Lightnews — Scholar-powered news

Shafik Yaghmour

@shafik.bsky.social

960 followers 240 following 660 posts

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

Mastodon: https://hachyderm.io/@shafik
Blog: https://shafik.github.io/

#cplusplus #compilers

Posts Replies Media Videos

Shafik Yaghmour

@shafik.bsky.social

Cursed Code of the day: godbolt.org/z/br5c8Msfo

#cplusCplus

godbolt screen shot:

Code:

bool f() {
return (char)256.f < (char)1.f;
}

Compiled w/ -O2

gcc result:

"f()":
mov eax, 1
ret

clang result:

f():
ret

December 4, 2025 at 2:35 AM

Shafik Yaghmour

@shafik.bsky.social

Great story on why the 50 year mortgages makes no sense at all: www.marketplace.org/story/2025/1...

According to the American Enterprise Institute, a 50-year mortgage would, initially, increase a homeowner’s buying power by 8%.

“Over time, maybe months, maybe years, that would fade,” said Edward Pinto, a senior fellow at the American Enterprise Institute.

Basically, if everyone’s buying power goes up, so does the price of housing. And a longer mortgage also means it takes longer to build up equity.

November 21, 2025 at 12:15 AM

Shafik Yaghmour

@shafik.bsky.social

C++ Retro Quiz #46

Answer here: hachyderm.io/@shafik/1155...

#Cplusplus
#Cpppolls

$Retro C++ quiz #46 #include <iostream> #include <string> int main() { using namespace std::string_literals; std::string s1("hello world",5); std::string s2("hello world"s,5); std::cout<<s1<<s2; // What is the output? } Without checking: A. hellohello B. world world C. hello world D. worldhello$

November 20, 2025 at 5:36 PM

Shafik Yaghmour

@shafik.bsky.social

"Learning with AI falls short compared to old-fashioned web search": theconversation.com/learning-wit...

Not shocking. What was interesting, is this held over many studies and w/ a large pool of participants.

#ai

The data revealed a consistent pattern: People who learned about a topic through an LLM versus web search felt that they learned less, invested less effort in subsequently writing their advice, and ultimately wrote advice that was shorter, less factual and more generic. In turn, when this advice was presented to an independent sample of readers, who were unaware of which tool had been used to learn about the topic, they found the advice to be less informative, less helpful, and they were less likely to adopt it.

November 20, 2025 at 4:58 PM

Shafik Yaghmour

@shafik.bsky.social

C++ quiz #307

Answer here: hachyderm.io/@shafik/1155...

#Cplusplus
#CppPolls

#307 Given the following in C++

template<class T>
void f() {
struct Y {
using type = int;
};
Y::type y; // Missing typename?
}

Without checking:

A. Yes
B. No

November 19, 2025 at 6:57 PM

Shafik Yaghmour

@shafik.bsky.social

Retro C++ quiz #46 Multiplying shorts C style

answer here: hachyderm.io/@shafik/1154...

#CppPolls
#Programming

$Retro C++ quiz #46 Given the following in C: unsigned short x=0xFFFF,y=0xFFFF; Without checking, assuming LP64. What is the correct format specifier X below and result: printf( “%X\n”, x*y); % | result ========= A. d | -131071 B. u | 4294836225 C. d | UB D. u | UB #CPolls #Programming$

October 31, 2025 at 4:44 PM

Shafik Yaghmour

@shafik.bsky.social

Retro C++ quiz #45 Multiplying shorts

Answer here: hachyderm.io/@shafik/1154...

#Cplusplus
#Cpppolls

Retro C++ quiz #45

Given the following in C++:

unsigned short x=0xFFFF;
unsigned short y=0xFFFF;
auto z=x*y;
std::cout << z;

Without checking, assuming an LP64 system what is the result:

A. 4294836225
B. -131071
C. fffe0001
D. Undefined behavior

October 31, 2025 at 4:42 PM

Shafik Yaghmour

@shafik.bsky.social

I am just reading an article trying to convince folks that data centers are not causing the rapid rise in power prices and this is ludicrous claim to make: www.bloomberg.com/graphics/202...

Al boom on the power market with unprecedented granularity. The locations and prices were tracked and aggregated monthly by Grid Status, an energy data analytics platform. Bloomberg analyzed this data in relation to data center locations, from DC Byte, and found that electricity now costs as much as 267% more for a single month than it did five years ago in areas located near significant data center activity.

