@sickdotdev Java gets way too much hate as a language.
Python doesn't get enough.
English
Eugene Von Furstenberg
33 posts
Eugene Von Furstenberg
@iusevimBTWW
Crypto investor💸 Manage a fleet of agents thanks to GStack🤖 100% all organic human slop
India Se unió Şubat 2024
4 Siguiendo1 Seguidores
@tgor23 @HSVSphere 4 commands that do the same thing? Just git add and git commit. Or use the editor features if you want to go hunk by hunk.
Having a VCS that only does what you tell it to is objectively simpler than whatever splitting an ethereal commit fuckery JJ has.
English
@iusevimBTWW @HSVSphere i don’t care how some other moron on google feels about it ?? literally everything you do with the index you can just do with a commit in jj and then u don’t have like 4 commands that do the same thing bc they need to address this worthless distinction. It’s objectively simpler
English
Git is just bad. Use jj, and stop complaining about concepts when the implementation is bad.
:3@Colonthreee
Version control like "git" is just creating more work for everyone involved. There is no reason for this to be so complicated, convoluted, opaque, and behaving like it knows best. It should not require the user to know a billion commands and "tricks" to use. It's unscalable.
English
@tgor23 @HSVSphere Its literally so simple. Shove changes into the index, and whatever is in there is in the next commit. So simple.
The first search result for JJ split for me is how the ux is terrible to use.
How the fuck is this an improvement over git?
English
@iusevimBTWW @HSVSphere the index is worthless dogshit compared to jj split / squash
English
@droidbuilds Right now codex, it changes every few weeks.
English
@chribjel I think the opencode team are chill people and have no doubt they're doing their best.
Unfortunately, models keep getting lobotomized, I just can't trust "we don't quantize" anymore. One of their providers must be lying. All the people saying Claude opus got dumber were RIGHT.
English
the opencode go plan might have attracted the worst possible type of customers
pays almost nothing, demands pretty much everything
my condolences to the opencode team
English
@YosarianTwo I agree with the whole it only matters what the CEO thinks thing, but also:
Notice how software quality has kept going down? Software as a field no longer rewards quality, it rewards speed. Vibe coded slop can get you into YC.
That's why LLMs hit software so hard.
English
"AI is not ever going to be intelligent or a threat to us, it's a glorified chatbot, but it's going to steal all our jobs" is a weirdly common opinion. Do people think it's likely LLM's can do so many jobs they cause mass unemployment without being a general intelligence?
ThatOneGuy006@That_One_Guy006
@YosarianTwo They aren't retaliating because of some imagined apocalypse scenario with your glorified speak and spell, they're attacking because you want them to starve.
English
@ThePrimeagen I've gone through a similar thing, slowly coming around to it, and I completely agree you can't break 20% improvement without slopifying.
English
Its crazy how many people think slowly changing your mind and testing your theories is bad or makes you dumb.
ThePrimeagen@ThePrimeagen
I am slowly coming around to AI assisted programming. I am genuinely trying to codify every rule about programming that I have and using that + several stages to build out small changes. Not sure the productivity changes, but I think I can see a modest gain in speed. I am also trying to be concerned about every line produced, not just slop trebucheting code over the wall.
English
@usr_bin_roygbiv @samswoora Your mom fell down stairs holding you when you were a toddler
English
There’s no explaining why AI is good when someone loses their livelihood forever. People will love AI when cancer cures start coming but that’s years away even if we reform the FDA.
This isn’t a messaging problem, normies aren’t stupid - they see what the labs are doing
Joshua Achiam@jachiam0
The AI industry has largely failed to foresee the extremely foreseeable: most normal people think AI is scary because we had decades of science fiction about how scary AI is. We owe the public clear explanations of why we think this can be good and how to make it so.
English
@RaviTejaKNTS @Prathkum They are:
Integrating Gemini into their wider ecosystem.
Applying transformer architecture outside of language (like alphafold)
Releasing smaller LLMs that pack a punch
Creating custom hardware for ai use.
Winning the war, not the battle.
English
@iusevimBTWW @Prathkum So basically they are losing in this? This post is about AI. How come they are amazing in other things is a valid answer?
English
@LundukeJournal Worst take from you man. They have TECHNICAL reasons to do so.
