https://t.co/RgVs4mn23m
You have no idea how bad his code is. This guy is the most pretentious likes the smell of his own farts larping as a genius douchebag on twitter.
He actually thinks he's smart because he asks LLMs to write code using fancy algorithms that don't even make sense to use where they are.
He's a fucking dipshit retard and I'm embarrassed for anyone who thinks highly of him.
"not that you'd understand them" neither do you!
I ran some analysis on this Asupersync trash with an agent swarm (got through maybe 2% of the 1 million+ lines of trash)
your own repo has files committed in it that say it's trash btw lol (analysis from your own agents)
the big one is MODES_OF_REASONING_REPORT_AND_ANALYSIS_OF_PROJECT.md
it says asupersync has "metastasized" into something one person plus agents cannot maintain, and that ~30-40k lines of mathematical machinery have zero callers from any operational code. your own agents ran the grep.
what you built is what happens when you keep prompting an LLM for fancy-sounding theorems without understanding any of it or whether it belongs. persistent homology, sheaf theory, Dialectica, martingale drain certificates, spectral monitors, conformal calibration.
none of it is wired into anything.
it exists because you asked an agent "what's the most rigorous way to bound X" and pasted the answer. multiplied by every subsystem. the result is unusable trash. nobody who actually writes Rust would touch this with a five-foot pole, because the alternatives already exist: tokio, hyper, h2, quinn, sqlx, tonic, rustls. all of them years old. all of them deployed at planetary scale by AWS Firecracker, Cloudflare, Discord. all of them maintained by hundreds of people with security response teams. tokio has had `tx.reserve(); permit.send(v)` (your "novel" two-phase channel) since 2021.
you, one human, in 100 days, on nightly-only Rust, with "I do not accept outside contributions," with 6,211 .unwrap(), 1,041 panic!(), and 210 unreachable!() calls in your library code , roughly 7,500 places where the program just kills itself, in a language whose entire discipline is that library code does not panic, it returns errors and lets the caller decide, claim to replace all of them.
your README markets this for "financial, medical, infrastructure."
when someone asked you for paying users of your products, you skipped past 1.1 million lines of supposedly world-historic engineering, and pointed at a markdown-file subscription service lol.
It's not even about all the people you're fooling, it's about how much you're fooling yourself with this crap if you truly believe it. This shit is UNUSABLE GARBAGE.
And you point to it as proof that you're smart or something? (not that we'd understand your genius). I don't think you're an idiot necessarily, I think you've got very little self awareness.
The fact that you've spent (and keep spending) money on this is absolutely insane. No rational software developer would ever EVER use this.
@Hesamation@doodlestein Building the most disgusting ridiculous slop code you've ever seen.
Like bad beyond belief and anyone who actually checks ends up laughing and people still follow this clown.
@mitsuhiko@doodlestein@sseraphini Armin please take the time to clone his repo and ask your clanker to analyze the implementation.
just for the absolute fucking hilarity of how bad this guy's slop is.
It is so incredibly bad behind belief lol, he's like the ultimate slop scammer
https://t.co/ENvd8p627y
"not that you'd understand them" neither do you!
I ran some analysis on this Asupersync trash with an agent swarm (got through maybe 2% of the 1 million+ lines of trash)
your own repo has files committed in it that say it's trash btw lol (analysis from your own agents)
the big one is MODES_OF_REASONING_REPORT_AND_ANALYSIS_OF_PROJECT.md
it says asupersync has "metastasized" into something one person plus agents cannot maintain, and that ~30-40k lines of mathematical machinery have zero callers from any operational code. your own agents ran the grep.
what you built is what happens when you keep prompting an LLM for fancy-sounding theorems without understanding any of it or whether it belongs. persistent homology, sheaf theory, Dialectica, martingale drain certificates, spectral monitors, conformal calibration.
none of it is wired into anything.
it exists because you asked an agent "what's the most rigorous way to bound X" and pasted the answer. multiplied by every subsystem. the result is unusable trash. nobody who actually writes Rust would touch this with a five-foot pole, because the alternatives already exist: tokio, hyper, h2, quinn, sqlx, tonic, rustls. all of them years old. all of them deployed at planetary scale by AWS Firecracker, Cloudflare, Discord. all of them maintained by hundreds of people with security response teams. tokio has had `tx.reserve(); permit.send(v)` (your "novel" two-phase channel) since 2021.
you, one human, in 100 days, on nightly-only Rust, with "I do not accept outside contributions," with 6,211 .unwrap(), 1,041 panic!(), and 210 unreachable!() calls in your library code , roughly 7,500 places where the program just kills itself, in a language whose entire discipline is that library code does not panic, it returns errors and lets the caller decide, claim to replace all of them.
your README markets this for "financial, medical, infrastructure."
when someone asked you for paying users of your products, you skipped past 1.1 million lines of supposedly world-historic engineering, and pointed at a markdown-file subscription service lol.
