darkstarsys

Or maybe you did it because you had ideas for things computers ought to be able to do, and needed to get those ideas into the computer?

This is an excellent essay, and especially relevant when read together with the MIT report on the future of education in an AI era: https://aiandeducation.mit.edu/report/ (and my own https://oberbrunner.com/blog/ai-future-of-mathematics).

I'm heartened to see this kind of thoughtful, reasoned discussion around long-needed reforms in the educational community, with AI as the catalyst. We certainly have interesting times ahead.

Yes. I find it really interesting to consider what the machines do and will think of as intrinsically interesting to them. Will they develop their own theories of beauty, mathematical and otherwise?

Brynjolfsson: an AI “job apocalypse” is unlikely.

Also Brynjolfsson: It will likely impact entry-level jobs. There are huge implications for business and the economy, including some increases in living standards, but also, potentially, job disruption. We must act now to address this gap. We can’t just sit back and wait for the tsunami to hit us.

Also Brynjolfsson: If companies don’t hire people at the base of the pyramid, then they’re not going to have those people later when they need them. Now, AI can do a lot of those basic tasks. Some companies, such as law firms and investment banks, are not hiring as many people at the base. In the short run, I can see that logic, but, in a few years, where are they going to have the most experienced people? So they need to think harder about that.

So there's that. Basically what he's hoping is if everyone does what he recommends re: regulation and policy, then maybe the job apocalypse will become unlikely.

Yeah, came here to see if anyone still remembers lmgtfy. Back to the future, I guess!

Sorry, this article's argument just doesn't hold water. Yes, we want Claude to write "the best text" and having its word choice even very slightly varied could arguably be construed as "not the best." But "best" is highly subjective, always has been. Claude has never, and will never, write what _you_ consider the best version of a piece of text. It has many choices, influenced by all kinds of random, context-dependent weights and environmental settings. Slightly tweaking weights to prefer certain phrasings might even tilt it toward your idea of "best."

If I'm any example, I'm on the other side. I got into computers in the '70s: plotter art, BASIC, then Z-80s, forth, PDP11s, Amigas and so to now with GPUs and supercomputers and visualization. I'm recently retired and having enormous fun plowing through my lifetime backlog of "wouldn't that be a cool tool" (https://pcons.org) or a climate visualization (https://globe-viz.oberbrunner.com) or a travel packing list app (packzen) or a bunch of 3d printer tools... so on and so on. Totally co-engineered with Claude. What we're doing together, and getting out to the world as useful open source projects, is easily 10x what I could've done on my own, in the limited time available to me in this life.

If you think LLMs can't "jump," watch Terence Tao engage with ChatGPT to try to understand the intellectual leaps achieved by Claude Fable 5 in finding a counterexample to the Jacobian conjecture https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed... . After that you may revise your opinion.

Yes, emphatically! I wouldn't be where I am today without tools like Emacs and Linux. This is why I created [pcons](https://pcons.org) because software build tools (like all dev tools) need to be open, transparent, hackable and easy to understand.

Exactly the opposite for me. I've gotten into a whirlwind of productivity, working through my lifetime backlog of "somebody out to do thing X" — now I'm doing ALL of them (all open source naturally), plus parametric 3d printing, hardware projects and more.

The enshittification begins.

Exactly the same thing happened when Google appeared on the scene. (Not that this makes it any better!)

Indeed. Perhaps my article here will be of interest to some: https://blog.oberbrunner.com/blog/ai-math-as-humanities/

See also my log-scale timeline of the universe — hand-curated rather than a giant import, and I hope also a bit nicer/simpler UI: https://deep-timeline.org

Owners of capital, yes, but also owners of the means of production (which now means AI companies). See https://blog.oberbrunner.com/blog/ai-risks-taxonomy/#economi...

I'm on my third Surface Pro (an 11 this time) and basically love it. Tablet form factor, lightweight, multitouch, with a real desktop OS (I mean not ipad's iOS). Decent battery life. True, it's not perfect: slow to wake, the touchpad stops working once in a while & needs to be reset. Missing a GPS chip. But it runs Adobe, Resolve, Chrome of course, msys2, and Linux (WSL2) quite well. I love the absolutely gorgeous HiDPI screen. The software emulator system is a little weird (arm64/arm64EC/arm64X, with no true universal x64/arm64 binary) for software developers, but it basically works fine from a user perspective. I say all this as someone whose daily driver is an Arm M1 Macbook Pro, also nice but not a tablet and quite heavier. I don't use a dock, just a simple USBC hub with a magnetic USBC connector.

