Marco Zocca
@ocramz.bsky.social
ML, λ • language and the machines that understand it • https://ocramz.github.io
Pinned
Marco Zocca
@ocramz.bsky.social
· Jul 4
CERN for frontier AI >>>
the entire information infrastructure of the EU can be turned off by the US and they are c++pping themselves over Chinese buses
OK, so now the UK, Norway, Australia and Denmark are all investigating buses made by China's Yutong Group for security risks.
www.ft.com/content/07ec...
www.ft.com/content/07ec...
UK investigates whether buses made in China can be turned off from afar
Investigation comes after Norway found Yutong vehicles could be ‘stopped or rendered inoperable’ by Chinese company
www.ft.com
November 10, 2025 at 11:33 AM
the entire information infrastructure of the EU can be turned off by the US and they are c++pping themselves over Chinese buses
a panoramic collage of waves rolling in trades time for space
November 10, 2025 at 8:37 AM
a panoramic collage of waves rolling in trades time for space
trying to get typed output from LLMs is about as fun as chewing tinfoil
nonetheless, Qwen2.5 is pretty useful
nonetheless, Qwen2.5 is pretty useful
November 9, 2025 at 12:59 AM
trying to get typed output from LLMs is about as fun as chewing tinfoil
nonetheless, Qwen2.5 is pretty useful
nonetheless, Qwen2.5 is pretty useful
I don't think the parallel holds at all.
Not only you can unplug and reset an LM, but you can engineer its knowledge and steer it more or less at will, repeatably and without memory/histeresis.
I agree a precautionary principle should hold in general, but the particular differences matter a lot.
Not only you can unplug and reset an LM, but you can engineer its knowledge and steer it more or less at will, repeatably and without memory/histeresis.
I agree a precautionary principle should hold in general, but the particular differences matter a lot.
When I started my doctorate, I had to complete the standard research ethics training. Particularly with prisoner populations, I saw many parallels to AI. #ai #artificialintelligence #llms #largelanguagemodels #aiethics
When We Decide Who Can Feel
Should AI be protected by ethical research guidelines?
open.substack.com
November 8, 2025 at 3:34 AM
I don't think the parallel holds at all.
Not only you can unplug and reset an LM, but you can engineer its knowledge and steer it more or less at will, repeatably and without memory/histeresis.
I agree a precautionary principle should hold in general, but the particular differences matter a lot.
Not only you can unplug and reset an LM, but you can engineer its knowledge and steer it more or less at will, repeatably and without memory/histeresis.
I agree a precautionary principle should hold in general, but the particular differences matter a lot.
gods forbid i consider chatgpt as having legal personhood or indeed the framing of the 1A to be as universal as USians make it to be
November 8, 2025 at 3:26 AM
gods forbid i consider chatgpt as having legal personhood or indeed the framing of the 1A to be as universal as USians make it to be
even your favorite author writes a platitude sometimes 🥶
Like all walls it was ambiguous, two-faced. What was inside it and what was outside it depended upon which side of it you were on.
November 3, 2025 at 2:45 PM
even your favorite author writes a platitude sometimes 🥶
having deep dives on stuff I know close to nothing about, getting what looks like well structured sensible answers I can study, with examples that make sense in turn
I’m still not close to being able to explain the output to someone else, but I guess this counts as progress?
I’m still not close to being able to explain the output to someone else, but I guess this counts as progress?
October 29, 2025 at 5:59 AM
having deep dives on stuff I know close to nothing about, getting what looks like well structured sensible answers I can study, with examples that make sense in turn
I’m still not close to being able to explain the output to someone else, but I guess this counts as progress?
I’m still not close to being able to explain the output to someone else, but I guess this counts as progress?
What’s the sound made by 1000 microsoft VPs igniting spontaneously?
"I DON'T NEED YOU TO FUCKING REWRITE WHAT I'VE JUST WRITTEN!"
October 29, 2025 at 3:43 AM
What’s the sound made by 1000 microsoft VPs igniting spontaneously?
For once, there’s an easy fix: port your backend to Haskell, add stm and go on with your day : hackage.haskell.org/package/stm
October 28, 2025 at 12:16 PM
For once, there’s an easy fix: port your backend to Haskell, add stm and go on with your day : hackage.haskell.org/package/stm
Reposted by Marco Zocca
I’m in a wifi-starved situation and rely on mobile data for everything for a few days
I think I maxed out my plan and the mobile provider is throttling my traffic at the application level. Like, no youtube but text apps work? Github actions showing blank results? Wtf
I think I maxed out my plan and the mobile provider is throttling my traffic at the application level. Like, no youtube but text apps work? Github actions showing blank results? Wtf
October 28, 2025 at 9:33 AM
I’m in a wifi-starved situation and rely on mobile data for everything for a few days
I think I maxed out my plan and the mobile provider is throttling my traffic at the application level. Like, no youtube but text apps work? Github actions showing blank results? Wtf
I think I maxed out my plan and the mobile provider is throttling my traffic at the application level. Like, no youtube but text apps work? Github actions showing blank results? Wtf
Slow is smooth, smooth is fast
Automated Guided Vehicles at Long Beach Container Terminal manufactured by Terex Port Solutions.
