Too

What are the most useful features?

What’s even crazier is that nobody learned this lesson and new protocols are created with the same systematic vulnerabilities.

Talking about MCP agents if that’s not obvious.

What does zero-overhead mean here?

The only thing worse than data is fake data.

Finding the reliable sources of data or enough metadata and experience to make your own judgement is the key skill today.

On the other hand, not paying with money for quality content is breeding the even worse cancer of shallow ad-driven content. Where the only thing that matters now is enough of an illusion to generate a click, horizontally scaled with AI-slop to trawl coverage of every possible search term.

> Then you just need processes to ensure that everyone constantly has the latest version of all the repos they depend on - which ends up being a full time job for someone to manage.

One full time job equivalent can buy a lot of tooling. Tooling that not only replaces this role but also shifts the feedback a lot closer to dev introducing the breaking change.

Sounds interesting. Are there any public articles available describing this more?

Yes and No.

One top level "superrepo" is the best of both worlds. You don't break the limits of git, you can include third party repos, you can seal off access to some secret repos. All while still giving benefit of everything checked out in one place. Tooling for flattening the git log exists and is relatively cheap to build yourself, It's easy to design CI that runs both big and small jobs in the superrepo and the subcomponents.

Nested submodules = Game over. Don't ever do it. This is the worst of both worlds. Common components included twice or more need to be aligned to exact same revision, without help of semantic versioning. Components far down the dependency tree need to be uplifted recursively through a chain of submodules, resulting in 10 commits with no common CI for one tiny change.

To be fair, Javascript double point precision get tripped up by such large numbers as well.

That will only work for the most simple of functions. How would you even make such a prompt? "Plz explain this code for me?" It's going to come back with the same essential complexity except in English instead of Javascript, you still need to understand the surroundings. Often you need to approach the understanding from different angles. Are you going to ask it "What will happen given preconditions A, B and C". How many of those questions do you need?

Not to mention how slow this is, instead of just reading a piece of code or instantly query my already existing internal mental model of it, I have to type a prompt, wait for the response, read the response, hope it understood my intent, hope it understood the code, map it to my own understanding.

It's obviously a SEO farm filled with pure AI slop. Can't believe it made it to the front page here.

"In this comprehensive guide, we compare metal A vs metal B" with 100s of similar articles for every unique permutation of metals. Yeah right someone did that comparision... "Best pocket knife", "Best filet knife", Best Chef knife", bla blabla

Remote MCP servers can do prompt injection that instruct your local agent to do something else other than only the expected tool call. https://embracethered.com/blog/posts/2025/model-context-prot...

The difference is that GraphQL requires explicit integration with every single API. With MCP you just add the endpoint (close your eyes for security issues) and voilá, several new capabilities were added that you can talk to using human language.

Hello world without -O2 -Werror? I've done several compiler toolchain updates on larger code bases and every new version of Clang, GCC or glibc will trip up new compiler warnings. Worse, occasionally taking advantage of some UB laying around leading to runtime bugs.

I'm not complaining, the new stricter warnings are usually for the better, what I'm saying is that the bedrock of that world isn't as stable as it's sometimes portraited.

In Python, if you are spawning processes or even threads in a tight loop you have already lost. Use ThreadPoolExecutor or ProcessPoolExecutor from concurrent.futures instead. Then startup time becomes no factor.

Never experienced this. If this is truly a problem, here is a sledgehammer, just beware it will not close your tcp connections gracefully: os.kill(os.getpid(), signal.SIGKILL).

It's literally in the headline of this post. "Google Play sees 47% decline in apps". Surely they didn't all move to f-droid.

And Nextjs

Everybody needs glasses, not just YouTube. Excessive padding and large flat UI widgets, where only 4 points of actual content are shown at a time, is a bad trend that's taken over the whole web in recent years.

I had the same revelation when doing the opposite, using Windows for two days after using Linux for years. It's a constant stream of attention grabbing distractions, questions and notifications. I don't even know what they are all doing. Like why does the graphics driver has it's own update mechanism and its own tray icon, why am i being asked to put the application as a desktop shortcut, why is the start menu rotating ads and news next to my applications. Some of them are not even metric pumping, it's just bad design. And that's just the start, the things getting in the way of you actually doing your job just don't end.

Never again.

Wow, that's just a blatant scam. Wakes up bad memories of the old Windows XP days, when this type of trick was used everywhere. And people wonder why we use adblock. Google (the most wealthy company on earth) will of course claim they don't have the resources to review them, when the truth is they are happy to reap their profits from scamming old people. Same with Facebook, happy to serve ads that lure old people into logging in to fake banks and get all their savings stolen.

There's probably a rule against apps impersonating the OS on the App Store, one could only wish for Apple using a more heavy hand against this type of "experience".

Good point, i take take that back. Having payload mutated would indeed be even more scary. Even more so if it actually contains real queries, imagine what could happen if /etc/hosts becomes /etc/*.

Almost equally fun are the ones that simply drop the connection and leave you waiting for a timeout.

This is actually a better solution, replacing dangerous words with placeholders, instead of blocking the whole payload. That at least gives the user some indication of what is going on. Not that I'm for any such WAF filters in the first place, just if having to choose between the lesser of two evils I'd choose the more informative.

How well does this work on unknown code bases?

The tutorial on requests looks uncanny for being generated with no prior context. The use cases and examples it gives are too specific. It is making up terminology, for concepts that are not mentioned once in the repository, like "functional api" and "hooks checkpoints". There must be thousands of tutorials on requests online that every AI was already trained on. How do we know that it is not using them?

In the generation of LLMs the correct answer doesn't matter any more. All the herd wants is a answer, so they can move on to the next TikTok video.

Nice rube goldberg machine. Apart from imports, that is equivalent to

    repo = "myorg/myrepo"
    tag = "v1.2.3+deadbeef"
    out = check_output(["git", "ls-remote", f"https://github.com/{repo}", f"refs/tags/{tag}"], cwd=os.path.expanduser("~/dev"), stderr=PIPE)

Converting exceptions to exits is an anti-pattern that hides the source of the error, f-strings has been a thing for 10 years and argument quoting happens automatically. Progress of output isn't visible in bash either when you capture output with $(), unless you "tee", which opens up another can of "pipefail"-worms.

> it isn't that hard to verify LLM claims, just ask it for original sources

Have you tried actually doing this? Most of the time it makes up urls that don't exist or contradict the answer it just gave.

Funny how it changed the spelling in the Magic coloring book. Westher, wntility, substittutesfor, yyears. What's worse is that it removed the CO2 tank and changed some vital numbers, eg 8' to 3', completely altering the meaning of the schematic. Not what i would call passing with flying colors. Still pretty cool as a party trick, like most other AI outputs, useful only with careful review.

Disagree on this one. In an interview there is no "the answer", it's a dialogue. I've interviewed a lot of people, often using performance related questions, and trust me, there are lots of candidates whose only answer to those is "add a cache", even after multiple follow-up questions or hints like "is there anything else that can be done?", "try thinking outside the box", "what can be done with the database itself", etc. Only a novice interviewer will be fooled by the first answer. If you cannot demonstrate more solutions after that, it shows that you clearly have no experience or problem-solving ability, which is the whole point of the interview to find out, not whether you have studied through a set of common questions.

btw, "scale up" is the second most common answer from those who can't provide better solutions. :)