Hey! I’m Dumky, Developer & Agent Experience Engineer at MotherDuck, the serverless analytics platform built on DuckDB. With a background as a Senior Analytics Engineer, I’m passionate about making data engineering accessible and practical. Personal blog at dumky.net
You can basically SELECT * the internet by querying the public Common Crawl dataset on S3 with DuckDB. It's a great way to explore how the AI-pilled web (Vercel, Lovable, Replit, Bolt, Cloudflare, GitHub) compare to the traditional Wordpress-Blogspot hegemony.
There is a surprising amount of science and technology to make your smartphone work on a train (h/t Deutsche Bahn). On a recent train ride from Berlin to Amsterdam I Clauded into a rabbit hole and learned about 5G in a way that I could not have imagined before: visually, in a socratic conversation, with analogies and sources throughout. Not sure I had to know why standalone 5G is better than new radio (NR) 5G bolted on LTE towers, but it definitely satisfied my curiosity.
Compared to what exactly? Snowflake? Hiring an engineer to deploy DuckDB? A hobby project? FWIW I work at MotherDuck so obviously biased, but curious to hear what makes you say that.
I mean, it is one thing to build things for yourself or your company this way. It's a whole other thing to dump it on the rest of the world. I'm still flabbergasted by the fact that maintainers of projects I highly value have to shut down contributions due to the amount of AI slop contributions they receive, making it impossible to maintain the project.
100% agree with the curators part. I think this is often implicit, but we look for signals of quality, whether that's Github stars or a person we trust. I feel the sense of what is a good curator has shifted, or even the curators are overwhelmed. Similar to the enshittification paradigm: once you find a good source of curated content, let's say Substack, then it grows and needs its own curation.
I've been overwhelmed with the flood of interesting new things being released. I tried to put a number on it and across the board from Github repos to package registries to Show HN submissions there is just such an immense increase in output, especially since Claude Code and Codex. I think everyone has felt this, but I have not yet found a way to deal with filtering out the genuine quality from the AI slop.
It’s interesting you mention linguistics because I feel a lot of the discussions around AI come back to early 20th century linguistics debates between Russel, Wittgenstein and later Chomsky. I tend to side with (later) Wittgenstein’s perception that language is inherently a social construct. He gives the example of a “game” where there’s no meaningful overlap between e.g. Olympic Games and Monopoly, yet we understand very well what game we’re talking about because of our social constructs. I would argue that LLMs are highly effective at understanding (or at least emulating) social constructs because of their training data. That makes them excellent at language even without a full understanding of the world.
Most of our jobs consist of working with tools. Yet it’s very hard to get insights into which tools are required most, are growing in your area, etc. So I decided to keep track of tools and technologies mentioned in the data space by keeping track of job openings for the last two years. Now I’ve opened up that data set. Here’s an analysis for jobs per data warehouse: https://selectfrom.work/insights/data_warehouses
By far the stupidest version of this to me has been Snowflake’s implementation of previews. This is a database, where people preview the content of a table, not in an app, not on a phone, and someone thought it was a good idea to make that an image. I have no idea who ever thought this was a good idea, but here i am constantly tricked into thinking I can select some preview data, only to realise I have to go on a 10 clicks and a SQl query diversion to get it done.
I agree that there is a line at using someone else’s data to make a profit, but it is kind of ironic that you mention Google, because their exact business model is scraping websites to feed their search results and litter it with ads to make a profit. For me there is a big line between aggregating publicly available data (search results, reviews, news, job postings, etc. ) and intentionally violating terms of service like signing up for fake accounts an harvesting user data. So entitled maybe not (sites can try to prevent you from scraping), but if you make something publicly available you shouldn’t be surprised when people use it in ways you may not originally have intended (within legal boundaries of course).
Am I missing something? This “hack” requires you to go to his site first, then use the back button and then click on a (fake) competitor link. How is he ever going to get people to his site in the first place? And if it’s through paid ads, why not create a fake paid ad that directs you straight to his fake site in the first place? All sounds very much like a marketer who uses the veil of “security researcher” to hide a scam.
All you have to do is start a new country called Http, convince ICANN to adopt it as a new TLD (will need a lot of persuasion) and serve “http” as a dotless domain. But, you know, anything for a beer… (fyi: the host name is the part after @ and before the port number indicated by “:”)
You’re right, but only if the company wouldn’t track whether you’ve seen or even received that message. So yes, general or even contextual messages would be allowed, but “You haven’t seen X in 9 days” would imply processing personal data for marketing purposes.
The current state of browser tracking preventions also means that you’re unlikely to identify conversions from the same user that saw your experiment after a week or sometimes even 24 hours.
Most, or at least a lot, of the prefetching is for third party libraries (think jQuery, Google Fonts, Facebook Pixel, etc). There’s a general speed advantage for users caching commonly used libraries and fonts across sites. Nonetheless I believe prefetch will still have a speed advantage even when the cache is segregated.
There’s definitely high demand for “technical” web analysts at the moment. That usually means someone who isn’t afraid of html and JavaScript and can help less technical marketeers and analysts implement their measurement requirements and analytics tools either through a tag manager like Google Tag Manager or directly in code.
It’s true, but GA also does a pretty good job at filtering out bot traffic for example, so it depends on what your definition of more accurate is. There are also ways to send GA hits through a sub domain to make it look self hosted and bypass ad blockers.