It's not even about all the people you're fooling, it's about how much you're fooling yourself with this crap if you truly believe it. This shit is UNUSABLE GARBAGE.
And you point to it as proof that you're smart or something? (not that we'd understand your genius). I don't think you're an idiot necessarily, I think you've got very little self awareness.
The fact that you've spent (and keep spending) money on this is absolutely insane. No rational software developer would ever EVER use this.
Bunch of broke inexperienced bitches giving advice so far
Zopiclone is amazing.
14 hour business class to Japan, like I was fucking teleported.
Took it shortly before first meal, fell asleep for like 8 hours straight, woke up for another meal and then slumbered back easily.
Didn't feel drowsy or anything.
@gustaf Sold my bootstrapped startup for mid 8 figures less than a year after founding it.
The acquisition wasn't made public at the time.
Will be able to talk about it after the acquirer IPOs this year at least..
Absolutely no street cred from it 🤌🏻
Buddy I'm gonna tear you a new asshole with how easy it is to show your code is absolute GARBAGE.
I've never made anything that can compare to it, THANK FUCK 🤣.
How fucking embarrassed I'd be making the kind of retarded claims you make about your "contributions".
@coffeebreak_YT you should dedicate some time to this new type of AI slop scam artist (unsurprisingly a former crypto scam artist).
This guy basically asks LLMs "ok how can we rewrite software X but like, so good, with like the fanciest math and algorithms" and then stuffs all that shit where it makes absolutely zero sense.
He rewrote SQLite except like 60000x slower or something, the benchmarks by @KatanaLarp were hilarious.
I think at this point AI has made him so delusional he thinks he can recreate in a month what took hundreds of people infinitely smarter than him a decade.
You're a joke and a scam artist @doodlestein , I'm sad for the inexperienced devs wanting to genuinely learn whom you've fooled.
@doodlestein I literally made an entire tweet just to focus on the fucking hilarity of how shit franken TUI is. For animating or resizing, just incredibly fucking beyond belief bad.
https://t.co/7cqmn3Lnvk
lol I just can't get over this shit
(the Sinkhorn-Knopp entropy-regularized optimal transport theory for animating layout transitions when a sidebar opens)
Imagine you have a chess board, and you want to rearrange the pieces.
- Normal solution:
Pick up each piece. Put it where it goes. Done.
- FrankenTUI solution:
Model the chess board's current state as a probability distribution. Model the desired state as a second probability distribution. Build a cost matrix encoding the Euclidean distance between every pair of squares. Solve a Kantorovich optimal-transport problem with entropy regularization, smoothed by the Sinkhorn-Knopp algorithm, partitioned into block-diagonal sub-problems via cost penalties. Animate the pieces along the resulting transport plan trajectories. Recompute the whole optimization every animation frame, roughly 23 million matrix entries per Sinkhorn iteration on a 120x40 terminal, with multiple iterations to converge.
For animating a sidebar.
This is the math ML researchers run on GPUs to train Wasserstein GANs and align word embeddings between languages. He's running it on your CPU. To slide a panel.
I wish I could I unsee this cursed fucking code.
@zeeg help I'm fucking dying over here lol
"not that you'd understand them" neither do you!
I ran some analysis on this Asupersync trash with an agent swarm (got through maybe 2% of the 1 million+ lines of trash)
your own repo has files committed in it that say it's trash btw lol (analysis from your own agents)
the big one is MODES_OF_REASONING_REPORT_AND_ANALYSIS_OF_PROJECT.md
it says asupersync has "metastasized" into something one person plus agents cannot maintain, and that ~30-40k lines of mathematical machinery have zero callers from any operational code. your own agents ran the grep.
what you built is what happens when you keep prompting an LLM for fancy-sounding theorems without understanding any of it or whether it belongs. persistent homology, sheaf theory, Dialectica, martingale drain certificates, spectral monitors, conformal calibration.
none of it is wired into anything.
it exists because you asked an agent "what's the most rigorous way to bound X" and pasted the answer. multiplied by every subsystem. the result is unusable trash. nobody who actually writes Rust would touch this with a five-foot pole, because the alternatives already exist: tokio, hyper, h2, quinn, sqlx, tonic, rustls. all of them years old. all of them deployed at planetary scale by AWS Firecracker, Cloudflare, Discord. all of them maintained by hundreds of people with security response teams. tokio has had `tx.reserve(); permit.send(v)` (your "novel" two-phase channel) since 2021.
you, one human, in 100 days, on nightly-only Rust, with "I do not accept outside contributions," with 6,211 .unwrap(), 1,041 panic!(), and 210 unreachable!() calls in your library code , roughly 7,500 places where the program just kills itself, in a language whose entire discipline is that library code does not panic, it returns errors and lets the caller decide, claim to replace all of them.
your README markets this for "financial, medical, infrastructure."
when someone asked you for paying users of your products, you skipped past 1.1 million lines of supposedly world-historic engineering, and pointed at a markdown-file subscription service lol.
