nattaylor

I'm drawn to this but it has several foot guns like INSERTs will fail for late arriving data if the constraints have been updated etc

Reminds me a little of htmz

htmz is a minimalist HTML microframework for creating interactive and modular web user interfaces with the familiar simplicity of plain HTML.

https://nattaylor.com

Building an email-to-calendar-feed service for all the mails from the multitude of services and attachments that I get related to my kindergartener.

My read is that no customers will leave since they are much more interested in news coverage -- and this helps the AP focus more on news.

This is a tangent, but I wonder if they feel that they are just creating LLM training data and that few readers (even of Sunday papers) will actually read their reviews.

The base model is Qwen2.5-VL-3B and the announcement says a limitation is "Model can suffer from hallucination"

I wish there were some explain plans in either post, since I don't get what's going on.

If the query uses the index, then the on the fly tsvector rechecks are only on the matches and the benchmark queries have LIMIT 10, so few rechecks right?

Edit: yes but the query predicates have conditions on 2 gin indexes, so I guess the planner chooses to recheck all the matches for one index first even though it could avoid lots of work by rechecking row-wise

Is pre-training in FP8 new?

Also, 10M input token context is insane!

EDIT: https://huggingface.co/meta-llama/Llama-3.1-405B is BF16 so yes, it seems training in FP8 is new.

S3 Tables is designed for storing and optimizing tabular data in S3 using Apache Iceberg, offering features like automatic optimization and fast query performance. SimpleDB is a NoSQL database service focused on providing simple indexing and querying capabilities without requiring a schema.

This is very cool. Kuzu has a ton of great blog content on all the ways they make Kuzu light and fast. WebLMM (or in the future chrome.ai.* etc) + embedded graph could make for some great UXes

At one time I thought I read that there was a project to embed Kuzu into DuckDB, but bringing a vector store natively into kuzu sounds even better.

500 errors for me

Doesn't compression make any minification gains negligible?

Sorry, I didn't mean to suggest your solution was brittle -- I actually quite like it and want to adopt it!

But I do think the legacy browser behavior with the ID attribute as window properties is very brittle for the reasons you suggest

If you like brittle things, the id attribute is already made into an attribute on the window for legacy reasons

Edit: My tone may have indicated that parent's solution was brittle. It's not!

https://we.phorge.it/ is a community fork that appears pretty active.

I was also very fond of Phabricator (all though my team preferred GitHub style pull requests) but I haven't had a need for it recently, so I haven't tried phorge myself.

On Chrome, I solve my too-many-tab issues with an extension [0] that closes the LRU tab once a threshold is reached (10 for me). I find the tabs I need are open and wide enough, and the tabs that autoclose were not useful anymore. About once a month I'm doing a research task where I actually want many tabs and I turn it off temporarily.

[0] https://chromewebstore.google.com/detail/max-tabs/ghhcibaghj...

I have a use case for this that I'm excited to try. I'm glad AlexG has put so much effort into this. Even the docs are pretty good!

My pyenv python3.12.2's sqlite won't load extensions even after installing with what I think are the correct command line flags. Argh!

My brew installed python3.12's sqlite will load extensions though, so I can proceed.

I was checking this out on my train ride home and I'm pretty excited to take it where it's been tonight. Well done team!

Yes, speed. 1,000,000+ QPS and 1,000s of ads to evaluate is common. At that scale it's very distributed, so replication is another challenge.

In a way, its a search problem where bid≈relevance and targeting_match≈recall, so I've seen Solr used here too.

A friend built a system where the ad campaigns with all their targeting rules were Lua scripts and it was also fast and simple in a glorious way!

In my experience with Snowflake support about a year ago, an administrator of the customer's account had to explicitly grant access to Snowflake in order for the Snowflake team to see or do anything -- and if I recall correctly the access had an expiry.

Boring things like running the proxy while I did manual QA / ran automated tests.

I quickly realized that if I wanted an up to date spec then I should do it properly in the application

Reminds me of https://github.com/alufers/mitmproxy2swagger which I discovered from this thread https://news.ycombinator.com/item?id=31354130

I generated some specs from that!

I ran into trouble keeping them up to date.

This is particularly interesting to me for Android/iOS. I can't even picture the use case where there'd be enough data on the device for the row-based format to b a bottleneck, but maybe some case that involves many, many aggregations

pg_analytics looks promising here https://github.com/paradedb/paradedb/tree/dev/pg_analytics

https://pypi.org/project/languagemodels/ can load some small models but forming JSON-reliably seems to require a larger-ish model (or fine tuning)

Aside: I expect Apple will do exactly what you're proposing and that's why they're exposing more APIs for system apps

In my experience it excels at the stated use case of OLAP in the cloud with compute separate from storage and claim it's unambiguously good for this.

For OLTP it would be unambiguously bad (although maybe there's hope with Unistore, which I haven't tried.)

Many workloads are a mix and so it can become ambiguous whether it's a great fit / great value / whatever you're defining good/bad-ness by

Photopea is amazing, so I have high hopes for Vectorpea

Whoever came up with the 1-3 month game changer oughta get a promotion

I think this parquet virtual table extension is a pretty neat example of the power of extension https://cldellow.com/2018/06/22/sqlite-parquet-vtable.html