@shafik@hachyderm.io avatar

shafik

@[email protected]

Compiler Engineer clang front end. Interests: C++, C, and undefined behavior. Martial Artist, Book Worm and Dad

Blog: https://shafik.github.io/
Bsky: https://bsky.app/profile/shafik.bsky.social (@shafik.bsky.social)
Twitter: https://twitter.com/shafikyaghmour (@shafikyaghmour)

This profile is from a federated server and may be incomplete. View on remote instance

@shafik@hachyderm.io avatar shafik , to random

This is a damning article from the Wikipedia editors on GenAI articles written for Wikipedia: https://wikiedu.org/blog/2026/01/29/generative-ai-and-wikipedia-editing-what-we-learned-in-2025/

Far more insidious, however, was something else we discovered: More than two-thirds of these articles failed verification. That means the article contained a plausible-sounding sentence, cited to a real, relevant-sounding source. But when you read the source it’s cited to, the information on Wikipedia does not exist in that specific source. When a claim fails verification, it’s impossible to tell whether the information is true or not. For most of the articles Pangram flagged as written by GenAI, nearly every cited sentence in the article failed verification.
ALT
@shafik@hachyderm.io avatar shafik , to random

Retro C++ quiz

<iostream>
<string>

int main() {
using namespace std::string_literals;
std::string s1("hello world",5);
std::string s2("hello world"s,5);

std::cout<<s1<<s2; // What is the output?
}

Without checking:

A. hellohello
B. world world
C. hello world
D. worldhello


@shafik@hachyderm.io avatar shafik , to random

There is nothing like the scorn of a developer who bought into something and then realizes that they were very wrong.

"Where's the Shovelware? Why AI Coding Claims Don't Add Up": https://mikelovesrobots.substack.com/p/wheres-the-shovelware-why-ai-coding

My argument: If so many developers are so extraordinarily productive using these tools, where is the flood of shovelware? We should be seeing apps of all shapes and sizes, video games, new websites, mobile apps, software-as-a-service apps — we should be drowning in choice. We should be in the middle of an indie software revolution. We should be seeing 10,000 Tetris clones on Steam. Consider this: with all you know about AI-assisted coding and its wide adoption, if I showed you charts and graphs of new software releases across the world, what shape of that graph would you expect? Surely you’d be seeing an exponential growth up-and-to-the-right as adoption took hold and people started producing more?

My argument: If so many developers are so extraordinarily productive using these tools, where is the flood of shovelware? We should be seeing apps of all shapes and sizes, video games, new websites, mobile apps, software-as-a-service apps — we should be drowning in choice. We should be in the middle of an indie software revolution. We should be seeing 10,000 Tetris clones on Steam. Consider this: with all you know about AI-assisted coding and its wide adoption, if I showed you charts and graphs of new software releases across the world, what shape of that graph would you expect? Surely you’d be seeing an exponential growth up-and-to-the-right as adoption took hold and people started producing more?
ALT
@thomasfuchs@hachyderm.io avatar thomasfuchs , to random

So these people charge money for their super serious reporting

New York Times article excerpt: Most experts acknowledge that a takeover by artificial intelligence is coming for the video game industry within the next five years, and executives have already started preparing to restructure their companies in anticipation. After all, it was one of the first sectors to deploy A.I. programming in the 1980s, with the four ghosts who chase Pac-Man each responding differently to the player's real-time movements.
ALT
shafik ,
@shafik@hachyderm.io avatar

@thomasfuchs @glennf

I don't think it is that they don't care, it is safe to stay within a particularly framing because that is what everyone is else is doing.

Standing out from the crowd means being wrong carries a much greater penalty.

It is a variation of "Nobody ever gets fired for buying IBM"

Having a solid tech reporting team that had the depth to stick their necks out would require significant investment.

@shafik@hachyderm.io avatar shafik , to random

"The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity": https://machinelearning.apple.com/research/illusion-of-thinking

Nothing surprising for me here but I salute the folks willing to go against the grain and ask the hard questions. Folks can now look at this work and maybe have a better answer to how limited these models really are.

