Emily Riederer

Please tell us this is a blog or book outline 👀

Role-Based Access Control for Quarto sites with Netlify Identity | Emily Riederer A quick tech note on Netlify’s managed authentication solution

Not sure if you are talking about securing backend stuff like API keys or front end webpages but I’ve done the latter with Netlify Identity and Quarto www.emilyriederer.com/post/quarto-...

I have the privilege of working with fed agency folks. They've been jerked around, forced to move, had offices taken, lost resources, put on leave, fired, re-hired, *shot at*, and now furloughed. Still they return each day to help make America a bit more informed, healthier, and safer.

Tomorrow I have to take 25 blurbs of text, copy paste them one by one into a UI, get an ID and link, and then go paste those into some docs

I’m reminded how much more I value of a good REST API or RPA/headless browser (Playwright!) over AI for the kind of work I want to delegate to a robot

Thank you!! Cannot wait to read and try 👀 Slides is the last place I always find myself reaching back to WYSIWYG. This may help me finally make the move

1. Bluesky doesn’t downrank links
2. You can choose your own algorithm from among thousands
3. The whole app is open source

This keeps it a free and open platform by design, leading to better click-throughs and traffic to your site.

We 💙 the open web.

If all the world were a monorepo The R ecosystem and the case for extreme empathy in software maintenance

Really insightful post from Julie Tibshirani (spotted in LinkedIn, can't find on Bsky) reflecting on #rstats 's unique governance structure and what can be learned for other languages

jtibs.substack.com/p/if-all-the...

No kidding! And yet it sometimes feels like a no win between things that are too flimsy or too rigid. Like, I want super robust tools without limiting the creativity that comes from flexibility ⚖️

It’s almost like the meme of “random effects implies the existence of fixed effects” except true — like fixed effects is almost hard to explain because they are normal-ish and it only makes sense to give a name to them if you’ve already talked about random effects?

Very fair. Hoarding with chipmunk like intensity tho. Or maybe our winter hasn’t come yet? No joke, in the waining days of Twitter I sometimes would scroll my bookmarks instead of the feed

Thrilled to once again be able to bookmark posts I will never once revisit like a chipmunk hoarding nuts before winter 🐿️

The really killer feature that the former place never had would be to search or organize your bookmarks 🙏🏻

This is fantastic!!

Are there any limitations of the Shiny app framework (eg golem, flex dashboard, etc) so long as webR has the packages?

Not too familiar with Electron, but in theory could this also allow pkging w a duckdb db or something for persistent state?

That’s brilliant with WASM! When that first came out I was hoping you could just “ship” apps in a zip folder, but found you still need to launch via CLI for browser security. This is a great middle ground

So thrilled you're working on this!! A few years back I tested a number of abandoned projects in the vein. Would be an incredible feature for distribution!

What Are Software-Defined Assets? Software-defined assets (SDAs) shift pipeline thinking from steps to outcomes, letting teams manage assets with clarity, observability, and context.

I think data eng practices actually give a slightly better model -- specifically, I like pulling on the thread of @dagster.io 's software-defined assets philosophy to focus more on what you are trying to produce vs the process

dagster.io/blog/softwar...

(3/3)

A few big diffs IMHO:

- Fundamentally interactive/HITL. Automate too rigidly and you get generic AutoML vs best fit-for-use solution
- More focused on generating artifacts (datasets, model objects, etc.) than functionality
- Work less "chunkable" w/ implications for project planning

(2/)

Definitely worth thinking a lot about where the abstraction holds and where it breaks! I've both extremes

I think there are table-stakes code hygiene that 100% apply (modularity, DRY, version control, docs, style guides, etc.)

But some big differences too

(1/)

Amusing how 99% of people using LLMs forget how these things work:

They are advanced probability machines. They generate the next most likely token (word) based in the input and their training.

Under the hood, it’s a giant matrix multiplication that has eerily good output.

Tonight’s the night: PyData Seattle’s CFP & Labor Day Flash Sale will CLOSE tonight at midnight! 😱

This is your final chance to share your technical talks with our technical people, and even get 25% off your ticket! Don’t let this opportunity slip by, GO GO GO! pydata.org/seattle2025

Python documentary is a fantastic long weekend watch! Love the balance in the story between the tech and community 🍿

➡️ Deadline approaching—only one month left to send in your papers and presentation proposals for #CDSM2025!

🚨 𝗖𝗮𝗹𝗹 𝗳𝗼𝗿 𝗣𝗮𝗽𝗲𝗿𝘀: 𝗖𝗮𝘂𝘀𝗮𝗹 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗠𝗲𝗲𝘁𝗶𝗻𝗴 𝟮𝟬𝟮𝟱 🚨
📅 𝗡𝗼𝘃 𝟭𝟮–𝟭𝟯, 𝟮𝟬𝟮𝟱 (𝗩𝗶𝗿𝘁𝘂𝗮𝗹)
📥 Submission Deadline: 𝗦𝗲𝗽𝘁 𝟯𝟬, 𝟮𝟬𝟮𝟱

@libbyheeren.bsky.social — a particularly relevant cross over: bsky.app/profile/amst...

I feel like there’s a Producers joke here but I’ve never seen it

Ok but like where do we donate!

The amount of alpha in this post is off the charts

www.seangoedecke.com/good-system-...

Screenshot of cases from the U.S. CDC’s website showing the number of national measles cases indexed by rash onset date.

Are they going down?

a woman says they 're the same picture while sitting in front of a window ALT: a woman says they 're the same picture while sitting in front of a window

Right! Same with the Twitter-era #rstats and dbt communities. So much love for software best practices, reproducibility, craft but surprisingly little overlap

Makes sense not all AE need models, but I'm sometimes surprised how little DS lean into databases versus python

Finally (awkwardly) wrote something up! bsky.app/profile/emil...

Screenshot of "Limitations" section from linked blog post

NGL it's not perfect or right for all use cases (e.g. batch only, limited model types) so YMMV. Still some rough edges and bugs, too. I might not rush to deploy in enterprise tomorrow but definitely a project to watch and something I can definitely imagine using in some personal/volunteer work

5/5

Screenshot of "MLOps Challenges" section from linked blog post

Screenshot "Deploying to the database" subsection of linked blogpost

Screenshot of "Key Features and Design Choices" from linked blog post

Exciting because:

- dbt has some of the boilerplate needed for MLOps (tests, logs, orchestration)
- DBs integrated with other systems like CRMs/dashboard so its easy to serve predictions from there

But gotta tweak both orbital and dbt to get the most benefit, with an assist from {sqlglot}

4/