tk90

I'm working on exactly this! I'm building a small model that classifies the correct DOM node containing an HTML's article content/title/date/author (given a raw html with a lot of noise/chrome). A fun learning exercise :)

30KB model, 40-50ms inference. Pretty happy with the results so far!

I can see an entire industry of tiny models like this, now that we have AI to help us do the grunt setup work (validation/training data creation, data cleaning, etc). Or just use a general classifier like Jev/Kev ha

> I find the post cursor groks are phenomenal at frontend web development, though Claude is much better at backend ruby.

Wonder if we'd benefit from a much more specialized + task-specific benchmarks to paint a clearer picture like this. A benchmark solely for frontend, ruby, hardware, etc.

closest thing I've seen is https://github.com/tobi/qmd, though it requires installing 3 local embedding models (300MB, 640MB, 1.1GB).

A couple weeks ago my Ghostty session crashed and found that it was using 40GB of RAM(!) - glad this was resolved!

Pretty cool, especially the voice and background music - feels just right.

I asked it about pointers in Rust. The transcript and images were great, very approachable!

"Do not let your computer sleep" -> is this using GPU on my machine or something?

If you found this interesting, I highly recommend reading "The Righteous Mind" by Jonathan Haidt. It's deeply impacted how I think of morality and politics from a societal and psychological point of view.

Some ideas in the book:

- Humans are tribal, validation-seeking animals. We make emotional snap judgments first and gather reasons to support those snap judgments second.

- The reason the political right is so cohesive (vs the left) is because they have a very consistent and shared understanding and definitions of what Haidt calls the 5 "moral taste receptors" - care, fairness, loyalty, authority, sanctity. Whereas the left trades off that cohesive understanding with diversity.

thank you!

Hi Peter! Thank you for doing this. Are you seeing any issues/concerns with Green Card holders traveling internationally and returning to the US in the past few months?

Is there a Rust equivalent to this? I'd like to dive into a rust "api_starter" as someone with mostly NodeJS experience!

> Can they read code and debug it in their head?

> Strong engineers, however, can break out the two problems and solve them at the same time.

Disagree so much with this. A "good" engineer breaks a problem down, solves them one by one, and avoids mentally juggling multiple things at once if possible.

This post's headline should've been: "100% bootstrapped solo founder sells for $80M in 6 months"

Spark AI | Full Stack Engineer | Full-time | Onsite | San Francisco

Hi HN! We’re Spark, building an AI research tool to help deploy solar farms and battery plants faster.

Turns out the biggest challenge with building more solar farms is not construction - it’s navigating local regulation. We build AI agents that help large energy developers find and understand critical information to develop and invest in energy projects.

Please reach out if you:

* Love writing code, but also care about having significant impact.

* Want to work on an AI product that helps power more homes or learn about the energy industry!

* Want to be part of an early team and shape the product and culture.

We’re funded by top investors, including AI Grant (Nat Friedman and Daniel Gross), the founders of Brex, Plaid, Helioscope (acq. Aurora Solar), and more (to be announced).

Full Stack Software Engineer - https://www.workatastartup.com/jobs/73632

Exactly. Reminds me of all the "technically superior" crypto coins that failed, and what ended up winning were the popular memecoins like dogecoin. There's a lot to say about distribution and what "the masses" end up adopting, whether or not it's the "better" product!

This is super cool! Curious how you got the data, web scraping I assume?

"agent" is a buzzword. All it is, is a bunch of LLM calls in a while loop.

- 'rag' is meaningless as a concept. imagine calling a web app 'database augmented programming'. [1]

- 'agent' probably just means 'run an llm in a loop' [1]

[1] https://x.com/atroyn/status/1819396701217870102

tinder has a great PWA web app: tinder.com

out of curiosity, what's the technical solution? My guess: add an idempotency key to the request and a message queue? Then when you try to consume it, you check whether that request was made previously.

RAG?

If you want to implement this in a AWS/ALB (Load Balancer) setup, you have to:

- Remove the ALB's HTTP listener

- Remove port 80 from the ALB's security group

I don't see how a RAG that can cite sources from a document doesn't addresses this already (example [1])? Although I admit it's not a perfect solution yet...

[1] https://mattyyeung.github.io/deterministic-quoting

The "Recall" part is interesting - I wonder if they're running RAG locally to achieve that? Interested to see how it works under the hood...

Isn't there retrieval degradation with such a large context size? I would still think that a RAG system on 128K is still better than No Rag + 1M context window, no? (assuming text only)