edg5000

14 ms·

361.
by edg5000 1y ago
Why since ~2005? Browsers still function the same, just with more features. I can create a website with some HTML and tables and it will work fine. Many sites are still very simple (such as HN!) and are browsed just like it would be pre-200
362.
by edg5000 1y ago
I'm not saying regulation is bad, but overregulation is. It's counter intutive. See my reaction to arp242 for better articulated arguments :)
363.
by edg5000 1y ago
We all want safe products. Regulating products is a bit like the war on drugs. Drugs are bad, so you criminalize it. Seems logical, right? In reality, we have many case studies where this has actually worsened the situation, as it pushes th
364.
by edg5000 1y ago
With 7 hour tasks it might become worthwhile to invest in a RAM-based local solution with DeepSeek Coder? I've heard that you can run it with 300-700GB. With such long tasks, Claude may run out of usage, right? So queueing it up on a l
365.
by edg5000 1y ago
Maybe it's not the "expert", but "software engineer" part that works? Essentially it's given a role. This constrains it a bit; e.g. it's not going to question the overall plan. Maybe this helps it take a s
366.
by edg5000 1y ago
"(...) kids’ shorts with drawstrings longer than regulation length (...)" We really overrregulate things here in the EU. It was a great run, we've had access to great Chinese stuff for a while now. Maybe around 2007-ish the d
367.
by edg5000 1y ago
Wow! I will try that. Really cool. Never tried the mythical sub-agent feature, not sure if it was really a thing due to the sparse docs. The "You are an expert software engineer" really helps? Probably good idea to mention "s
368.
by edg5000 1y ago
Ah, I see.
369.
by edg5000 1y ago
You are using the "auto-switch back to Sonnet" mode right? Try just selecting Opus without the auto-switch, probably you'll get more Opus and may not run out of it. Anthropic is just being careful because Opus eats compute an
370.
by edg5000 1y ago
This was with Opus? What sort of tasks were you doing? I found that very large files (2.5MB source file) can really eat up tokens. Other than that, I've never run out with my 100 EUR plan, exclusively using Opus.
371.
by edg5000 1y ago
They have fixed capacity but want to provide a reasonable product for all. They've really struggled with defining and enforcing quotas. It seems they have issues predicting what their actual capacity is, how many users they want. Also,
372.
by edg5000 1y ago
I've used CC a lot and to great effect, but it never runs more than 10 mins (Opus). Completely independent for 60 min, sounds impressive. Can you share some insights on this? Really curious; I can also share recents prompts of mine.
373.
by edg5000 1y ago
@rwmj The Eclipse foundation is actually backed by some really big companies. Check out the sponsors page. (And a fun but irrelevant bonus fact: Eclipse was originally made by IBM)
374.
by edg5000 1y ago
Are you exaggerating about "literally hours"? If true, that implies you're making better use of CC than I am. For me the record is maybe 10 mins. Do you write extensive functional/technical instructions? E.g. how large i
375.
by edg5000 1y ago
My only negative thing about AI is that currently it's hard to do it locally, you rely on a third party who now suddenly has access to a large portion of the code. That is really a downgrade from before AI, when I had absolutely everyt
376.
by edg5000 1y ago
It was fun to read though, it flows oddly but it works. Great case study of appropriate AI use!
377.
by edg5000 1y ago
> I also now feel guilty for not providing the same level of detail to other people that I've tasked with something. Relatable! It's because long messages are for uncool nerds. Cool people write short ambiguous messages. But in
378.
by edg5000 1y ago
What do you hate about it? Because LLM companies take away the hardfought control gained by the the individual from their moated castle?
379.
by edg5000 1y ago
Agreed!
380.
by edg5000 1y ago
No, I think the book pirating thing is different than an individual pirating. It's like comparing genocide to murder. A few murders, not great. A few genocides? Really not cool. This book thing at Meta is something we should never forg
381.
by edg5000 1y ago
This is what I believed intuitively, but the reality is more nuanced. I'd rephrase it to: "We can't create something better than humans". A tractor can outplow a human easily just like an LLM can outwork a human on menia
382.
by edg5000 1y ago
Google has been crawling the web. I think it's about limiting to 1Hz per IP. This means a site with 100 000 pages will take 24 hours to crawl. The crawler could send requests as fast as possible, just distributed across IPs. If I put a
383.
by edg5000 1y ago
Maybe a good way to think about this is that humans like usability/automations because even through we can handle more complicated manual operations, "abstracting away" certain operations helps us free up context. E.g. I like
384.
by edg5000 1y ago
Claude Code, where it can use tools and iterate; if it makes mistakes it will know as well and retry, this is a massive boost from copy pasting into chat and getting the trust broken by the LLM confidently making mistakes. By having it be r
385.
by edg5000 1y ago
I've spent many year moving away from relying on third parties and got my own servers, do everything locally and with almost no binary blobs. It has been fun, saved me money and created a more powerful and pleasant IT environment. How
386.
by edg5000 1y ago
If I may speculate the opposite: With cost-effective energy and a plateau in AI development, the per-unit cost of an hour of AI compute will be very low, however, the moat remains massive. So a very large amount of people will only be able
387.
by edg5000 1y ago
If you have absolutely no idea what the code is doing, something is wrong. Imagine someone with no engineering experience telling an engineer to build a jet engine "make it spin fast". Presented with the designs, he has no idea we
388.
by edg5000 1y ago
Mostly agree. What works better is seeing the AI as a "high level to low level converter", in the same way Java is converted to machine code when it's ran. You describe exactly what you want it to report or do, and steer it
389.
by edg5000 1y ago
Treating it like a thinking partner is good. When programming, I don't treat it like a person, but rather a very high programming language. Imagine exactly how you want the code to be. Then find a way to express that unambiguously in n
390.
by edg5000 1y ago
About the data being sent to Google thing; one thing I noticed today with Claude Code it that it ran "cat" on my public key but it ran "head -n 5" on my private key (e.g. it avoids reading my private keys fully) (I was h

More ›