Hacker Newsnew | past | comments | ask | show | jobs | submit | wiz21c's commentslogin

Good news: we will leave oil

Bad news: according to the discussions here on HN it appears that there is no consensus on what the good mix of renewable/nuclear is. Therefore us, citizens, will be manipulated by politics.


The manipulation risk is real, but it usually comes from pretending there is a painless answer

> UK banks will get access to Mythos next week.

Makes me wonder what the military have access to since months...


I do greenfield in fluid dynamics and Claude doesn't help: I need to be able to justify each line of my code (the physics part) and using Claude doesn't help.

On the UI side Claude helps a lot. So for me I'd say I have a 25% productivity increment. I work like this: I put the main architecture of the code in place by hand, to get a "feel" for it. Once that is done, I ask Claude to make incremental changes, review them. Very often, Claude does an OK job.

What I have hard times with is to have Claude automatically understand my class architectures: more often than not it tries to guess information about objects in the app by querying the GUI instead of the data model. Odd.


My observation is so far, LLMs are not good at scientific computing.

The job of the programmer/Designer will be to answer questions about what the program can do/not do and tweak it outside of Claude's abilities. To be able to answer those questions (like: can we do this ? will it fail under pressure ? etc) requires a deep understanding of the programs which you only have if you actually build them (with or without AI).

So, less jobs for sure, but not like 50% less jobs.


the good news I guess are

1/ lean-zip is open source so it's much easier to have more Claude's eyes looking at it

2/ I don't think Claude could prove anything substantial about the zip algorithm. That's what lean is for. On the other side, lean could not prove much about what's around the zip algorithm but Claude can be useful there.

So in the end lean-zip is now stronger!


Tested egui and it's great provided you accept the "immediate mode" and its limitations. It's quite polished nowadays although it misses in some areas.


I'm sure that if a car appeared from nowhere in the middle of your living room, you would not be prepared at all.

So my question is: when is there enough training data that you can handle 99.99% of the world ?


within 20 years, everyone will be developing software which will be copyrighted partly by AI and be behind walled gardens. Sure you'll be able to do things locally but everything (security clearance, walled garden, government's control etc) but it will forever remain at the level of "tinkering".

If you are 50 years old or more, the computing you were born with (you own the computer, you own the programs) will be gone. Copyleft only makes sense if you own the computer.

That makes me sad.


they are activists, everything make them nervous. however I'm sure there are tons of past reasons to make them nervous...


don't forget the size of the search space...


this is why big tech is spending 500B on GPUs


that they don't even have the datacenters to plug them in, not the power generation needed to run them if they did


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: