Stormy Shippy

Thanks for the high quality swag @firecrawl - my default recommendation for fastest way to scrape and crawl sites and get screenshots. Just point your agent to their docs and off you go.

I feel sorry for my teammate, however, this reason for missing standup is comical: "Hey team, This morning I fell while shoveling snow. I am not able to sit properly so I wont be able to join the stand up."

@ardabrowski Do you happen to use the “not interested” interaction? I found that I had a political heavy feed a year or so back until I started to just say not interested and now my “for you” feed is quite good with only a modest amount of topic du jour.

The algorithm sucks. This is my feed now. All politics. I've been here for 10 years now, mostly for tech circles. Never followed politicians, on purpose. But I guess it doesn't matter now.

@nbaschez While I don’t have a name for it, you can use github.com/fxstein/ascii-… to fix it up automatically

we need a name for this

@dhasandev @SherryYanJiang it is quite common - you may want to incorporate something similar to this if you diagram in ascii a lot github.com/fxstein/ascii-…

@SherryYanJiang Damn, so it’s not just me getting those fucked up sidebars

one underrated thing i’ve been doing w/ opus 4.5 in cursor: i ask it to sketch the first design as terminal charts. super fast way to see structure + flows before i touch real ui. feels like thinking in ascii rn.

@kevinkern Love this approach myself, but out of the box you get quite a few off by one line errors that require a bit of back and forth revisions or a skill to minimize/reduce

Switched from figma to ascii

@ojoshe github.com/jlevy/flowmark

@ojoshe you continue to contribute to bloomreach all these years later...I just incorporated flowmark into part of our docs pipeline

Feature request for @meltyinc Conductor: Open in Sublime Merge by @sublimehq or @kaleidoscopeapp for some more robust workspace diff views -- @charlieholtz

@jobergum The hook made me want to read the surrounding context not in the screenshot: linkedin.com/pulse/building…

Search has a new user now, and I’m here for it

@sspaeti Reading this part felt a little out of place: “Based on your notes, I'll write two focused examples and add a third section pointing to other key implementations:”

I'm looking forward that you can dive in. Your feedback, questions, or corrections are always welcome and incredibly valuable 🙏🏻. Feel free to reply to me or try the new commentary feature—Simon

I am happy to share that I released my second pattern: «Data-Asset Reusability Pattern», focusing on efficient data reuse (dedp.online/part-2/5-dep/d…). This chapter explores reusability in data engineering with dedicated sub-patterns. But first, what is data-asset reusability?

@ojoshe Saving the next reader a google search: copier.readthedocs.io/en/stable/

Maybe it's due to the nondescript name, but I am starting to think one of the most underrated Python tools out there is copier. Being able to maintain projects of code or content off of an a template, modify your copy, and then later backfill template updates simply by running `copier update` and doing git merges just ... awesome. This feels kind of esoteric but let me try to explain why. As we've seen with the rise of Cursor-style code editors, any time you can remove repeated inefficiencies (like copying to/from ChatGPT) in your editing processes it is like good investments—it offers a compounding benefit. That's also why git+GitHub turned out to be so powerful: cheap and flexible branching and merging radically improves the way you build software. Don't have time to test something right now? Commit it in a branch! We also over time figured out package management (like npm or pypi), where you package code in pieces,manage dependencies across packages, then later update and resolving the dependencies and conflicts. This process is tricky and complex, and took (literally) decades to get these workflows right. But in spite of that, we've never had good tools that allow branching and merging of content/code (think of what you're really building) separately from tools/structures (think of everything that supports it—project build, dev tools, GitHub Actions, deploy scripts, etc.) Say you have a couple open source projects, and you want to upgrade the CI system or the linting. Do you fix them all at once? One at a time, but track them? Unless you have enough time you'll either be inconsistent or not do it at all. The same applies within enterprises, for example handling how you manage custom settings for different large customers. In copier, when you clone a template, you save the answers/customizations. Later, if the tooling improves, do `copier update` and then use git to review changes, resolve conflicts, and merge. Workflows can feel like the messy, unsexy parts of a project, but there's a reason what differentiates good and bad enterprise software is almost always whether they fit the workflows of a team, reducing when you're unblocked and tightening update loops. As move to more and more AI agent-type tools, we're managing even more workflows now. So it makes sense to think about what tools make our own workflows tighter and unblocked.

@vivekgalatage @pavpanchekha Pre-ordered the hardcover version: a.co/d/80vnfmz

As a usual practice with the new year, I recommend browser.engineering to anyone trying to learn how the browser works under the hood. This year, the print version of this amazing book will be available to order. All the best, @pavpanchekha and Chris, with the launch.

@marcbowes Glad this found its way to the right person. I figured it wasn’t related, but something about the release timings and the same number nagged at me and now I never have to think about it again.

@Stormy No :)

Is the Amazon Aurora DSQL limitation of a single transaction not modifying more than 10k rows a consequence of S3 Express One Zone’s 10k object append limit?

@tim_cook Good and expected news. It would be a shame and border-line criminal to pick up this show and not commit to seeing the book’s compelling and core plot lines through.

Excited to share that “Silo” will return for a third AND fourth season! We’re thrilled to support the imagination and inspiration out of the UK as they continue to create world-class films and series.

@JinaAI_ I’m a sucker for thoughtfully designed books like this. You’ve got my order.

This beautifully designed, 160-page volume brings together our most influential technical articles, featuring: - Comprehensive coverage of Jina AI's multilingual multimodal embeddings, rerankers and small LMs: jina-embeddings-v3, jina-clip-v2, jina-reranker-v2-base, reader-lm and jina-colbert-v2 - Insights into ColBERT late interaction and late chunking - Deep dives into modality and linguality gaps - Empirical studies exploring zero-shot classification and embeddings & rerankers vibes

2024 has been a wild year for search & us. Today we publish Re·Search, our 2024 yearbook—a carefully curated collection showcasing our most impactful technical achievements and breakthroughs in search foundation. Order yours now! Whether you're a longtime follower of our tech blog or new to the search technology, this volume compiles the pinnacle of our research and development in 2024.

@britg @rauchg I wanted to go this route but my city ordinances don’t allow them after talking to a retailer. Jealous.

@rauchg I bought an Endless Pool for this logistical reason - fits on a smallish patio endlesspools.com

This is the year I began switching from running to swimming 🏊. A logistical PITA but an infinitely superior form of cardio.

@petereliaskraft This idea might just be my entry point into kicking the tires with DBOS. I’ve got some Python scripts on scheduled GitHub Actions, but each invocation has to wait for a long running async job to finish from an external system, which gets expensive fast.

We recently had a user sign up for a paid account with a use case that surprised me–all they need to do is run a Python function every few minutes. The function doesn’t use any unique DBOS features, but instead calculates and posts the value of a financial index. They could have built this app without DBOS. So why use (and pay for) it? Simplicity. They wanted a serverless solution that wouldn’t require operations and maintenance, and what sold them on DBOS was that it “just worked.” They could take their existing function, add a “DBOS.scheduled” decorator to it, and it was running on a schedule locally. They could then run a “deploy” command and serverlessly deploy it with tracing and monitoring built in. DBOS helped them ship their tool in a couple days with full confidence it would keep running for as long as they need it.

@ilanbigio @OpenAI Would love to see these cross-hosted on YouTube for the overall convenience of being able to have them show up on my “watch later” list on the TV and I don’t forget about all this great content.

after designing and deploying ai solutions with 100s of companies we wanted to share our secrets. all of them announcing @openai build hours showcase learn about agents, evals, realtime, distillation, orchestration +++, directly from the solution architects & engineers @openai