Conclusion In this paper, we systematically examine frontier Large Reasoning Models (LRMs) through the lens of problem complexity using controllable puzzle environments. Our findings reveal fundamental limitations in current models: despite sophisticated self-reflection mechanisms, these models fail to develop generalizable reasoning capabilities beyond certain complexity thresholds. We identified three distinct reasoning regimes: standard LLMs outperform LRMs at low complexity, LRMs excel at moderate complexity, and both collapse at high complexity. Particularly concerning is the counterintuitive reduction in reasoning effort as problems approach critical complexity, suggesting an inherent compute scaling limit in LRMs. Our detailed analysis of reasoning traces further exposed complexitydependent reasoning patterns, from inefficient “overthinking” on simpler problems to complete failure on complex ones. These insights challenge prevailing assumptions about LRM capabilities and suggest that current approaches may be encountering fundamental barriers to generalizable reasoning. Finally, we presented some surprising results on LRMs that lead to several open questions for future work. Most notably, we observed their limitations in performing exact computation; for example, when we provided the solution algorithm for the Tower of Hanoi to the models, their performance on this puzzle did not improve. Moreover, investigating the first failure move of the models revealed
ALT
@shafik@hachyderm.io avatar shafik , to random

Click on your old gobolt/g links to preserve them: https://xania.org/202505/compiler-explorer-urls-forever

Bring out your old godbolt links

@shafik@hachyderm.io avatar shafik , to random

Stackoverflow bans LLMs: https://meta.stackoverflow.com/q/421831/1708801

I missed this but still important news.

There is nothing shocking in the rationale, the quality is just not high enough and the content just sounds right enough that the effort to review swamps out good content.

All use of generative AI (e.g., ChatGPT1 and other LLMs) is banned when posting content on Stack Overflow. This includes "asking" the question to an AI generator then copy-pasting its output as well as using an AI generator to "reword" your answers. Please see the Help Center article: What is this site’s policy on content generated by generative artificial intelligence tools? Overall, because the average rate of getting correct answers from ChatGPT and other generative AI technologies is too low, the posting of content created by ChatGPT and other generative AI technologies is substantially harmful to the site and to users who are asking questions and looking for correct answers.
ALT
@shafik@hachyderm.io avatar shafik , to random

Talk about saying the quiet part out loud.

Folks who don't understand AI like it more but once they understand it they find it less appealing. SOOOOOO companies should target low literacy customers and keep folks ignorant

https://journals.sagepub.com/doi/10.1177/00222429251314491

Like wow 🤯

As artificial intelligence (AI) transforms society, understanding factors that influence AI receptivity is increasingly important. The current research investigates which types of consumers have greater AI receptivity. Contrary to expectations revealed in four surveys, cross country data and six additional studies find that people with lower AI literacy are typically more receptive to AI. This lower literacy-greater receptivity link is not explained by differences in perceptions of AI’s capability, ethicality, or feared impact on humanity. Instead, this link occurs because people with lower AI literacy are more likely to perceive AI as magical and experience feelings of awe in the face of AI’s execution of tasks that seem to require uniquely human attributes. In line with this theorizing, the lower literacy-higher receptivity link is mediated by perceptions of AI as magical and is moderated among tasks not assumed to require distinctly human attributes. These findings suggest that companies may benefit from shifting their marketing efforts and product development towards consumers with lower AI literacy. Additionally, efforts to demystify AI may inadvertently reduce its appeal, indicating that maintaining an aura of magic around AI could be beneficial for adoption.
ALT
@shafik@hachyderm.io avatar shafik , to random

Teen Vogue yet again doing the needed work that few others in the media seem willing to do.

https://www.teenvogue.com/story/ice-watch-programs-immigrants-how-to-start

@molly0xfff@hachyderm.io avatar molly0xfff , to random

somewhere, someone is trying to dig through my years of wikipedia editing history to prove my sinister influence, only to be driven mad by having to paw through reams of edits on things like figuring out how we should spell the name of Michael Keaton’s character in Beetlejuice

shafik ,
@shafik@hachyderm.io avatar

@molly0xfff so if you edit Beetlejuice three times does that summon him?

@molly0xfff@hachyderm.io avatar molly0xfff , to random

having succeeded in purchasing positions of influence inside the White House, the billionaires and tech VCs have now set their sights back on Wikipedia

Shaun Maguire @shaunmmaguire Wikipedia has been ideologically captured for years It's important that this is finally coming to light The most extreme bias is probably against Israel Foundation AI models are being trained on this biased data as if it's ground truth
Elon Musk @elonmusk · 3h Defund Wikipedia until balance is restored!
Elon Musk @elonmusk · 3h Since legacy media propaganda is considered a “valid” source by Wikipedia, it naturally simply becomes an extension of legacy media propaganda!

Elon Musk @elonmusk · 3h Since legacy media propaganda is considered a “valid” source by Wikipedia, it naturally simply becomes an extension of legacy media propaganda!
ALT
shafik ,
@shafik@hachyderm.io avatar

@molly0xfff not being able to quote toot is so frustrating.

shafik ,
@shafik@hachyderm.io avatar

@molly0xfff folks need to step up and donate more.

