jeromechoo

This is a fantastic resource. The author only briefly covers library databases, but there's so much more in structured querying that could be worth covering.

For example, you can use Sparql to perform structured queries on Wikidata (a structured database version of Wikipedia) to get well beyond unstructured documents.

Here's every person in Wikibase that was born in NYC: https://query.wikidata.org/#%23Humans%20born%20in%20New%20Yo...

> You knew. And you signed off anyway. Because the alternative was losing the job, and the job was the mortgage, and the school fees, and the visa, and the version of yourself who'd fix it later once things stabilized.

I felt the pang in my bones reading this. All of us peons are just wading through this brave new world trying to do what we know is right but ultimately having no choice but to give in to life's needs.

This needs to be an Apple TV app!! Way more fun than watching fly by skyscrapers.

I think many developers worth their salt will argue the same. Cloud is and has always been a shortcut to buying your own hardware. Local models will get better and smaller. Qwen3-coder-next runs on a Spark and is as capable as Sonnet 4.5. Bonsai released a 1-bit model yesterday.

I also like the freedom of not having to ration a daily allowance of tokens.

There are a few. Try this one: https://ooh.directory/

The dilemma we're battling with here is the morality of avoiding most of your taxes if you can afford to hire the right people to manage your money.

Would it still be justified if we replaced "taxes" with "judgement in the afterlife"?

What does “bringing (RSS) back properly” entail in your eyes?

It’s still alive. Many sites still use it. Many people still subscribe to those sites. RSS reader apps are still being created to this day.

15/18 on food. Been to and eaten at all three countries. It was mostly instinct TBH. I’m not sure I can point out exactly what characteristics make a particular picture of food Korean/Chinese/Japanese.

That said I really love food. I cook all 3 types often and go out to eat all 3 types (and even regional variants) quite frequently.

One unobtrusive ad in the middle of an article isn't going to pay for a journalist and their camera person in a war zone.

By "news should be free" I think you probably mean "taxpayer funded".

This is an American-centric POV. I can’t speak for Europe but in many parts of Asia payments are most commonly done as a transfer of digital cash rather than credit. Whether it’s “consumer friendly” is not particularly relevant. People already accept it as a way to transact and there’s very little reason to switch to credit cards.

parse.com was my last straw building on "as a service" startups because of this. DaaS is not even particularly good for hobby projects anymore given how easy it is to work with sqlite.

Response from Clawdbot author when I said this: https://masto.ai/@jeromechoo/115928552690869904

My experience of 100% agentic coding has been roughly the same as yours. That said, starting an agent off on a task has been the single most productive step I have introduced to my workflow in awhile.

95% of the time the code doesn't even build but it gets all the jigsaw pieces in place and it's a million times easier to start deleting and moving pieces around than to start from scratch.

Building an index is easy. Building a fresh index is extremely hard.

Ranking an index is hard. It's not just BM25 or cosine similarity. How do you prioritize certain domains over others? How do you rank homepages that typically have no real content in them for navigational queries?

Changing the behavior of 90% of the non-Chinese internet is unraveling 25 years and billions of dollars spent on ensuring Google is the default and sometimes only option.

Historically, it takes a significant technological counter position or anti-trust breakup for a behemoth like Google to lose its footing. Unfortunately for us, Google is currently competing well in the only true technological threat to their existence to appear in decades.

I’m sure this is great on desktop but lack of mobile support today is kindof a bummer. It doesn’t even degrade gracefully.

I got a $1500 quote to replace a stuck thermostat in my car from my local mechanic. After some Youtubing I was able to replace the part myself with $150 in parts.

I got a $5000 quote to fix my AC 2 summers ago and no amount of Youtube was able to help me DIY a fix.

Maybe in the long term there will be more HVAC techs than auto mechanics. Somehow I don't think that's likely.

There are two paths to KG generation today and both are problematic in their own ways. 1. Natural Language Processing (NLP) 2. LLM

NLP is fast but requires a model that is trained on an ontology that works with your data. Once you do, it’s a matter of simply feeling the model your bazillion CSVs and PDFs.

