dreamglider

It's not that better than Opus 5.8, IMHO (highly biased since personal) - no more than 10% positive difference for the same prompt, and that 10% is not consistent, sometimes you get obviously better response / nicely done task, sometime you don't.

Thank you. Love the post. Best of all it does not try to state the obvious but just share how one feels about the cr@p hype which is soooo tiresome - not just about the LLMs, it was pretty much the same for mobile apps, big data, lab on a chip, quantum computing, fusion (ma favourite), monkey in a jar & etc. However the sad truth is that - hype will continue, seems like a default feature irrespective of context and domain, more hype = more $, more visibility, more reach. Probably the only good thing is the cray talk lead by Sam Altman, weirdest though - so many zombies believe it but again - expected, will have the same faith as the EOW according to the Mayan calendar... lol

That's actually something pretty useful for businesses that relay on *NIX. Why did you do it? Did you have to manage a lot of Linux devices with no easy way to set policies? How did you come with the feature list? Any option to replace the Internal CA with any of one's choosing? Any alerts (webhooks)? In either case - gonna play with it, seems cool. Thanks for sharing!

For the better part of the last 20+ years big corpos had the $ to throw and replicate virtually anything they wanted. They got the cash and man power, yet they didn't do it. Why? Because they don't care, they have business to run, they need to somewhat keep focus which can't happen when scattering attention all over the place. The difference now is that for relatively >simple< projects (4-6 months work of a team of 5-6 ppl) one can do it faster using LLMs. Basically - I can get faster to a place I could always go to but didn't (and still don't) want to.

One seems to omit the fact that LLMs are fundamentally designd for workload quite different for what they are being used right now. Sure you can improve them but can't escape / workaround the current NLP design endlessly. Then there's the irony - Internet did deliver on free (as close as it gets) and easy access to information (any). Did this make people smarter, more knowledgable, more tech savvy & etc? Nope, it didn't. Just like the libraries didn't (queues at libraries were and are a rare event). Big deal that the information is readily available when people do not know what do with it or care to do anything.

Ideas are cheap, the chances of having some truly unique idea that is also business feasable are not that big. It's not so much about the ideas but rather the ability to execute, follow through and well - make sales while constantly improving what you got. Staying silent, going dark - have their merit but only when the wheels are already turning and one is into acting, not into fearful hiding.

In either case - awesome blog post!

Seems it is not just about geeks VS general population but rather the classical case where success (ease of access) & popularity inevitably bring failure. Few open (loosely or not actively moderated) "spaces" left. No wonder given the general attitude of the, lol, "invaders". :D In the past - you pick an IRC server and a room and 4 out of 5 times you'll learn something interesting, have actual fun with ppl you don't know and just enjoy the interactions. Now similar experience can happen only in closed/invite only or hard-to-find groups. The mainstream ones (different "social media" services) seem filled with people who want only to show off while remaining as alienated (and as consequence hostile) from one another as possible. Good or bad - the old Net is dead. The new one is predominantly for making money and BS. The same trend can be observed virtually anywhere. In the past - people experimented with games, lots of cr@p titles but also pure gems, games that last. Today? AAA titles that repeat the "successful" pattern, over and over again. Anyway - it is what it is, the good thing is there are still meaningful places and people worth reading/listening to, just way harder to find through the noise. News.y seems one of the few remaining and open islands.

To bring this further - it's like the migration from villages to cities and towns - proliferation of alienation, loneliness, broken communities, fake smiles and treating anyone not part of your close circle as potentially hostile psycho ready to steal your kidneys, sneeze in your coffee or /dev/null ya. Anyway, no more laments for the past given the current situation presents interesting problems that nobody has solved-solved yet, perhaps because they won't make you a billionaire lol.

The few thousand of people (worldwide), who were into NLP (Natural Language Processing) and were the drive behind the ML "revolution" - they care, but for other reasons. In like 10 years from now, when you ask for a hotdog but get a sandwich - then you may care too. But hey - that's happening right now, with humans, anyway ! :D

Stating that there are "acts of intelligence" is not even wrong. Sorry, not an opinion, a fact. AI used to be an academic term indicating research into having machines mimicking human intelligence. Machine Learning (a.k.a the current hype) is about pattern recognition and has precisely zero to do with AI and/or any form of intelligence whatsoever. A rat has more intelligence than an LLM and bees - for sure. It's not about pretending - it's about facts. This statement is true if we have a shared understanding of what 'intelligence' means.

Nothing extraordinary, just interesting. You know - to advance / provoke one's thinking.

Do you write code? Yes. What code, examples? I code HELM charts! :)

Awesome particle.