@regehr@mastodon.social avatar regehr , to random

I went to find awesome floating point bit tricks and found a pile of UB :(

https://karmafx.net/docs/karmafx_floattricks.pdf

C/C++ functions freely casting between pointer types, violating the "strict aliasing" rules that these languages have. one example below. the rest are in the document on page 8 Fast Float Math Functions Absolute _ forceinline float fast_abs(float a) { unsigned int f = (* (unsigned int*)&(a)) &0x7FFFFFFE; return (* (float*)&(f)); }
ALT
shafik ,
@shafik@hachyderm.io avatar

@regehr

-fno-strict-aliasing FTW 😭

@eniko@peoplemaking.games avatar eniko , to random

i love programming C

C comment saying "look, stdlib.h in visual studio using clang-cl just throws a fuckton of fake errors, okay? i'm not responsible for this i just want my error list window to show actual errors." below it there's an if not defined for intellisense, which includes stdlib.h, but if intellisense is defined there are definitions for NULL and fake implementations of malloc, realloc, and free
ALT
shafik ,
@shafik@hachyderm.io avatar

@eniko

not that we don't have a ton of cursed code in C but this is a MSVC/Windows problem not a C problem.

This is also a problem that intellisense has a different frontend 😱

@shafik@hachyderm.io avatar shafik , to random

Quip on double negation from “Says Who”

An MIT linguistics professor was lecturing his class the other day. "In English," he said, "a double negative forms a positive. How-ever, in some languages, such as Russian, a double negative remains a negative. But there isn't a single language, not one, in which a double positive can express a negative." A voice from the back of the room piped up, "Yeah, right."
ALT
@chandlerc@hachyderm.io avatar chandlerc , to random

So, just for the sake of curiosity -- how bad is it to create a 563486 character long string literal?

Clang very helpfully tells me that it exceeds the maximum length guaranteed to work -- 64k -- but c'mon, do any compilers really care?

I'm feeling lazy about building the compiler-explorer example that checks this...

shafik ,
@shafik@hachyderm.io avatar

@chandlerc

The compiler is trying to be helpful Chandler 🧐

I would be curious how the other implementations handle it though if you ever feel up to creating a minimal reproducer.

@shafik@hachyderm.io avatar shafik , to random

Crazy graphic of the day.

Wage theft greatly eclipses all other forms of theft value: https://www.tcworkerscenter.org/2018/09/wage-theft-vs-other-forms-of-theft-in-the-u-s/

Graphic shows Wage Theft (when employers steal from employees) Vs Other types of theft Minimum wage violations $23.20 billion Overtime violations $8.80 billion Rest Break violations $4.00 billion Off-the-clock violations $3.20 billion Vs Larceny $5.30 billion Burglary $4.10 billion Auto Theft $3.80 billion Robbery $ 0.34 billion
ALT
@shafik@hachyderm.io avatar shafik , to random

I meet a microbiologist today

They were wag BIGGER than I expected 🥁

@shafik@hachyderm.io avatar shafik , to random

AI power center usage is causing the life of some coal power plants to be extended: https://wapo.st/3KSgQxx

It found data centers will account for 8 percent of total electricity use in the United States by 2030, a near tripling of their share today. New solar and wind energy will meet about 40 percent of that new power demand from data centers, the forecast said, while the rest will come from a vast expansion in the burning of natural gas. The new emissions created would be comparable to that of putting 15.7 million additional gas-powered cars on the road. “We all want to be cleaner,” Brian Bird, president of NorthWestern Energy, a utility serving Montana, South Dakota and Nebraska, told a recent gathering of data center executives in Washington, D.C. “But you guys aren’t going to wait 10 years … My only choice today, other than keeping coal plants open longer than all of us want, is natural gas. And so you’re going see a lot of natural gas build out in this country.”

In Omaha, where Google and Meta recently set up sprawling data center operations, a coal plant that was supposed to go offline in 2022 will now be operational through at least 2026. The local utility has scrapped plans to install large batteries to store solar power. These concrete developments in energy markets contrast with tech companies’ futuristic promises. A recent Goldman Sachs analysis of energy that will power the AI boom into 2030 did not even consider small nuclear plants or futuristic fusion generators.
ALT
@shafik@hachyderm.io avatar shafik , to random

TIL Binet's Formula for calculating the Nth Fibonacci number: https://mathworld.wolfram.com/BinetsFormula.html

It is pretty amazing that this formula gives us whole numbers.

Enjoy! 🤯

Binet's Formula Binet's formula is an equation which gives the nth Fibonacci number as a difference of positive and negative nth powers of the golden ratio phi. It can be written as = ((1+sqrt(5))^n-(1-sqrt(5))^n)/(2^nsqrt(5)).
ALT
@shafik@hachyderm.io avatar shafik , to random

#2 Christmas C++

Santa I want ...

Four panels, kids talking to Santa Kid: "For Christmas I want a dragon!" Santa: "Be Realistic" Kid: I want to understand all of C++ Santa: What color do you want your dragon
ALT