Make me a parser for this abstruse PDF that the bank never intended mortal eyes to understand", it says.
"Aye, but I am no mortal", responds the agent, and brings forth a beautiful, well-tested parser for that statement in no time.
And once that is done, dbu6, out of false pride, tries to categorize the bank transactions into the right buckets using deterministic rules. String matching, regexes, and so on.
But when all else fails, it runs back to the coding agent, asking it to classify the transactions into the user's accounts, or make new ones as the situation demands.
I have had a lot of fun building this. And even more fun using it - every month I upload my statements to dbu6, and I watch it beating our banks and UPI apps and let me take control of my personal financial data.
Jacob retweeted
[new blog post]
Writing as a thinking tool
muratbuffalo.blogspot.com/20…
Jacob retweeted
TLAPS-Bench aims to raise the standard of TLA+ spec from "checked" to "proved."
We keep being impressed by the rapid advances in the proof capabilities of frontier models in the past 6 months. Today, we're able to prove most protocol-level specs in TLA+ with reasonable cost. System-level specs are still nontrivial, but honestly are not out of reach especially given the crazy speed of AI.
What I like the most is Tilapia 🐟 TLA+ Proof Infrastructure for Agents 🐟 the name is coined by @muratdemirbas, and the proofs we generated using Tilapia for many system/protocol specs that were not proved before. No matter how models evolve, Tilapia will be a useful proof service for TLA+.
The project is a huge effort across 11 organizations. I got to collaborate with many core @tlaplus colleagues I admire. @lemmster proposed the project in May and @qiancheng9788 drives it since then. Great work!
Meet TLAPS-Bench: A new benchmark of 265 real-world #TLAPlus proof targets to push AI agents to crack machine-checked correctness proofs using #TLAPS
44.3% rate at pass@3
46 bugs found
Paper: arxiv.org/abs/2611.02849
Code: github.com/specula-org/TLAPS…
lnkd.in/p/g3ET9E8H
Jacob retweeted
How we improved bloom filters in parquet-rs [1], now adopted by ClickHouse [2]: pydantic.dev/articles/bloom-…
[1]: github.com/apache/arrow-rs/p…
[2]: github.com/ClickHouse/ClickH…
Jacob retweeted
Engineering introduction to distributed query execution and how it is implemented by DataFusion distributed.
datadoghq.com/blog/engineeri…
Jacob retweeted
The ALP encoding is now officially part of @ApacheParquet.
parquet.apache.org/blog/2026…
Jacob retweeted
Great blog from Time Saucer, a @ApacheDataFusio PMC member and core maintainer, about why agent dump PR actually hurts (along with some good suggestions on how to make the contributions actually helpful): timsaucer.com/2026/09/14/tha…
Jacob retweeted
Super excited to share "A Design Space Exploration of Async/Await". How differently does async/await behave across today's languages? (More than you might expect!) And what *is* "structured concurrency"? Check out our ~interactive~ blog post on the paper:
cel.cs.brown.edu/blog/design…
I've published Sapporta - a framework to build database applications targeted at power users.
It makes your applications agentic by default, and provides a rich data-grid for every table you define.
It is the 80s technologies, - dBase, FoxPro etc. - but with LLMs.
Try it out - the demos need no login.
Jacob retweeted
At TigerBeetle, as systems engineers, we:
1. write code, sharpening skill and second order understanding
2. read code, walking through the building to constantly improve things
3. review code, ensuring no code is merged without two engineers’ understanding and review.
Jacob retweeted
excited to get this out there in the hopes that other folks can benefit from our work here! a lot goes into making a good tokenizer
also wanna make sure I thank the tantivy folks, we've built our earlier tokenizers (everything before word_v4) atop their lovely open sourced work
so @morgallant has optimized FTS tokenization throughput to 423 MiB/s and open-sourced it (github.com/turbopuffer/alyze)
I keep telling him that it would be really high agency to get to DRAM bandwidth (~100 GiB/s), and he keeps getting annoyed at me and making it faster
Jacob retweeted
Another example of incorporating the latest Database Research (in this case Limit Pruning from ❄️) in @ApacheDataFusio
datafusion.apache.org/blog/2….
Jacob retweeted
I wanted to understand how to generate my own vector embeddings and understand semantic search in general without external services. It became a significant rabbit hole. I wrote a post on some of what I learned.
Jacob retweeted
You can use @ApacheParquet for Vector Search with embedded indexes:
> We don’t change the file format; we just tune it.
@MOVNTDQ explains how in blog.xiangpeng.systems/posts…
Jacob retweeted
DataFusion 52 is released datafusion.apache.org/blog/2…
Jacob retweeted
New paper by Nancy Lynch summarizing her career's influence on the field of distributed computing.
arxiv.org/pdf/2502.20468
If you don't know who she is, she's the L in FLP and DLS. @MarcJBrooker has a good summary article: brooker.co.za/blog/2014/05/1…
Jacob retweeted
Been working on a tiny LLM service to help me write prompts just like regular well-typed application code.
Here's a sample use case - map freeform text to an address form:
Jacob retweeted
For everyone interested in data infra, want to get a quick sense of how big data works, how data systems are designed, and what the tradeoffs are, start with this share from @MOVNTDQ, really nice intro!
intro-data-system.xiangpeng.…
Jacob retweeted
Excited to announce a new side project, a power user terminal UI for your personal finances: moneyflow.dev/
For years I've used personal finance tools like Mint and now Monarch. The data cleaning can be slow and tedious, so I made this to speed that up!


























