Matt Rasmussen

new jynpost jyn.dev/what-is-a-bu...

We are excited to share GPN-Star, a cost-effective, biologically grounded genomic language modeling framework that achieves state-of-the-art performance across a wide range of variant effect prediction tasks relevant to human genetics.
www.biorxiv.org/content/10.1...
(1/n)

Build systems are awesome, terrifying, and unloved. They are used by every developer around the world, but are rarely the object of study. In this paper we offer a systematic, and executable, framework for developing and comparing build systems, viewing them as related points in landscape rather than as isolated phenomena. By teasing apart existing build systems, we can recombine their components, allowing us to prototype new build systems with desired properties.

Finally got around to reading Build Systems à la Carte (dl.acm.org/doi/pdf/10.1...), it's really great (like everyone says).

I love papers like this, that take a well-known, messy, unloved area and propose a useful organizing principle for it.

8/ Absolutely DETERMINED to make mutable collections key in a dict? Make them the default parameter of a lambda:

>>> {{}: {}}
TypeError: unhashable type: 'dict'
>>> x = lambda y={}: y
>>> x()[1] = 2
>>> x()
{1: 2}
>>> {x: {}}
{<function <lambda> at 0x000001E45E80E440>: {}}

Explicit note types that match and advance stages of sensemaking.

“Patterns of Hypertext-Augmented Sensemaking” by Zhu et al. (2024)

dl.acm.org/doi/pdf/10.1...

Text screenshot: "Ventures?" "—Please choose an option—" [group group-801] [object Object] [/group] "Sign me up for your mailing list"

This field is mandatory. I don’t know what it means.

Waste my time, I’m gonna waste yours.

As our debut Bluesky post, we’re excited to share our new paper (first author Zachary McCaw) in HGG Advances on scrutinizing the practice of using a ratio trait (numerator / denominator) for GWAS. www.cell.com/hgg-advances...

Our insitrocytes are the heartbeat of our shared vision. Together, as builders, scientists, engineers, we solve complex challenges every day, laser-focused on delivering better medicines for patients. Hear directly from some of our colleagues about how we work.
Visit www.insitro.com/people/.

Program 0900-1000: Compilers Database query compilation: our journey Viktor Leis & Thomas Neumann A YJIT interview Maxime Chevalier-Boisvert Can we democratize JIT compilers? Haoran Xu Safe and productive performance with user-schedulable languages Jonathan Ragan-Kelley 1000-1100: Understanding programs How debuggers work Sy Brand Debugging compiler-optimized code: how it works and doesn't Stephen Kell Side-Eye: ask your programs anything Andrei Matei Let’s run a million benchmarks Yao Yue Rocket science of simulation testing! Aleksey Kladov 1100-1200: Wild ideas Back to modularity Daniel Jackson DB usability: as if Jonathan Edwards Twizzler and far out memory sharing: precise abstractions Daniel Bittman Programming without pointers Andrew Kelley Throwing it all away - how extreme rewriting changed the way I build databases Tyler Neely 1200-1230: Programmers are people A case for feminism in programming language design Felienne Hermans Malloy, mic drop, peace! Michael Toy 1230-1300: Lightning talk buffet 1300-1400: Query languages ??? Jeff Shute ??? Tobias Brandt ??? Yury Selivanov AquaLang: a dataflow programming language Klas Segeljakt A polymorphic data model for SQL using algebraic types Steve McCanne 1400-1430: Databases Use of time in distributed databases —don’t fall behind the times Murat Demirbas ??? Thomas Ballinger Thank goodness we're not building a database Peter van Hardenburg 1430-1500: Wasm Thinking in wit Dan Gohman Bringing the WebAssembly standard up to speed with SpecTec Dongjun Youn

The program is up and tickets are live at www.hytradboi.com/2025/#program.

I'll keep updating this thread as the last 10 talks hit camera-ready.

Such a package index would encourage open source contribution of prompts plus pre-made code fragments

This might really help when building apps using coding assistants. Having assistants write so much code from scratch still feels unpredictable. What I really want is to be able to suggest from an index of code fragments and prompts which things to glue into my app.

