Anthony
1.5K posts
You’re not supposed to be sore all the time… lol
🍝@butlerspasta
No one talks about how staying consistent at the gym means living with sore muscles 24/7.
English
@TylerJnstn many current jobs will go away
i think we will find a lot of new ones, though they may look very different
English
I honestly miss the Sam Altman that used to call out his peers for downplaying this risk.
Sam Altman@sama
i think a lot of people are going to be busier (and hopefully more fulfilled) than ever, and jobs doomerism is likely long-term wrong. though of course there will be disruption/significant transition as we switch to new jobs, the jobs of the future may look v different, etc.
English
@Marco777Polo @TheFizzyMind @levelsio Yeah, could be, I have one on every floor they fluctuate in the low 400s. If me and the kids/wife/dog are all in a smaller common area I will see it rise to around 550 after a movie etc. but quickly falls back.
English
@Antmage @TheFizzyMind @levelsio Oh crazy low! Woods sounds like a dream!
Global is 431ppm (Hawaii). I never saw any value (outside) in a city below 470ppm on a calibrated device. City still has CO2 emissions sources...
Maybe your very lucky. Devices could also be off like ±100ppm (which is little!).
English
I still haven't solved the CO2 bedroom challenge
You open the window and you wake up from a 6am garbage truck or barking dogs and sunlight
You close it, you suffocate in 1200 ppl at 5am
I guess you really need some mini tube in your wall with a vent that opens and closed based on internal CO2 but how do I build that?
English
@Marco777Polo @TheFizzyMind @levelsio Mine stays around 400~ open bedroom doors and plenty of ventilation. Also living in the woods might help.
English
@TheFizzyMind @levelsio If you leave it will decrease back to 400ppm-ish.
But impossible to keep while you are there.
Otherwise you need active systems. And you get other downsides like super dry air in winter. With active moisturizer you risk mold - it gets complicated and a professional level.
English
@planefag the premise doesn't even hold, even deeply buried libraries ultimate exist to serve an end-user somewhere, somehow, eventually even if that user is the same company writing the software.
English
This is the exact kind of guy I'm talking about btw. I'm arguing that Github could stand to put a few hyperlinks in more visible places on the page and he's acting like I'm demanding the fucking firstborn of every FOSS coder in existence.
Theater kids
Anon-kun@Anonkun80564929
@planefag This just kinda reads like an entitled rant. The world does not revolve around you, and if you want to get into FOSS, then you are going to need to learn how things are done in that world. I did it when I was 7. Also, lmao cygwin. Stop using windows for unix-made stuff dingus.
English
@theodorvaryag @effectfully The bureaucracy still takes longer than project though.
English
@effectfully it's not just the turn-around time for smaller stuff, I can take on insane milestones and deliver them in less time with much tighter perf and correctness bounds.
English
I've been developing a massive AI psychosis.
"Hey GPT, I need a breadth-first level-aware tree generator with a clean API that I can use to specify skewness"
"Sure thing, I stitched together the obscure applicative you suggested, the free applicative for layering and the generator applicative, added a metric for skewness and ensured the results match the expectations"
I wouldn't fucking know how to do that myself without spending a whole evening on it. Bastard did it in minutes.
English
💯 Yep mine had assembled an architecture group for that purpose, and it does look like it has potential. I received approval to rewrite a project late last year, and part of it included embedding the rails and invariants that force responsible AI usage.
It took about 2 months before it supplanted the original 8 figure/7 year application in both form and function.
The bureaucracy of approvals required will take longer than the applications creation.
English
Yes. Me too. Except 25 years. You’re exactly right that tech stack issues don’t go away with AI.
We have a unique opportunity right now to build new stacks. “Innovation” and “continuous improvement” teams are desperate to implement AI at all costs, and that cost can include building entire separate AI and data ingestion infrastructure to create datasets AI can run against.
I’m getting 7-8 figure buildout costs approved for the first time since 2007, because companies are desperate not to be left behind in the AI arms race.
English
I'm 22 years old and Claude Code is deteriorating my brain.