October 27, 2025 at 6:24 PM
Slow is smooth, smooth is fast
Good stuff. Scientific software and in particular optimization codes are everywhere, yet the magic sauce is often proprietary and found heuristically.
Analysis such as this one is net progress for the field.
Analysis such as this one is net progress for the field.
The simplex algorithm is super efficient. 80 years of experience says it runs in linear time. Nobody can explain _why_ it is so fast.
We invented a new algorithm analysis framework to find out.
We invented a new algorithm analysis framework to find out.
Beyond Smoothed Analysis: Analyzing the Simplex Method by the Book
Narrowing the gap between theory and practice is a longstanding goal of the algorithm analysis community. To further progress our understanding of how algorithms work in practice, we propose a new alg...
arxiv.org
October 27, 2025 at 10:12 AM
Good stuff. Scientific software and in particular optimization codes are everywhere, yet the magic sauce is often proprietary and found heuristically.
Analysis such as this one is net progress for the field.
Analysis such as this one is net progress for the field.
Slop hogs, riddle me this: the Sonnet 4.5 I use with my work account is over the top enthusiastic, the one with my personal account plain and factual. Code quality and quantity per prompt roughly comparable.
verdict:
Claude is smart and even too chatty. Fills hundreds of lines in a go as well as self-celebratory SUCCESS.md logs.
4o : "here's that bs you asked. Let me know if you need 10 more lines."
Claude is smart and even too chatty. Fills hundreds of lines in a go as well as self-celebratory SUCCESS.md logs.
4o : "here's that bs you asked. Let me know if you need 10 more lines."
let's see if this agentic coding is any good
October 27, 2025 at 10:08 AM
Slop hogs, riddle me this: the Sonnet 4.5 I use with my work account is over the top enthusiastic, the one with my personal account plain and factual. Code quality and quantity per prompt roughly comparable.
Reposted by Marco Zocca
Are there any posts about how OpenBLAS really absolutely was the big bang moment for what is now the explosion of python and data science and the way we see the world running so much AI and ML all came from that shit getting released and unseating the intel kernel math library
October 26, 2025 at 4:42 AM
Are there any posts about how OpenBLAS really absolutely was the big bang moment for what is now the explosion of python and data science and the way we see the world running so much AI and ML all came from that shit getting released and unseating the intel kernel math library
Codespaces + agent mode = prototype all the things
October 25, 2025 at 6:24 AM
Codespaces + agent mode = prototype all the things
Reposted by Marco Zocca
My new article on the illusion of AI productivity.
We often feel more productive with AI, but are we really?
Read here newbeelearn.com/blog/ai-prod...
#AI #Productivity #FutureOfWork #DigitalTransformation #AItools
We often feel more productive with AI, but are we really?
Read here newbeelearn.com/blog/ai-prod...
#AI #Productivity #FutureOfWork #DigitalTransformation #AItools
Does productivity really increases with ai or it just feels like it?
story of how the tool's save frame from video was useful for landing page
newbeelearn.com
October 25, 2025 at 4:52 AM
My new article on the illusion of AI productivity.
We often feel more productive with AI, but are we really?
Read here newbeelearn.com/blog/ai-prod...
#AI #Productivity #FutureOfWork #DigitalTransformation #AItools
We often feel more productive with AI, but are we really?
Read here newbeelearn.com/blog/ai-prod...
#AI #Productivity #FutureOfWork #DigitalTransformation #AItools
You can actually feed the article TeX source to copilot as a reference and it will produce code and examples. Neat
Taking Haiku 4.5 for a spin, this time with a scientific computing codebase.
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
October 24, 2025 at 1:21 PM
You can actually feed the article TeX source to copilot as a reference and it will produce code and examples. Neat
Reposted by Marco Zocca
Computer science courses that don't exist, but should (2015) [Discussion]
Computer science courses that don't exist, but should (2015)
Computer science courses that don't exist, but should (2015)
prog21.dadgum.com
October 24, 2025 at 5:52 AM
Computer science courses that don't exist, but should (2015) [Discussion]
Can confirm, for one-off tools the New Way is an incredible force multiplier.
In org settings, it shifts the power balance between waiting and doing.
The problems as always start when that little prototype becomes load-bearing in any sense (must be maintained, other people depend on it etc)
In org settings, it shifts the power balance between waiting and doing.
