emschwartz

Definitely put your finger on what I’ve been feeling.

What AI did was give every engineer a small team of tireless, fast, occasionally-wrong direct reports. And with the team came the manager’s problem. The discomfort engineers are feeling right now isn’t an AI problem. It’s a delegation problem, and delegation is the oldest unsolved problem in our discipline.

Very good synthesis.

Hilarious. Nice work

Could you put some screenshots in the readme? Thanks!

If you haven’t read Entangled Life: How Fungi Shape Our World, I can’t recommend it enough.

https://www.merlinsheldrake.com/entangled-life

I’ve been using MixedBread, which is a pretty old model at this point. Recently, I tried comparing it to some newer models and was disappointed that the results weren’t dramatically and uniformly better.

You probably can’t go wrong if you pick a recent one that scores decently well on benchmarks and is at the right price point (or memory requirement) for whatever you’re trying to do.

Thanks for that! I added the first one to the feedback board here https://feedback.scour.ing/142 and I'm tweaking the recommendations to avoid topics like Swedish filesystems.

Let me know if you have any other suggestions!

Developer here. Thanks for posting! I'm glad you've found it useful!

Let me know if you have any feedback on it.

Thanks for the write-up! Comprehension debt definitely feels like one of the biggest issues when I'm leaning heavily on coding agents.

It might fall into the too-heavy prior art category you mention at the end, but I quite like Superpowers (https://claude.com/plugins/superpowers and wrote a short review of why I like it: https://emschwartz.me/a-rave-review-of-superpowers-for-claud...).

Yeah I agree with the implementation plan not being that useful for me to read. I often just tell it to go after reading the design doc.

Seems kind of crazy, but I like it

This is great. It would be nice to be able to add a custom dictionary for words that are often misunderstood.

Pick an embedding model that supports binary quantization and then use a SIMD-optimized Hamming Distance function. I'm doing this for Scour and doing about 1.6 billion comparisons per second.

https://scour.ing

https://emschwartz.me/binary-vector-embeddings-are-so-cool/

Seems neat. Could you plug it in to DataFusion to use their SQL layer?

This is exciting work. I especially appreciate that it was trained to support binary quantization (I wrote about how cool binary quantized vector embeddings are in https://emschwartz.me/binary-vector-embeddings-are-so-cool/)

This sounds interesting.

Reading your website and the investor deck, one of my main questions was "who is behind this (and is this just AI-generated)?" It would be useful to put more of a bio on there.

Not having macOS/Windows support is going to make it hard to develop with. Would it be possible to build some kind of shim on top of other libraries that mirrors the API, even if it doesn't match the performance?

Also, one of the advantages of using a popular HTTP server stack is getting lots of battle-tested middleware that other people have developed and tested. Is there any way to leverage any of that or do you need to build everything from scratch? Granted, that is certainly somewhat easier in the AI era, but still.

I believe so. When you call `raw_sql`, the API doesn't provide a way for you to specify which parts of the query are parameters, so it just passes that exact string in to exec.

You should try Scour (https://scour.ing)!

You specify your interests as free form text, it ranks articles by how closely they match, and you can consume your Scour feed as an RSS feed to read it in NNW.

Disclaimer: I’m the developer

In Deno Sandbox, secrets never enter the environment. Code sees only a placeholder

The real key materializes only when the sandbox makes an outbound request to an approved host. If prompt-injected code tries to exfiltrate that placeholder to evil.com? Useless.

That seems clever.

Developer here, happy to answer any questions you might have about Scour!

Also, feedback on the product or the docs is very welcome!

I know the purpose of these pieces is to drive interest in companies that Amplify has invested in, but I appreciate the technical depth they do it with.

I'm seeing a Cloudflare Error 1102: Worker exceeded resource limits

It's working again for me

If you want to try filtering your RSS feeds based on topics you’re interested in, you should give Scour (https://scour.ing) a try. I was never a huge RSS user because I ended up with too much of the infinite unread email feeling. I built Scour to help find the signal among the noise.

You’re welcome! I wanted it to add to Scour (https://scour.ing) but glad it was helpful for someone else too!

Here's that list as an OPML file (for importing into a feed reader): https://gist.github.com/emschwartz/e6d2bf860ccc367fe37ff953b...

It includes the 92 of those blogs that have RSS/Atom feeds.

Duplicate https://news.ycombinator.com/item?id=46454413

How did you decide on pgvector versus a dedicated vector database? I’m curious because it didn’t sound like your architecture included Postgres before (though maybe I’m wrong about that).

I agree that Opus feels like a step change.

Claude code with sonnet was pretty good but needed a lot of back and forth to get to the right solution. Opus feels closer to a colleague, maybe not your absolute best colleague, but far from your worst.

It's very fun to try to make that limited data model do more than it was intended to.

When I did a little bit of contracting for TigerBeetle, I was working on these learning exercises and came up with some fun ingredients that could be used if you're trying to push what's possible https://github.com/tigerbeetle/tigerlings/pull/2

Glad you think so!

There isn't a way to block feeds (yet), but you can subscribe to specific feeds, which basically acts like creating your own list of allowed feeds.

Separately, I'm going to be working on letting you exclude content from certain domains (which was requested in https://feedback.scour.ing/33).