It's not even about all the people you're fooling, it's about how much you're fooling yourself with this crap if you truly believe it. This shit is UNUSABLE GARBAGE.
And you point to it as proof that you're smart or something? (not that we'd understand your genius). I don't think you're an idiot necessarily, I think you've got very little self awareness.
The fact that you've spent (and keep spending) money on this is absolutely insane. No rational software developer would ever EVER use this.
"not that you'd understand them" neither do you!
I ran some analysis on this Asupersync trash with an agent swarm (got through maybe 2% of the 1 million+ lines of trash)
your own repo has files committed in it that say it's trash btw lol (analysis from your own agents)
the big one is MODES_OF_REASONING_REPORT_AND_ANALYSIS_OF_PROJECT.md
it says asupersync has "metastasized" into something one person plus agents cannot maintain, and that ~30-40k lines of mathematical machinery have zero callers from any operational code. your own agents ran the grep.
what you built is what happens when you keep prompting an LLM for fancy-sounding theorems without understanding any of it or whether it belongs. persistent homology, sheaf theory, Dialectica, martingale drain certificates, spectral monitors, conformal calibration.
none of it is wired into anything.
it exists because you asked an agent "what's the most rigorous way to bound X" and pasted the answer. multiplied by every subsystem. the result is unusable trash. nobody who actually writes Rust would touch this with a five-foot pole, because the alternatives already exist: tokio, hyper, h2, quinn, sqlx, tonic, rustls. all of them years old. all of them deployed at planetary scale by AWS Firecracker, Cloudflare, Discord. all of them maintained by hundreds of people with security response teams. tokio has had `tx.reserve(); permit.send(v)` (your "novel" two-phase channel) since 2021.
you, one human, in 100 days, on nightly-only Rust, with "I do not accept outside contributions," with 6,211 .unwrap(), 1,041 panic!(), and 210 unreachable!() calls in your library code , roughly 7,500 places where the program just kills itself, in a language whose entire discipline is that library code does not panic, it returns errors and lets the caller decide, claim to replace all of them.
your README markets this for "financial, medical, infrastructure."
when someone asked you for paying users of your products, you skipped past 1.1 million lines of supposedly world-historic engineering, and pointed at a markdown-file subscription service lol.
It's not even about all the people you're fooling, it's about how much you're fooling yourself with this crap if you truly believe it. This shit is UNUSABLE GARBAGE.
And you point to it as proof that you're smart or something? (not that we'd understand your genius). I don't think you're an idiot necessarily, I think you've got very little self awareness.
The fact that you've spent (and keep spending) money on this is absolutely insane. No rational software developer would ever EVER use this.
lol I just can't get over this shit
(the Sinkhorn-Knopp entropy-regularized optimal transport theory for animating layout transitions when a sidebar opens)
Imagine you have a chess board, and you want to rearrange the pieces.
- Normal solution:
Pick up each piece. Put it where it goes. Done.
- FrankenTUI solution:
Model the chess board's current state as a probability distribution. Model the desired state as a second probability distribution. Build a cost matrix encoding the Euclidean distance between every pair of squares. Solve a Kantorovich optimal-transport problem with entropy regularization, smoothed by the Sinkhorn-Knopp algorithm, partitioned into block-diagonal sub-problems via cost penalties. Animate the pieces along the resulting transport plan trajectories. Recompute the whole optimization every animation frame, roughly 23 million matrix entries per Sinkhorn iteration on a 120x40 terminal, with multiple iterations to converge.
For animating a sidebar.
This is the math ML researchers run on GPUs to train Wasserstein GANs and align word embeddings between languages. He's running it on your CPU. To slide a panel.
I wish I could I unsee this cursed fucking code.
@zeeg help I'm fucking dying over here lol
I'm not even sure he knows at this point? In the sense that if I was his lawyer I could see myself arguing an insanity defense based on how fucking delusional he is.
I mean I also believe he's a massive piece of shit scam artist but I would genuinely hear someone out explaining why he may just be experiencing a novel type of psychotic break.