If you know me, you know my line: LLMs lie. A lot.
In 2025, Columbia's Tow Center proved AI search engines botch citations.
I re-ran their test in 2026: 800 answers, and I clicked every link. About 70% wrong, and half the links broken or fake.
My addition: a checkpoint that clicks every link before an answer reaches you, checks the quote is on the page, the site, the date. Fabricated links reaching the reader fell to zero. Wrong answers fell from about 7 in 10 to about 1 in 8.
The engines lie just as much; the lies just can't get through.
The receipts, case by case: https://t.co/i2ZQLyjph1
The code: https://t.co/9cZBVwOtxy
I ran this 800x across 4 AI search engines on 200 real news excerpts.
Wrong ~7 in 10. Over half the cited links broken, dead, or not the source.
The fix isn't a smarter model. It's refusing to take its word for it.
https://t.co/SEExxYhqax
An AI search engine told me a story came from the LA Times. Gave me the link: https://t.co/aVk7ynecPV
It doesn't exist. The LA Times never ran it. Real source: Entertainment Weekly.
It invented a source, borrowed a real paper's name, and faked a URL to make it checkable.
The fix isn't a better employee. It's a checkpoint: nothing's marked done until something independent checks it against what was ordered.
I ran it on AI citations. Alone, engines were wrong 7 in 10. Through the checkpoint, 0 fabrications reached anyone.
https://t.co/GYLFtdQMzD
Ordered dinner on the Sonic app. It said "delivered." Half of it wasn't in the bag.
Nobody lied. Every step reported success. The food still wasn't there.
That's the exact problem I just spent a month measuring in AI. ๐งต
๐๐ผ Did you know in Python, functions are like VIPs at a concert? ๐ค๐ They're called first-class functions! Why? Because in Python, functions get the red-carpet treatment.
๐ Here's what makes them 'first-class':
You can pass them around like party invites ๐ - assign them to variables, or pass them as arguments to other functions.
They can return from other functions like a boomerang comes back to you
You can modify them on the fly, just like adding a filter to your favorite selfie
* Example Time: Let's say we have a function print_greeting that says "Hello!". We can pass this function to another function greet_twice to double the fun!
def print_greeting():
print("Hello!")
def greet_twice(func):
func()
func()
greet_twice(print_greeting)
Output:
Hello!
Hello!
This flexibility makes Python super versatile. It's like having a Swiss Army Knife in your toolkit!
Next time you code in Python, remember - functions are not just lines of code, they're the stars of the show!
If you like my posts, please follow me for more.
#Python #Programming #CodeFun #FirstClassFunctions
โ ๏ธ๐ Attention Pythonistas! ๐ Ned Batchelder has crafted an insightful note on default values in functions, highlighting how they can sometimes yield surprising results. ๐ค๐ก Definitely worth a quick read for a deeper understanding.
๐ Check it out here: Ned Batchelder's Blog on Default Values in Python Functions https://t.co/mUdWksRiq0
If you like my posts and want to see more of them, just follow me.
#Python #ProgrammingTips #NedBatchelder #PythonDefaults ๐๐จโ๐ป๐
Educate: https://t.co/Y7drFlpnO6
๐ "Exploring Dual Thread Pools in Python for Optimal Performance" by Itamar Turner-Trauring ๐
๐ค Ever wondered why you need different thread pools for efficient data processing in Python? Let's dive in! ๐
1๏ธโฃ CPU-bound tasks: Maximize your CPU! If you have 8 cores, use 8 threads to fully utilize them. More than that? You might face slowdowns due to context switching. ๐ฅ๏ธ
2๏ธโฃ Network-bound tasks: Here, CPU cores aren't the focus. It's all about handling multiple network operations concurrently. Async loops like asyncio work well, but sometimes threads are needed. Think requests to a server! ๐
๐ Mixing CPU and Network-bound tasks? Don't stick to just one thread pool. Have one for CPU tasks (sized by core number) and another for network tasks (sized by concurrency level). This approach prevents resource limit hits and ensures efficiency. ๐
๐ง Remember, each application has unique needs. Use these insights as a starting point for optimizing your Python data processing jobs. ๐ก
Happy coding! ๐จโ๐ป
#Python #ParallelProcessing #ThreadPools #DataProcessing #PerformanceOptimization
The python Africa community has sent an open letter to @ThePSF . We raise concerns over issues that have negatively affected our communities and many of us feel hurt. Please take a moment to understand where we are coming from: https://t.co/x0IY8X1jxp
๐ Exciting Update on "Mastering Python"! ๐
Hello, X community!
I'm thrilled to share that I'm nearing the completion of Chapter 3 of my upcoming book, "Mastering Python". This journey has been incredibly enriching, and I can't wait to bring this comprehensive guide to all of you.
But that's not all โ once Chapter 4 is finalized and the technical reviews are in, "Mastering Python" will be officially available for pre-order! This is a big milestone, and your support means the world to me.
Want to be the first to know when pre-orders go live? Sign up now to get notified! Plus, as a token of my appreciation, early subscribers will receive an exclusive sneak peek at Chapter 1. This is your chance to get a head start on mastering Python.
๐ Sign up here: https://t.co/5ormrfQrBQ
Thank you for being part of this exciting journey. Stay tuned for more updates and get ready to elevate your Python skills with "Mastering Python"!
#Python #Programming #BookLaunch #PreOrder #TechCommunity #developer #morioh #programmer #softwaredeveloper #computerscience #webdev #webdeveloper #webdevelopment #pythonprogramming #pythonquiz #ai #ml #machinelearning #datascience