What happens to humanity when AIs are better at being human than most humans? (More patient, more empathetic even if it's simulated empathy, more knowledgeable)

I spend a lot of time doing this too -- cutting up band rehearsal tapes into songs and exporting all. As weird as Audacity's UI is, I haven't found anything better than it at this.

Absolutely gorgeous as an art piece.

I'm newly mostly-retired as a VFX software developer & CTO. I'm writing about AI, climate change and more in my blog, https://oberbrunner.com, running Long Now Boston (https://longnowboston.org) to promote long-term thinking, and working through my lifetime backlog of "wouldn't it be great if somebody wrote this" ideas using Claude, at https://github.com/garyo.

You should check out my new open source software build tool, https://pcons.org.

I'm mostly retired from a lifetime as a graphics programmer and CTO, and now I'm working through my lifetime of fun backlog projects. https://pcons.org, https://deep-timeline.org, https://pelorus-nav.com, https://packzen.org, https://github.com/garyo/sea-surface-temp-viz, https://globe-viz.oberbrunner.com/ and lots more. All open source and free.

Sure, make money from software. I did. But when you have enough and it's time to give back, open source it.

As a heavy AI-assisted open source code creator (and someone with 40+ years of dev experience), this seems wrong-headed to me. I think it is an excellent policy, as they say, to "value contributors over their contributions," but this policy excludes all potential contributors who use the latest tools. It will eventually doom zig to a smaller "artisanal" pool of contributors, rather than welcoming newbies and helping them become better open-source developers.

But what about copyright? Asking for a friend.

I hate CMake as much as anyone, but I actually did something about it: https://github.com/DarkStarSystems/pcons. As one of the original developers of SCons, I tried to get the best parts of SCons and CMake into a simple, modern, python-based software build tool. I take toolsmithing seriously, so I plan to build pcons into a strong, useful open-source project.

I find correction is rarely necessary with Opus 4.6. Definitely not so much that "it's not really the LLM writing it anymore." More like it's the author and I'm the editor (in this limited case -- of course architecturally the ideas are all mine.) But I totally respect that my prompt style, the type of app I'm writing, and other factors could be influencing my success vs. others' lack of success.

I can't agree here. https://pelorus-nav.com/ (one of my side projects) is 95-98% written by Claude Opus 4.6, all in very nice typescript which I carefully review and correct, and use good prompting and context hygiene to ensure it doesn't take shortcuts. It's taken a month or so but so worth it. And my packing list app packzen.org is also pretty decent typescript all through.

This post is excellent, and accurately describes my experience writing pcons (pcons.org) as a side project. I was one of the original developers of SCons and have wanted to rebuild it better for more than a decade. All the same roadblocks Maganti describes kept me from starting it, and Claude Opus 4.6 suddenly opened the door, and now it's live and people are starting to use it as a cmake or scons replacement. My experience over the last few months mirror Maganti's in many ways: ease of refactoring, investigating many more design ideas, getting frustrated with blind alleys and its not understanding the big picture, and ultimately getting a product I'm proud of.

Vision, taste and good judgment are going to be the key skills for software developers from now on.

I agree with this article, and I think it goes further than astrophysics or even physics. As agentic LLMs are starting to prove long-standing open mathematical conjectures and even invent new ones, I fear we may reach a point in mathematical research (which, as Hogg's article describes astrophysics, has no "right edge") where the machines are just better than us. At that point do we just mostly lose interest? You can see it already in the Go-playing community. Why study for years to be a "pretty good" 9P when machines will always play better? What PhD student will spend years on a hard, interesting, deep math problem? Sure, a few will. Some people become monks too. But not many.

Yes, and this is an incredibly powerful idea. Running fluid flow simulations inside an optimization loop (monte carlo + gradient descent) revolutionized aircraft design, nuclear simulations and geophysics. When the tool being called updates the LLM's training data, or runs experiments that the LLM can learn from, then that potentially becomes a self-improvement loop.

straight and elpaca etc. are just as vulnerable. Maybe more so.