Linux is meant to be a general purpose OS. i486 is so old you should and frankly have time to move over to a different OS. Like a BSD.
English
Support for "older" computers is critical for the future of computing.
Linux is dropping 486 support... and that is a massive mistake.
English
@vivoplt They're too focus on doing actual good shit, especially their flagship deepmind.
You got turboquant, alphafold, gemma4, TPUs, I'm convinced the only reason Gemini sucks is because google sees no future in the largest LLM race and focuses on other aspects instead.
English
@vaxryy I would not like to see desktop OSs become like phones. I absolutely do not want the manufacturer to determine my OS.
That's not to say we can't learn from their security, but we can do better than the current situation on phones.
English
unpopular opinion
even without grapheneos mobile os'es (android and ios) are the most secure os'es out there.
Yes, without grapheneos apple/google are definitely sniffing on you but from an attacker's perspective any desktop os is much, much, much less secure
HSVSphere@HSVSphere
I like how people who have NO idea about security telling you to "stop using your phone" Brother, your phone may be the most secure device you have on you. Especially if it's @GrapheneOS
English
@jake_researcher @BenjDicken It's 27 years old, so that area of code likely hasn't been touched. It's in the TCP stack too, not exactly a fast evolving part of the code.
English
Mythos: finds 27 year old kernel bug
Also Mythos: writes a bad SQL query
English
English
Hi! Big Anthropic hater here. Fundamentally disagree.
Anthropic is spending way more than they are getting here. It’s causing a compute crisis for them that’s destroying their reputation in record time. The cost to train the thing is probably higher than the revenue will ever be, especially considering opportunity cost.
The security concerns are absolutely real here. Getting a 27-year old 0-day in OpenBSD by running a for loop is not a joke. We’re a free model iterations away from every piece of software we rely on being compromised.
I respect Anthropic for eating the cost here and giving the software world a few weeks to get ahead of the inevitable crisis looming overhead.
English
The whole Mythos cyber security story is likely just a psyop to have an excuse to not serve frontier models to the public.
Reasoning:
1) Other labs can't distill it.
It's annoying when you have a dominant SOTA model, and 2 months later Chinese labs sell the same SOTA model for 1/50th of the cost
2) Compute constraints- so you have to choose between enterprise and vibe coders.
Enterprise have like 1% monthly churn.
Vibe coders cry and threaten to have their mommy buy them a mac mini for local models whenever their rate limits are cut.
3) Big enterprise pay a hefty premium for slightly better performance and corporate polish.
Dario wears suits and is reliable.
Sam is a compulsive liar.
Chinese models are politically a no-no.
xAI seems to focus more on hentai-gen.
Google seems to focus more on defending their ads moat + consumer distribution.
Anthropic has a quasi-monopoly on enterprise and can charge quasi-monopoly prices
(which explains their ARR)
They HATE you. They don't want anything to do with you.
But they can't say it, so they of course have to be like: "Mythos is a very bad h4ck0r! We can't give you Mythos cuz u gonna get h4cked!"
Reboting@Reboting_HFT
@scaling01 Not really, I think the moment Mythos is released to the general public it'll be distilled by China. We will have a open-weight Mythos level model by the end of the year imo. I think a bigger issue is who the hell is going to be able to run the damn thing.
English
@peterwildeford Not even anthropic, the biggest exxaggeraters of all time, managed to spin that they hacked openbsd like they "hacked" freebsd, best they could do is a DoS.
English
"Mythos found a 27-year-old vulnerability in OpenBSD—which has a reputation as one of the most security-hardened operating systems in the world and is used to run firewalls [...] The vulnerability allowed an attacker to remotely crash any machine running the operating system"
Anthropic@AnthropicAI
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing
English
@kenwheeler The reason google isn't winning the LLM race is because they're too busy making the world a better place (eg. Alphafold, gemma4)
English
ok then cure cancer you fucking dorks
unusual_whales@unusual_whales
BREAKING: Anthropic's new model, Claude Mythos, is so powerful that it is not releasing it to the public, per NYT, Instead, it is starting a 40-company coalition, Project Glasswing, to allow cybersecurity defenders a head start in locking down critical software.
English