October 25, 2025 at 11:42 PM

Shafik Yaghmour

@shafik.bsky.social

Apparently today is Mole Day: en.wikipedia.org/wiki/Mole_Day

Happy Mole Day!

This seems like a apropos time to share "A Mole of Moles": what-if.xkcd.com/4/

While it sounds quite amusing it gets very grim very quickly.

The mole planet is now a giant sphere of meat. It has a lot of latent energy (there are enough calories in the mole planet to support the Earth’s current population for 30 billion years). Normally, when organic matter decomposes, it releases much of that energy as heat. But throughout the majority of the planet’s interior, the pressure is over a hundred megapascals, which is enough to kill all bacteria and sterilize the mole remains—leaving no microorganisms to break down the mole tissues.

October 23, 2025 at 2:50 PM

Shafik Yaghmour

@shafik.bsky.social

C++ quiz #306 This

Answer here: hachyderm.io/@shafik/1154...

#Cplusplus
#Cpppolls

#306 Given the following in C++

struct A {};

struct B {
void f(this A); // A
void f(this int); // B
};

Without checking, which is well-formed?

A. A
B. B
C. Neither
D. Both

October 22, 2025 at 2:59 PM

Shafik Yaghmour

@shafik.bsky.social

"Your diet’s impact on the planet depends on where you live. Look up your city.": wapo.st/4ocwoy8

I get a lot of pushback from people when I say cutting beef is the best thing you can do. Apparently there are a lot of influencers who say beef is not that bad.

Bar graph showing alternative to cut carbon hoofprint

Swap half your beef for chicken 33%
Cut food waste in half 16%
Meatless Mondays 14%

Combined impact 51%

On average, 73 percent of every city’s hoofprint comes down to beef. Not only is it more popular, producing beef generates far more planet-warming pollution compared with raising hogs and chickens. However, raising cattle in feed lots can be more energy efficient, despite its other environmental and ethical costs. And the type of feed a farm uses and how it manages waste can tip these cows’ emissions the other way.

The distance that cattle feed is transported, as well as how far cows travel to be slaughtered, also matters, since this can span thousands of miles.

October 21, 2025 at 4:36 PM

Shafik Yaghmour

@shafik.bsky.social

Cursed diff highlighting here.

How can github be so bad.

#programming

Diff it higghlights the middle paren as opposed to the outher paren

- assert(!Success || !Trap.hasErrorOccurred() &&
- "Substitution failures must be handled "
- "by CheckConstraintSatisfaction.");
+ assert((!Success || !Trap.hasErrorOccurred()) &&

October 20, 2025 at 3:07 AM

Shafik Yaghmour

@shafik.bsky.social

C++quiz #305

Answer here: hachyderm.io/@shafik/1153...

#Cplusplus
#CppPolls

#305 Given the following in C++

auto f();
int f(); // Well-formed?

Without checking:

A. Yes
B. No

October 16, 2025 at 4:48 PM

Shafik Yaghmour

@shafik.bsky.social

Retro C++ quiz #42 vector cast?

Answer here: hachyderm.io/@shafik/1153...

#CppPolls
#cplusplus

Retro C++ quiz #42

Given the following in C++:

#include <vector>

void f() {
(std::vector<float>)(5,2);
}

Without checking, is the code:

A. Well-formed
B. Ill-formed
C. Undefined Behavior

October 10, 2025 at 5:13 PM

Shafik Yaghmour

@shafik.bsky.social

The Usual Arithmetic Confusions: shafik.github.io/c++/2021/12/...

#cplusplus

There are a lot of aspects of C++ that are not well understood and lead to all sorts of confusion. The usual arithmetic conversions and the integral promotions are two such aspects.
Certain binary operators (arithmetic, relational and spaceship) require their operands to have a common type. The usual arithmetic conversions are the set of steps that gets operands to a common type. While the integral promotions brings integral types smaller than int and unsigned int to either int or unsigned int depending on which one can represent all the values of the source type. This is one of the areas in C++ that comes directly from C, so pretty much all of these examples applies to C as well as C++.

October 9, 2025 at 10:54 PM

Shafik Yaghmour

@shafik.bsky.social

"A small number of samples can poison LLMs of any size": www.anthropic.com/research/sma...

