Yagnesh Revar

Computing about computation to compute better computers so we can compute more efficiently about computation to computer better computers ...

[Yagnesh Revar @_max_entropy]

How many unspoken set of assumptions does it take to make this true?
(1) Lack of awareness about specific matter and calculus skills
(2) Change is a step function and not observable until it flips, or
(3) Nobody excludes folks who know a bit of math.
Any other guesses?

[Elon Musk @elonmusk]

True

[Robin Debreuil @debreuil]

@scaling01 The taming of electricity. We went from not knowing it existed, to making it talk to us, all in 100 years or so.

[Yagnesh Revar @_max_entropy]

This is such an underrated observation. I wish more people knew this and were honest about it.

[François Chollet @fchollet]

If you can learn one thing that's genuinely novel to you, you can learn anything.

[Yagnesh Revar @_max_entropy]

If there's any airline website that deserves an award, I'd definitely give Polish Airlines a LOT of votes!

[Yagnesh Revar @_max_entropy]

Built as an experiment in how far modern AI can compress the product-design loop. The accuracy, speed and, the cost of translation surprised me. More write-up in the article.

[Yagnesh Revar @_max_entropy]

Here’s the result of the product design experiment I mentioned earlier.

Σ-list, a prototype of an intelligent shopping list, reimagined.
https://t.co/004d0fIwS4

[Yagnesh Revar @_max_entropy]

Those who've been computer programmers in their journey: how has your experience been with the latest AI coding agents? I've been trying out Claude for the past few weeks, and I'm honestly blown away by its borrowed skills in coding a web application.

[Yagnesh Revar @_max_entropy]

It’s likely Anthropic may have to make their products worse at coding performance. Because Claude sometimes codes so well it’s almost like watching Netflix. And Netflix Standard with ads (equivalent to Claude permissions) costs $8.99/month, while Claude Max starts at $100!

[Yagnesh Revar @_max_entropy]

The definition of what science means is likely to evolve rapidly as we build increasingly powerful AI agents. As a result, efforts focused on the responsible use of these systems will become even more crucial.

[Yagnesh Revar @_max_entropy]

@pickover Olivia need not apply any force given the text above

[Yagnesh Revar @_max_entropy]

@pickover Btw, what's the task?

[Yagnesh Revar @_max_entropy]

@pickover Is the spider also contributing to the task?

[Yagnesh Revar @_max_entropy]

@yunta_tsai Right, like what does even same kids mean :)

[Yagnesh Revar @_max_entropy]

Say specification was mostly around A, and the generated code achieves something close to A, but what are the guarantees that it won't do other things? Any thoughts on it? Have you found a good prompt or tool for this purpose?

[Yagnesh Revar @_max_entropy]

Those who've been computer programmers in their journey: how has your experience been with the latest AI coding agents? I've been trying out Claude for the past few weeks, and I'm honestly blown away by its borrowed skills in coding a web application.

[Yagnesh Revar @_max_entropy]

It's not impossible to draft and/or generate a large number of test cases, but how can we ensure the generated code is devoid of unspecified, especially undesirable, behaviors?

[Yagnesh Revar @_max_entropy]

@claudeai
What was the rationale for using ">" as a prompt indicator for natural language input? For users familiar with shell environments, it can be confusing and may blur the distinction between the regular terminal and Claude. There are other hints though, thankfully.

[Yagnesh Revar @_max_entropy]

@karpathy I don't get it.