Every single day for the last 6 months I've had 6 to 8 Claude Code terminals open, waiting for a response just so I can hit 'enter' 75% of the time. And it's doing something to me.
In convos with a couple of friends, it's been a point that's been brought up pretty frequently.
None of us feel as sharp as we used to.
I don't know if it's just us, or others in their 20s are feeling the same thing, but it's something I've been thinking about a lot.
P.S. I know this is a problem with my reliability/usage of it, not Claude Code itself, but the effects are real nonetheless
English
Maybe, but I am in the same decade, in this industry for 20 years, EoD most places never cared about the code, just the rate in which features can be rolled and the associated maintaince/support cost balanced with liability.
The systems level problems historically were the hardest to solve, tech stacks chosen just because of the market or in house familiarity, 9.9/10 we all knew where and what the issues are and will be in the future, it was always about balancing time and ambitions.
Now the time tradeoff on the software side has shifted to where more ambitious projects can be done by smaller teams, it will be interesting to see where the industry is in 10 years for sure.
English
@Dgkalx1267 @astnkennedy He’s making shit up. This guy is a 40 year old from Bangladesh.
English
@Matthew02388332 @niccruzpatane Yeah, I tried 3 times to use it, all three were light rain, 2 it got stuck and I had to get wet to get it anyway, the one though, magic.
English
@Antmage @niccruzpatane yeah maybe depending on your layout. i was taking more about ASS. super cool to see an unmanned car drive through the grocery store parking lot but too slow / nerve wracking to be practical.
English
According to Ashok, the new V14 Lite for HW3 will include pretty much all the features as HW4 V14 (start from park, presumably arrival options, etc). It’s expected to release at the end of June.
That will improve the HW3 experience drastically. I’m looking forward to doing a head to head between both HW3/HW4 V14 when it rolls out.
English
@Matthew02388332 @niccruzpatane Summon is super useful for not having to go to the garage to get the car out…
English
@niccruzpatane start from park / auto park feels like a party trick (like summon), vs a real quality of life improvement. FSD is a game changer for my long commute … i really don’t mind pulling out into the street before activating it.
English
@theodorvaryag @aramh 🎯 I said something similar at the advent of the models becoming useful; the top end will disproportionately benefit for a span
English
@aramh I'm happy enough with how they enable me to learn faster. The code I can actually keep is gravy.
English
If you use LLMs to do work on your behalf (i.e. you know what needs to be done), these don't apply to you. But if you use LLMs to figure out the best way to solve a problem, they might.
Yes, of course that 15kLOC one shotted in a single 12h session is never going to be good enough to ship (but if you use a theorem prover it will be correct!). But it might be good enough to understand what works and what doesn't so you can have a more informed second or third approach.
Aram Hăvărneanu@aramh
All folklore about LLM best practices is wrong. I have found that: - the longer the task, the better it performs. Up to a point, sure, but that point is ~12-24h of work). - longer threads (with multiple compactions, etc) improve performance. Models don't even get good until they explore millions of tokens worth of misunderstanding. I avoid starting a new thread as much as possible. - precise instructions kill competence. models need space to think. - plan mode is absolutely terrible. - optimal request size is 600-1000 words. - MCP tools degrade performance a lot. Some things are only accessible through MCP, tough luck then, but for things that accesssible through other low tech means, prefer that to MCP. YMMV, but for research type of work that is at the edge of feasibility and understanding for both people and machines, I stand by what I said.
English
@elonmusk @MikeBenzCyber A lot of improvements needed to claim “Accurate”, better, sure.
English
@MikeBenzCyber The private sector accurately values labor
English
For reference, the head of the CIA only makes $230K / year.
Alec MacGillis@AlecMacGillis
"Sheryl Cowan, 57, was making $272,000 a year as a senior VP at a U.S.A.I.D.-funded nonprofit when she was let go at the end of March 2025. Last month she had an online interview for a $19-an-hour job managing a Penzeys Spices store in Falls Church, Va." nytimes.com/2026/04/21/us/…
English