Black hats have to be trying (Why wouldn't they, the potential gains are large). If it's really this easy then likely malicious code is already making its way into real code bases, maybe at scale.

😱

#ai

In a joint study with the UK AI Security Institute and the Alan Turing Institute, we found that as few as 250 malicious documents can produce a "backdoor" vulnerability in a large language model—regardless of model size or training data volume. Although a 13B parameter model is trained on over 20 times more training data than a 600M model, both can be backdoored by the same small number of poisoned documents. Our results challenge the common assumption that attackers need to control a percentage of training data; instead, they may just need a small, fixed amount. Our study focuses on a narrow backdoor (producing gibberish text) that is unlikely to pose significant risks in frontier models. Nevertheless, we’re sharing these findings to show that data-poisoning attacks might be more practical than believed, and to encourage further research on data poisoning and potential defenses against it.

October 9, 2025 at 10:21 PM

Shafik Yaghmour

@shafik.bsky.social

C++ Retro quiz #41

Answer here: hachyderm.io/@shafik/1153...

#CppPolls
#cplusplus

Retro C++ quiz #41

Given the following in C++:

struct B {
int x:4;
};

void f(int &x);

void b() {
B b1{1};

auto &[y] = b1; // A
y = 5; // B
f(y); // C
}

Without checking, the code is ill-formed (requires a diagnostic) at point:

A. A
B. B
C. C

October 3, 2025 at 4:33 PM

Shafik Yaghmour

@shafik.bsky.social

C++ quiz #304 Relational and nullptr

Answer here: hachyderm.io/@shafik/1152...

#Cplusplus
#CppPolls

#304 Given the following in C++

void f(void *p) {
p > nullptr; // Well-formed?
}

Without checking:

A. Yes
B. No

September 25, 2025 at 3:49 PM

Shafik Yaghmour

@shafik.bsky.social

"AI-Generated “Workslop” Is Destroying Productivity": hbr.org/2025/09/ai-g...

New vocabulary word just dropped

WorkSlop: AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task

#Ai

In collaboration with Stanford Social Media Lab, our research team at BetterUp Labs has identified one possible reason: Employees are using AI tools to create low-effort, passable looking work that ends up creating more work for their coworkers. On social media, which is increasingly clogged with low-quality AI-generated posts, this content is often referred to as “AI slop.” In the context of work, we refer to this phenomenon as “workslop.” We define workslop as AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.

Here’s how this happens. As AI tools become more accessible, workers are increasingly able to quickly produce polished output: well-formatted slides, long, structured reports, seemingly articulate summaries of academic papers by non-experts, and usable code. But while some employees are using this ability to polish good work, others use it to create content that is actually unhelpful, incomplete, or missing crucial context about the project at hand. The insidious effect of workslop is that it shifts the burden of the work downstream, requiring the receiver to interpret, correct, or redo the work. In other words, it transfers the effort from creator to receiver.

In another case, a frontline manager in the tech sector described their reaction: “It was just a little confusing to understand what was actually going on in the email and what he actually meant to say. It probably took an hour or two of time just to congregate [sic] everybody and repeat the information in a clear and concise way.”

September 24, 2025 at 4:23 PM

Shafik Yaghmour

@shafik.bsky.social

"Why Good Summaries are Important to a Pull Request": shafik.github.io/software%20d...

#programming

Why Good Summaries are Important to a Pull Request
Code Reviewers are a precious resource in an open source project, often they have limited bandwidth. A critical goal when drafting a Pull Request (PR) should be, how does one minimize the amount of time it takes a reviewer to do a review. Especially we want to minimize context switches, which can have a large impact on productivity. Today I am going to argue that the summary is one of the most important aspects of your pull request to get right. A good summary should reduce the need for the reviewers to go to their editor to examine the larger surrounding code and minimize the amount of context switching a reviewer has to do.

Another way to think about it, is that you’re telling a story to your reviewers. Just like in any conversation, if you leave out key parts or details your audience will get lost. Perhaps think of it like a good joke, if you have to explain the joke after you tell it, it was not a good joke. You want the PR to land well the first time through.

Ultimately the more effective you are at explaining your change the better feedback you will receive from your reviewer(s).

September 22, 2025 at 4:21 PM

Shafik Yaghmour

@shafik.bsky.social

"As businesses wait for a tariff ruling, some have run out of time" www.marketplace.org/story/2025/0...

Small businesses are not doing well under tariffs.