Here's a google collab notebook built from this. It takes ~2 hours on A100 GPU if you have collab pro. Might work on free account as well.

https://colab.research.google.com/drive/1dklqzK8TDPfbPbyHrk3...

Research Paper https://arxiv.org/pdf/2405.19040

...and if you're in the vanishingly small overlap of folks reading this comment and people interesting in attending an academic talk in Denver next Wednesday, the official conference page for the paper is https://popl25.sigplan.org/details/POPL-2025-popl-research-p...

(The ArXiV preprint has the exact same content)

Oh, I'll be on the live streams!

https://popl25.sigplan.org/attending/live-streams

I want to say that the cultural changes inside of the ACM to make historical research open access and to have excellent live streams of the conferences is just so damn wholesome and wonderful. Thank you ACM and the people inside the ACM that made this happen.

And in case someone from the ACM is reading this, the live streams are very useful for physical attendees. I was attending Splash! and there were a ton of talks where I would have needed to change rooms, wanted lots of desk space for notes and research. It was somewhat ironic attending half a day from a vacation rental. :)

The presentation (a PDF with the slides and the talk transcript) are now linked from https://typesafety.net/rob/blog/fclp-at-popl

Enjoy

let button = document.querySelector('.main-btn'); document.querySelector('.main-btn'); if (button) { let clickCount = 0; const interval = setInterval(() => { button.click(); clickCount++; if (clickCount >= 10000000) { clearInterval(interval); console.log('Clicked 1000 times!'); } }, 10); // Adjust the interval (10ms) if necessary } else { console.log('Button not found!'); }

MCP is a protocol for tool usage, where as KAG is for knowledge representation and information retrieval.

Build a tool on top of the LLM layer for a specific use case. That'll get you up to speed. You haven't missed much.

Exactly. Avoid intentionally throw-away effort and instead attempt to build something specific and practical. Learn by doing.

Assume you are the author, I found your website is such a treasure. I couldn't believe your website wasn't listed on HN before. Next time please post on HN when you write.

I am not, just a huge fan of prolog and works of passion like this.

+1, this entire website is a fascinating find. Interestingly, in terms of formatting, all the long-form texts on the site are centered (as opposed to left align). I wonder what's the logic behind this decision (surprisingly, it is not at all annoying to read, though).

https://meetfebin.com/apps/filmflow Uses GPT for visualizing and analyzing the emotional pulse of films.

Loved your previous book, will the new book "Practical Math for Programmers" will be released this year?

Sorry, but no. I've got too much going on. But I am at ~200 pages of a draft

Somethings that helped me.

1. I realised the hard way that you can't focus on too many things at the same time.

2. Ruthlessly cutting down pursuits to one thing helps drive results.

3. Focus on areas you have circle of competence in.

4. I struggle with abstract ideas, and jump to projects without thinking the details, scoping abstract ideas and writing down concrete POA helps in completing things.

5. Working hard on staying loyal to my resolutions. I took a resolution last month to write a book on rust for beginners. I constantly hold this thought every day in my head. It helps a lot in making progress and saying no to distractions.

> 2. Ruthlessly cutting down pursuits to one thing helps drive results.

This worked really well for me until I actually settled in with a long term partner, who was rightfully frustrated when I'd try to cut out some of the daily responsibilities that are required to be a good roommate and partner.

When you're single you have a lot more freedom to squirrel away for a while on a problem, but it's a tremendously selfish thing to do when your life is intertwined with another person who depends on you to juggle your daily obligations like everyone else.

1. Chat GPT3 or Similar Tech Enabled Personalised Social Engineering Attacks From Stolen Data.

2. A new social network will rise that will have an explosion of reels made from Stable Diffusion tech.

3. Apple creating a dedicated program to encourage developers to build apps leverage M1/M2s GPU Capabilities.

4. Reserve Bank of India creating a CDBC enabled savings account program to allow citizens to hold money without intermediary banks.

> 4. Reserve Bank of India creating a CDBC enabled savings account program to allow citizens to hold money without intermediary banks.

It won’t be a savings account program in the sense of earning any interest. The Reserve Bank of India is quite cautious not to hit the private banking system by introducing competition with it, and hence will only offer zero interest accounts (and possibly experiment with small negative interest rates after a few years).

2. Why or how? People already hate computer generated art/content/copy etc. It makes the media feel fake/useless/unoriginal.

Let's take a reel trending on Instagram, even though same dance moves is copied countless times, it still gets views.

Prompt Engineering if utilised by artists who knows how to steal, they can deceive people that generated output is original.

A social network that can streamline these into a simple processes can massively onboard new artists who can create art with drastically less efforts.

I think social media + deception is the opposite of the recipe for success going into the next unicorn. (Unless it’s obviously intended, for example having an avatar or a game character as your profile.)

Hence why BeReal became a thing but didn’t take off because it became too easy to fake (user is given a few seconds to correct front photos) + it’s super boring to look at ugly photos from a dark sofa.

People want beautiful hi res real media or completely animated.