thomasfromcdnjs

Great work. Resume parsing is hard business. I've been using https://turbo.build/repo in recent projects, looking through your setup, it would probably be worth checking out.

I updated the themes page a couple weeks ago -> https://jsonresume.org/themes/

But I could also try resurrect the theme that you were using.

I still maintain the project, currently rebuilding a lot of the infrastructure.

Ive been having fun integrating AI stuff into it. Going to polish them more soon.

https://registry.jsonresume.org/thomasdavis/jobs <- this is purely a prototype i built on the weekend, it uses vector databases to match your resume to Who is Hiring hn post.

https://registry.jsonresume.org/thomasdavis/interview <- this one preprompts your resume so you can interview yourself

https://registry.jsonresume.org/thomasdavis.letter <- Cover Letter (attempts to write a cover letter that matches your resume)

https://registry.jsonresume.org/thomasdavis.suggest <- Resume Suggestions (it asks ChatGPT to give precide feedback for improvements you could make to your resume)

The creator here, can help you out. I moved it all to vercel functions running on node 18. Broke a bunch of themes in the process.

How to build a custom CMS and new frontend codebase in 2 months that can handle 3 billion hits per year in a corporate setting.

Would be more interested in how your work flow was managed.

I love using clicky.com, homepage looks old school but the app is modern.

Has that simple, show-people-what-they-actually-care-about dashboard vibe, but has lots of hidden functionality around.

It also let's you see the lifetime of a single visitor in an easy fashion -> https://i.imgur.com/xeOnn2V.png

I still install GA on my sites but I never look at it.

I think metricwave.com and counter.dev style analytics are a bit too simplistic, would be nice to see just a few more features on top such as reports, alerts and integrations

Jokes aside, just use pnpm.

Commercial grade apps where I do throttle CPU.

If you are only re-rendering what is required, it's just a webpage at rest.

That is so beautiful.

I worked at the checkouts of a supermarket for a year. We had about 50 donation boxes for children, diseases, animals etc

Depending on how much you spent, you would receive X amount of tokens to put in the boxes.

My job was to clean the boxes and tally it etc

95% of tokens were in the animal boxes.

Yeah, that sounds so strange. I've had thousands if not tens of thousands of components, reacting to objects that contain millions of data points that felt native.

Whenever someone compares dextroamphetamine or methylphenidate to modafinil, I don't take them seriously.

They don't begin to compare in effect.

Agreed.

The author is upset for XYZ "other" reasons, I would have recommended they slept on it and not post that publicly.

Are you saying that it return only JSON before? I'm with the other commenters it was wildly variable and always at least said "Here is your response" which doesn't parse well.

Technical question: What code in the websiteHandler.ts is responsible for spidering the website in question?

https://github.com/openchatai/OpenChat/blob/main/llm-server/...

aha love this

I've built diagram like tools several times, constantly umming and ahing between Canvas and SVG.

SVG seems to get you pretty far, but I always end up with Canvas in the long run.

Fun fact: Google docs is actually a canvas.

Agree heavily with the article. As an app developer, I actively spend my time walking away from frameworks/languages/libraries that start mashing the server weirdly into client code. Each to their own but I think quasi-RPC implementations are a relic of the past.

Treating the network layer as an explicit abstraction to components just feels a better developer experience in the long run. The data abstractions are easier to reason about, and caching albeit always hard, is a lot easier to think about.

I don't even know _why_ React/Next is pushing Server Components. (Part of me wants to join the conspiratorial bandwagon like the author)

There is a lot of engineering work and reasoning to make sufficiently complex prompts.

Depending on your business case, the 4096 tokens given to you have to go quite far. Vector embeddings are not "easy" to work with. Trying to splat together a range of techniques to craft a good prompt is hard™.

Adding in Actions (e.g. using headless browsers to open pages etc) is also pioneering territory.

Sucks that OpenAI currently has the market, but there is still plenty of reasons to develop on top of it.

An example from today was a YC company and they had an ad on the front page.

https://deep-ivy-ltd.breezy.hr/p/055a500d0e8701-ml-research-...

45-60k

● Very strong Python skills, with deep expertise in PyTorch, TensorFlow or JAX

● Very strong skills in recursive programming. Check out the Ivy Container class

45-60k would be a Wordpress developer in my opinion, I'd imagine deep expertise in ML cost a lot more.

Pretty much a 100% yes rate, it was novel at the time, early 2020.

I made a Tinder bot in the first month of GPT-3's public API release. The ethics made me squirm so I always asked for permission to turn it on and told people when it was on and off. The models moral alignment was way more "friendly" back then so turned it into some pretty saucy chats. The model was better than me at replying, still single though.

(It was simply holding onto the last 10-20 messages and injecting it into the prompt)

I've got my own invented version of this style. I will boil them in a cm of water or something at the perfect time it evaporates I then add the oil.

I don't like them like the Americans do but there is lots of to play with it. (I've even started boiling onions whole before I use them in a bunch of other dishes)

Highly recommend it.

I've never been interested in the automated matching aspect of HR but this piece has me inspired to attempt the challenge.

To add a point, my favorite reason to use recruiters, is having them do the negotiation process for me.

High quality post as usual.

I'd wager that "Seaflooding" will just happen somewhere anyway, likely some where we ain't really looking.

And it will be fine. Mankind has seemingly surpassed weather patterns, and it now the 21st century.

Baha I had the same what moments.

Already down aha I lol'ed

I barely though it was violent.

Everyone should read it for the writing style alone, it's great.

Great work, the space needs some more tooling in this direction.

I keep going back and fourth between the two.

I have absolutely no proof but sometimes feel like the responses I get are weaker if there is no white space in the structured data.