@flaneur2024.bsky.social
1.1K followers 210 following 140 posts
maintainer of SlateDB loves Rust, Datasys, Cloud Infra, AI https://flaneur2020.github.io
Posts Media Videos Starter Packs
flaneur2024.bsky.social
recently Qwen3-Next-80B-A3B has become my goto model for asking "how-to" questions in the work. it's not considered as a smart model but it's really dog fast 😲
flaneur2024.bsky.social
thanks! hope it can be useful 😁
flaneur2024.bsky.social
yeah, it's an optimistic numbering scheme, if there's any username conflict, I'll just back off to the next year and try to register the account again. 😂
flaneur2024.bsky.social
what an amazing journey! after working through this step by step, we finally have a functional transaction API with SSI support. hope it proves useful! 😁
Reposted
sungkim.bsky.social
You already know the answer, but it’s nice that someone put in the manual effort to create a benchmark

Can AI file your taxes? No

"TaxCalcBench: Evaluating Frontier Models on the Tax Calculation Task"

Paper: www.arxiv.org/abs/2507.16126
Repo: github.com/column-tax/t...
GitHub - column-tax/tax-calc-bench: Code & data for TaxCalcBench
Code & data for TaxCalcBench. Contribute to column-tax/tax-calc-bench development by creating an account on GitHub.
github.com
Reposted
calavera.dev
Job update: a couple of weeks ago, I joined @tensorlake.ai full time. I’m having a lot of fun building the product with @diptanu.bsky.social and the rest of this wonderful team.

We have a few open positions if you’d like to work with us: www.linkedin.com/jobs/search/...
Reposted
chris.blue
On to new things!

"Ingest, query, and share telemetry data with your engineers and customers at a fraction of the cost."
flaneur2024.bsky.social
early return / goroutine leak is my favourite interview question, it can help reveal a candidate’s awareness of resource management, which is essential for building robust systems 😲
Reposted
Reposted
chris.blue
Chris @chris.blue · Sep 5
1/ SlateDB v0.8 is now available! This is release includes OpenDAL object store support, serializable snapshot isolation, first-class Go bindings, Python binding improvements, deterministic simulation tests, performance improvements, and tons of bug fixes. Details below. 👇
flaneur2024.bsky.social
now i believe all the dinosaurs are big chicken 🤔
Reposted
chris.blue
Chris @chris.blue · Sep 5
2/ Snapshot isolation — @flaneur2024.bsky.social has been hard at work on snapshots/transactions. 0.8 now has `DbSnapshots`, which provide a consistent point-in-time DB view. Sequence numbers are now core to SlateDB and will be used for many features (including transactions) going forward.
flaneur2024.bsky.social
instead of just looking at scores, we can build up an internal 'reputation' for each model based on how it performs day-to-day 🤔
flaneur2024.bsky.social
it seems there's no gold-standard benchmark for coding models at the moment. imho, the only way to know if one is any good is to throw a (real world) task at it and see if it can handle it smoothly 🤔.
flaneur2024.bsky.social
thanks to the help from @xuanwo.io for reviewing this, there are still some follow-up tasks that need to be continued 😀
flaneur2024.bsky.social
(i can only let it work independently for at most 10 minutes at the moment 🤔)
flaneur2024.bsky.social
in the age of vibe coding, the measure of a top programmer should be how long they can get Claude Code to work continuously without human intervention. if you can get Claude to code independently for a full 24-hour workday, you deserve top dollar in the market 🤔
24/7: Late Capitalism and the Ends of Sleep – June 4, 2013
by Jonathan Crary
Reposted
sungkim.bsky.social
z.ai's slime

It is built with native SGLang integration, carrying its full inference optimizations straight into training, is an LLM post-training framework for RL scaling, providing two core capabilities:
Reposted
avi.im
v @avi.im · Aug 10
I went through DuckDB's WAL, and it does everything I was asking for in my blog post:

1. Per record checksum
2. Explicit error on checksum failure
3. Configurable behavior
4. Partial recovery
5. Safe truncation of the WAL only when WAL contents are checkpointed
avi.im
v @avi.im · Jul 24
Published a new post: "PSA: SQLite WAL checksums fail silently and may lose data"

This is a follow up to my previous posts. When SQLite encounters checksum failures in WAL, instead of raising an error, it drops all subsequent frames; even if they are not corrupt. It's not a bug
PSA: SQLite WAL checksums fail silently and may lose data

This is a follow-up post to my PSA: SQLite does not do checksums and PSA: Most databases do not do checksums by default. In the previous posts I mentioned that SQLite does not do checksums by default, but it has checksums in WAL mode. However, on checksum errors, instead of raising error, it drops all the subsequent frames. Even if they are not corrupt. This is not a bug; it’s intentional.
flaneur2024.bsky.social
however, non-linear FFNs are terrible at learning the identity function. It's like trying to jump up to the 15th floor and then land back exactly where you started. with ResNet, the layer simply adds its input to its output, making the `identity` function trivial to learn 😲