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

The first substack I've actually considered subscribing to. Apparently I am a manchild who enjoys this sorta writing, and few others do. :D

I got 2 500s and quickly popped over to codex since I was just doing documentation and planning work and didn't neeeeeeed claude. He's already back for me. :)

To another's point, yeah, it's sad that Claude taking a brief hiatus just halts workflow. I imagine my whole team just started packing for a beach trip. Then again, I suppose if the CPA office computers go down, everyone doesn't dust off giant ledgers and quill pens and start working manually.


I want to like Claude but I keep having to pop over to codex and I feel at some point I'll stop starting with Claude and just use Codex from the start.

Claude to plan, codex to implement. Claude's giant context is great for reading large amounts of code but it is currently incapable of following instructions/guidelines.

GPT 5.4 is working pretty well for me, both in Copilot and Codex vscode extensions

If you create a plan it follows it closely


I love that this has become my competition in the job market

yall hilarious

neeeeeeeeeeeeeeeeeeeeeeeeed


The big difference is the number of 9s you get from LLM APIs versus your CPA office computers. Building a business around something with 0-2 nines is a nightmare.

That is something hospitals do when the computer systems go down haha

I keep waiting for LLM chats to "steer" to a specific vendor's solution (in exchange for that vendor's substantial fee of course) -- so when I ask "is my UPS repairable?" i might get tips to fix, replace the battery (with $VENDOR's chinesium nonsense perhaps), or straight lied to and told the UPS is now e-waste, but consider $VENDOR's sale on UPS's right now over at this link. Perhaps an affiliate link? who knows!

I pay for LLMs so I hope they don't leak that crassness into paying clientele -- but... how would I know if they did it subtly? I wouldn't! :/


An app that can run queries on all LLM’s at the same time and then figure out which answer is less like an advert might be the solution lol.

A bit like ad block plus.

Or using open source models.


This is my take too, although I'm not prepared for a max400 reality to replace the max200, but... I hate all of the whingeing. Piggies at the buffet line seem to be the loudest on this subject.

Same for me. I have been max200 for like 10 months, i think my usage is reasonable, and i never seem to get the quality throttling that gets complaints daily. I assume they are targeting certain piggies at the Claude buffet, who then squeal the loudest.

As far as I can tell, the people affected are primarily those using their Claude code tokens for openclaw or similar and burning as many tokens as possible


Same here, maybe we're grandfathered into a good plan or something.

We are all quickly becoming allergic to AI writing.

To fool us into thinking writing is not AI generated, we will create "human-ifying" filters to the LLM. This will introduce common keystroke, grammar, and spelling issues that surely no automation would ever create on its own.

Soon the writing most vaunted and trusted will be the writing that appears written by a 4 year old with a crayon.

Sigh.


Glad to see it. We enjoyed the movie greatly but I am sure 50% of that was just relief that the movie adaptation didn’t F the story up :)

The audiobook is the peak experience still I think. Highly highly recommended


In terms of menace potential, any private plane will lose to a van full of fertilizer and a baddie intent on causing destruction. It's a matter of scale.

Little planes, like this one [1] just don't do damage on the same scale as airliners.

[1] https://en.wikipedia.org/wiki/2010_Austin_suicide_attack


Most private planes taking off from commercial airports (the ones where TSA generally operates) are much larger than a Piper Dakota.

(But regardless, it’s not clear that the TSA is even performing that kind of calculus.)


A G650 still loses to a motivated U-haul. :)

No argument though, just saying it's a hard problem, and the scaling issue makes it somewhat awkward to deploy security resources in proportion to the threat.

I don't have a solution. I'm not exactly thrilled with the current setup, but I try to stay quiet since I can't think of anything better.


Government building codes already anticipate the "van full of fertilizer" attack, as a result of the Oklahoma City bombing in 1995. Federal building security is a separate matter though, with its own agency called FPS that predates DHS and TSA by decades.


What about a private plane full of anfo


I daily a M1 Max / 64GB / 2TB and have for some time.

My current company bought me an M4 Pro / 24GB / 512GB.

There is a significant amount of dust on it. I use it every once in awhile when I need to access my corporate Apple ID which is tied to the thing. I'll be really sad if they force me to use it on the daily. Luckily nobody seems to be checking usage.

I know it's not what you asked, but I think it's where you're headed -- an adequately specced Mac instead of two half-spec. :) I can't imagine splitting a workload between two physical machines again -- like when I needed a PC and the M1. It's a real PITA to juggle boxes.


One of my favorite books -- I had no idea there was a real-life inspiration for it (Balzampleu!) This will get me to re-read it, it's been too long. :)


I was aware that Aramis and of course the various royals and aristocrats were real, but not the individual soldiers. Loved this novel growing, seems like the Count of Monte Cristo is seen as more 'serious' literature, but the Three Musketeers will always have a special place in my mind.


> I was aware that Aramis and of course the various royals and aristocrats were real

It's more that their names were real, but their descriptions and their actions in the books are almost entirely fictional.


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

Search: