[Space Bites+] Weird New Planet // New Moon in A Bag // Superstorm Hit Mars

Without closed captions, I am not following you..

I've always found this argument very strange: why do you use Claude Code, just learn how to code? Why do you use Midjourney, just learn how to paint! Why do you use Suno, just learn how to make music? Like, don't get me wrong, there's a million good reasons not to use these tools. But the fact that given unlimited time you could learn to do everything yourself is not one of them.

Sylvester Stallone's image was used for the Rocky planet. Who's images was used for the gassy planet?

I know this is not your main thing, but given how useful you're finding AI coding tools, and given that you are a good explainer and basically a computer-literate non-dev, it would be amazing if you'd consider doing a stream showing how *you* use AI tools to create your tools.

Fraser, I've enjoyed your reasonable and nuanced thoughts for years, and I am onboard 99% of the time, but I don't think you're in the same ballpark of thoughtfulness that you set for all your other content when you declare that unauthorized training on software code is somehow free of ethical concerns that apply to other intellectual property. Basing this on code "seeming like math" doesn't seem well-thought-out. Perhaps it would help to think about all the different things code can accomplish, beyond utilitarian tools. Or, consider that open source software that was donated to humanity in good faith can now be exploited for commercial gain, or at the very least stripped of attribution, without the option for any vibe coders to rectify this. Can we stop the change to LLM coding? No. But, should all intellectual property that is popularly associated with analytical thinking be dismissed as not worthy of some ethical considerations when it comes to fair use?

Hey Fraser: Just saw Project Hail Mary in theater! You will love it, because it follows the storyline of the book almost exactly. And Ryan Thomas Gosling was excellent. Fun Movie,

"Learning how to code" has changed dramatically throughout the history of computer programming: - First you had to learn how to punch the right holes in cards to drive a series of mechanisms to add, subtract, multiply and divide series of data to "compute" the outcome you needed - The first paradigm shift was to type machine-code into a text-editor to tell a CPU how to "compute" instead of punching the holes (CPUs were also more flexible in what they could do). You could still punch holes in cards if you wanted to, but a more productive mechanism now existed. - The next paradigm shift was to write code in a higher language (BASIC, Pascal, Fortran, etc) that was compiled into machine-code. This really paved the path to abstracting the inner workings of the machine and focus on describing the problem. You could still write machine-code or even punch holes in cards if you wanted to.... - Several "paradigm shifts" happened in software, more and more abstracting the underlying machine and offering more and more productive ways to describe the problem to solve, allowing for solving more and more complex problems effectively. - The latest paradigm shift is you can now describe the problem in a human language to describe the problem to solve - that is what "learning to code" means in 2026!