Those Python issues are things I had to deal with earlier last year with Claude Sonnet 3.7, 4.0, and to a lesser extent Opus 4.0 when it was available in Claude Code.
In the Python projects I've been using Opus 4.5 with, it hasn't been showing those issues as often, but then again the projects are throwaway and I cared more about the output than the code itself.
The nice thing about these agentic tools is that if you setup feedback loops for them, they tend to fix issues that are brought up. So much of what you bring up can be caught by linting.
The biggest unlock for me with these tools is not letting the context get bloated, not using compaction, and focusing on small chunks of work and clearing the context before working on something else.
Having read the post, I feel the title here doesn't do it justice. It is also missing half of the title.
Title on HN at time of comment: "Silicon Valley Could Learn a Lot from Skater Culture"
As far as the article, I wasn't aware of the schism in the 80s around skateboarding culture. I was too young, but even looking back through documentaries and other media about the history, it seems to have been revised to ignore it.
I think the article is overall positive, in that there are things to learn so long as we don't ignore the context and the bad side of what has happened in the past.
In the case of the JS solution, it runs horribly and is highly inefficient.[0]
For non-hot code, it is perfectly fine, but in this case, inside of a main loop like this it is a waste of cpu cyles. Compare the amount of forests here:
$ node orig-magicForest.js 117 155 106
total forests: 1522899
{ goats: 0, wolves: 0, lions: 223 }
total time: 816ms
$ node new-magicForest.js 117 155 106
total forests: 428
{ goats: 0, wolves: 0, lions: 223 }
total time: 6ms
I think there is a disconnect between what an artist might expect to receive and what a listener would actually pay.
Even before streaming radio became the standard way of listening to music, people didn't pay every time they played a song. They paid for the rights to listen to it as much as they wanted, but they only paid that once.
Now that we have streaming, consumption is different, and artists actually get paid per listen. They can't expect to get the same amount per listen that they got per song on a CD years ago.
You really shouldn't get into music for the money, because there isn't money anymore. The fact that there was money ever was a blip on the radar due to new technology allowing an industry to explode. Technology continued on though, and made the scarcity of music artificial.
This blog post is way off. The issue isn't that Google Maps couldn't find Sambisa Forest, it is that it has an issue with how it handles auto complete and pressing enter without selecting the choices it brings up. It quite frankly looks like a bug. You can see for yourself here: http://i.imgur.com/vX1xqp2.gif
I disagree that the blog post is way off and I think the issue is that google can't find Sambisa Forest (it can't - as your gif demonstrates). It can find, "Sambisa Forest Reserve," which is great, but not the point of the blog.
They are saying that google's ability to take natural language requests (sambisa forest, in-n-out east bay) and come up with an answer that corresponds what you searched for has seriously degraded. Personally, that is consistent with my experience.
What is causing the problem (bug, difference in focus, etc) is also worth discussing. Why does google have so much trouble finding answers to queries that seem straightforward? Why can it suggest "Sambisa Forest Reserve" (based on "sambisa forest") and then have no idea how to find "sambia forest?" The title might be a little grandiose, but I think it's a fair question.
What if they already get a placement fee? That would mean they make money on the front and backend, which if they dropped their tuition fees, would drop their income.
In the Python projects I've been using Opus 4.5 with, it hasn't been showing those issues as often, but then again the projects are throwaway and I cared more about the output than the code itself.
The nice thing about these agentic tools is that if you setup feedback loops for them, they tend to fix issues that are brought up. So much of what you bring up can be caught by linting.
The biggest unlock for me with these tools is not letting the context get bloated, not using compaction, and focusing on small chunks of work and clearing the context before working on something else.
reply