malgorithms

> Without sample-and-hold circuitry, this meant the stereo channels were approximately 11 μs out of sync.

Am I correct in assuming this wouldn't be perceivable on a stereo, based on the speed of sound?

You’re probably in the process of learning this now, but in case you’re just getting started: effective multiple myeloma treatments are being approved at an astonishing rate. It was a death sentence 20 years ago. Now it’s indefinitely treatable for many people.

A close family member was diagnosed 5 years ago and went through a stem cell transplant at Dana farber and the cancer still hasn’t returned…although statistically by now I believe it should have. But when it does return there is now a massive menu of next treatments for her that will likely hold it at bay.

Things are changing so fast now that I’m not sure the stem cell treatment is the first step.

Good luck to your dad.

I'd been waiting/hoping for someone to make something like this. Well done. A few ideas that will not clutter the UI but you might consider:

  - when a user changes the score slider, encode that in the URL with a hash tag, so they can bookmark the page with their preferred settings
  - a left button allowing me to step back to yesterday's news
  - to simplify newsletter signups, just accept an e-mail address right on that page
  - your **advanced** options:
    - have GPT score each news story across common labels: science, politics, entertainment, news, etc. Then allow these as a filter. If I want to see the top science stories of the day, that should be easy.
    - have GPT write a 2 sentence summary of each story as a lead-in after the headline title
    - a user/saved whitelist/blacklist of news sites
    - any advanced setting should be shareable. For example, if someone puts the effort in to make a page with just Australian news sources, focused on sports, with a minimum score of 5.0, they could save that with a title that can be shared for anyone.

Congrats on a well-executed project.

According to GPT-4, and I think this is a good summary: A "Foundational Model" in AI programming refers to large-scale machine learning models that serve as a basis for a wide range of applications and tasks. These models are pre-trained on massive amounts of data and can be fine-tuned or adapted for specific tasks, domains, or applications.

One of the most well-known examples of a foundational model is the GPT (Generative Pre-trained Transformer) series developed by OpenAI. GPT models, like GPT-3 or GPT-4, are trained on large datasets containing diverse text from the internet, which enables them to generate human-like text, answer questions, translate languages, and perform various other tasks.

Foundational models are significant in the AI field because they allow researchers and developers to create a wide range of applications and solutions without having to train a new model from scratch for each specific task. This approach saves time, resources, and computational power while still providing a high level of performance across different tasks.

I say this because investors sit on boards and work alongside these lawyers once they invest. They learn pretty quickly who is great and can often give very direct recommendations: a specific lawyer, not just a firm. At the startup stage you should choose the lawyer, not the firm.

If you’re talking to investors at all, ask them for recommendations. That’s a good place to start.