LLMs are slow but way easier to start as ontologies can be generated on the fly. This is a double edged sword however as LLMs have a tendency to lose fidelity and consistency on edge naming.

I work in NLP, which is the most used in practice as it’s far more consistent and explainable in very large corpora. But the difficulty in starting a fresh ontology dead ends many projects.

I agree with you but the only time execs were held personally responsible that I can remember is Enron and that was on insider trading charges. There is so much plausible deniability and near unlimited appeals built into the system that it would take mountains of evidence and several hundred Lina Khans for an actually responsible executive to be tried for something like this.

I joined my current company to work on Growth. I was added to Gitlab, and for the first 3 months I pushed all my commits as MRs that my manager reviewed and merged into main. Standard procedure.

One day I needed to get a hotfix out to prod STAT. I pinged my manager to accept the MR and explained all the testing I've done. He said I could just accept it myself if I wanted it up now.

Turns out I've had the permission to push to prod since day one. The only red tape I had to cross was my own confidence.

This is awesome. I built a real time whiteboarding app for teachers over 10 years ago on the backbone of the original Firebase service.

It was so fast I was able to build basic collision physics of letter tiles and have their positions sync to multiple clients at once. What a shame to be killed by Google.

I haven't had a need for real time databasing since, but this is inspiring me to build another collaborative app.

Too many people must've turned on their Windows 11 computers at once.

I wrote a tiny self-hosted price tracker that doesn't require a server to run.

It idles as an automator calendar alarm and runs on schedule like a calendar event. When a price drop is detected, it displays a notification using osascript.

When my macbook is asleep, it waits to check the price on wake.

The only input required is the URL of the product page. You can include a target price if you like.

No need to sell your soul to price tracking apps!

Mac only. Because I only use a mac.

This actually looks 100% playable in the office.

I’ve had 3 SD cards fail on me in the last year. I now avoid using them as serious long term storage.

To be fair, these SD cards were exposed to fairly extreme Texas temperatures. One in a car dashcam, the other in an outdoor camera.

We maintain a list of around 1000 major publishers across the world and we crawl it every 15 minutes. For every other publisher (smaller blogs, etc..), they come through our global crawl.

The list itself isn’t particularly hard to maintain. What’s hard are the myriad of rules and configurations required to crawl and scrape each publisher. We built a model that extracts article data and it does a good job figuring out headlines, images, authors, and text.

Scraping rules are very self-manageable if you're planning on crawling just a few publishers. But jt gets exponentially more difficult to crawl hundreds.

Back when I managed a small coworking space, one of our most loyal members paid the premium for 1 of 2 private offices we had where he would show up to every day and shut the door for 8 hours working in the dark. He liked our rowdiness outside but hated being in the middle of it.

People know what they need to be productive. Everything else just gets in the way.

Ultimately this has very little to do with “the right office layout” and far more to do with how much trust your employer affords to its employees to manage themselves by default.

I don't mind it. But as others have already said, a single website can easily break a web ring. Why not a simple "I'm Feeling Lucky" button that takes you to a random website in the ring?

Nice! Yes, coreference resolution is surprisingly absent even in enterprise NLP.

Depends on what you mean by "better". With more accuracy?

Not a paper, but NLP extraction into knowledge graph representations do exist and are in use today. For example, here's a general purpose NLP model that links organization and people entities (among others) to each other based on factual relationships described in the text — https://demo.nl.diffbot.com

This semantic extraction can be extrapolated to most any trainable context. A useful one I've worked with involved mapping supplier-partner relationships. A well built supply chain graph can identify every layer of risk in a single supply chain and provide the provenance to back it up.

IMO, the biggest blocker to more mainstream use of Knowledge Graphs (even in the commercial world) is an actually intuitive interface for knowledge exploration. The real market innovation behind GPT isn't its 175 billion parameters, its the prompt interface that makes ChatGPT so universally accessible.

I actually do this for every service I put my email down for. It’s been about 2 years since I started.

Fortunately (unfortunately?) my email has only been sold once, and it wasn’t as egregious as you might think.

Amplitude, the user analytics company, sold my address to at least 3 companies who simply started emailing me as if I’ve always been a subscriber to their newsletter.

I do use their free plan though so I’m not mad about it.