But if you wanted to package up a partial table schema and some prompt about how it could be rendered in a UI, then the package manager+coding assistant could auto merge the partial schema into the user's existing schema and even make the right UI updates to show the new fields

Libraries and packages indexes (especially language specific ones) have been a boon for code reuse. But they can only be used for code that can be cleanly abstracted. I often encounter chunks of code I wish could reuse as a lib, but I can't because I can't abstract it well enough

I've been thinking about what libs and pkg indexes might look like as coding assistants become the norm. What if instead of packages providing a post_install.sh script (to regex config, add symlinks), they had a post_install.md prompt to help integrate the code into your app.

Building Bauble I made something that I think is pretty neat, and I want to tell you about it. This is a little hot air balloon made out of alternating layers of brass and bronze that lock together with these angle...

building bauble

ianthehenry.com/posts/bauble...

Jupyter Kernel Architecture - Blog by Roman Glushko How Jupyter executes, inspects, completes and debug code

I stumbled upon this excellent post on the @projectjupyter.bsky.social kernel architecture, that is probably the nicest narrative writeup out there I've seen on the topic (please let me know of others!). I'm going to make a PR on our docs to link to it:

www.romaglushko.com/blog/jupyter...

OK. Practical question times. How are you adjusting your research given progress in reasoning style models? Also how are you adjusting the way you work?

SQL Workbench - Rapid prototyping SQL Queries & Data Visualizations An online SQL Workbench based on DuckDB that can query and visualize remote CSV, JSON, Parquet and Arrow data, as well as local files.

I rebuilt sql-workbench.com over Christmas.

It now uses CodeMirror instead of Monaco, thus working on mobile now, too. As well as an update to Perspective v3 and moving to Shadcn UI sidebar from a custom one…

Groove Salad from SomaFM A nicely chilled plate of ambient/downtempo beats and grooves.

current vibe: somafm.com/groovesalad/

AI-generated tools can make programming more fun

New blog post! About how using AI to build custom debug views can make it more fun to do the programming ourselves.

www.geoffreylitt.com/2024/12/22/m...

Diagram with large number: 2.7.123 First “2” is commented: Proud version. Bump when you are proud of the release Second “7” is commented: Default version. Just normal/okay releases Third “123” is commented: Shame version. Bump when fixing things too embarrassing to admit

I propose we replace semantic versioning with pride versioning

Tree Calculus One operator. Trivial semantics. Turing complete. Intensional.

So, this is cool:

treecalcul.us

A very simple intensional calculus. A simple, privileged programming language where functions can be analyzed, serialized, etc.

Foursquare places data is live in the hive 🐝 🍯

@hachej.bsky.social @seifert.blue

A lot of no-code lab software (LIMS, automation scheduling) suffers from trying to reinvent a software practice concepts in a GUI. Things like branches, test environments, API integrations are either not supported or reinvented in strange ways.

Implicit in the design of most live and visual programming interfaces is the ethos “if everyone would just pull the rest of the world into my environment…”

I'm reviewing LLM frameworks and every. single. one.:

- declares that a prompt is some function of state that produces an output
- then invents a convoluted DSL

So why is this not the obvious way to do it?
- self-documenting
- typed inputs and outputs
- easy separation of template from context

Introduction - Model Context Protocol Get started with the Model Context Protocol (MCP)

Anthropic released an interesting thing today: an attempt at a standard protocol for LLM tools to talk to services that provide tools and extra context to be used other the models modelcontextprotocol.io

DASL — Data-Addressed Structures & Links A small set of simple, standard primitives to work on content-addressed data.

Have you ever wanted to use content-addressing primitives like CIDs or something IPLD/CBOR-like but you found the optionality in IPFS a little unwieldy and expensive to implement? Or, you know, interested in ATProto?

We've stared drafting a little thing for you: dasl.ing

go.bsky.app/UZB3Ndi

my braindump on how Observable Notebooks work at the runtime level
observablehq.com/@tomlarkwort...