The problems as always start when that little prototype becomes load-bearing in any sense (must be maintained, other people depend on it etc)
One of our students vibed an extremely useful map visualizer in a few hours. We would just never have done it before because it would have been a week of work. We are in a new world.
October 24, 2025 at 5:54 AM
Can confirm, for one-off tools the New Way is an incredible force multiplier.
In org settings, it shifts the power balance between waiting and doing.
The problems as always start when that little prototype becomes load-bearing in any sense (must be maintained, other people depend on it etc)
In org settings, it shifts the power balance between waiting and doing.
The problems as always start when that little prototype becomes load-bearing in any sense (must be maintained, other people depend on it etc)
Reposted by Marco Zocca
What Happened To Running What You Wanted On Your Own Machine? | Hackaday https://hackaday.com/2025/10/22/what-happened-to-running-what-you-wanted-on-your-own-machine/
What Happened To Running What You Wanted On Your Own Machine?
When the microcomputer first landed in homes some forty years ago, it came with a simple freedom—you could run whatever software you could get your hands on. Floppy disk from a friend? Pop it in. Shareware demo downloaded from a BBS? Go ahead! Dodgy code you wrote yourself at 2 AM? Absolutely. The computer you bought was yours. It would run whatever you told it to run, and ask no questions.
Today, that freedom is dying. What’s worse, is it’s happening so gradually that most people haven’t noticed we’re already halfway into the coffin.
## News? Pegged.
_There are always security risks when running code from untrusted sources. The stakes are higher these days when our computers are the gateways to our personal and financial lives._
The latest broadside fired in the war against platform freedom has been fired. Google recently announced new upcoming restrictions on APK installations. Starting in 2026, Google will tightening the screws on sideloading, making it increasingly difficult to install applications that haven’t been blessed by the Play Store’s approval process. It’s being sold as a security measure, but it will make it far more difficult for users to run apps outside the official ecosystem. There is a security argument to be made, of course, because suspect code can cause all kinds of havoc on a device loaded with a user’s personal data. At the same time, security concerns have a funny way of aligning perfectly with ulterior corporate motives.
It’s a change in tack for Google, which has always had the more permissive approach to its smartphone platform. Contrast it to Apple, which has sold the iPhone as a fully locked-down device since day one. The former company said that if you own your phone, you could do what you want with it. Now, it seems Google is changing its mind ever so slightly about that. There will still be workarounds, like signing up as an Android developer and giving all your personal ID to Google, but it’s a loss to freedom whichever way you look at it.
## Beginnings
Sony put a great deal of engineering into the PlayStation to ensure it would only read Sony-approved discs. Modchips sprung up as a way to get around that problem, albeit primarily so owners could play cheaper pirated games. Credit: Libreleah, CC BY-SA 4.0,
The walled garden concept didn’t start with smartphones. Indeed, video game consoles were a bit of a trailblazer in this space, with manufacturers taking this approach decades ago. The moment gaming became genuinely profitable, console manufacturers realized they could control their entire ecosystem. Proprietary formats, region systems, and lockout chips were all valid ways to ensure companies could levy hefty licensing fees from developers. They locked down their hardware tighter than a bank vault, and they did it for one simple reason—money. As long as the manufacturer could ensure the console wouldn’t run unapproved games, developers would have to give them a kickback for every unit sold.
By and large, the market accepted this. Consoles were single-purpose entertainment machines. Nobody expected to run their own software on a Nintendo, after all. The deal was simple—you bought a console from whichever company, and it would only play whatever they said was okay. The vast majority of consumers didn’t care about the specifics. As long as the console in question had a decent library, few would complain.
_Nintendo created the 10NES copy protection system to ensure its systems would only play games approved by the company itself, in an attempt to exert quality control after the 1983 North American video game crash. Credit: Evan-Amos, public domain_
There was always an underground—adapters to work around region locks, and bootleg games that relied on various hacks—with varying popularity over the years. Often, it was high prices that drove this innovation—think of the many PlayStation mod chips sold to play games off burnt CDs to avoid paying retail.
At the time, this approach largely stayed within the console gaming world. It didn’t spread to actual computers because computers were tools. You didn’t buy a PC to consume content someone else curated for you. You bought it to do whatever you wanted—write a novel, make a spreadsheet, play games, create music, or waste time on weird hobby projects. The openness wasn’t a bug, or even something anybody really thought about. It was just how computers _were._ It wasn’t just a PC thing, either—every computer on the market let you run what you wanted! It wasn’t just desktops and laptops, either; the nascent tablets and PDAs of the 1990s operated in just the same way.
Then came the iPhone, and with it, the App Store. Apple took the locked-down model and applied it to a computer you carry in your pocket. The promise was that you’d only get apps that were approved by Apple, with the implicit guarantee of a certain level of quality and functionality.
Apple is credited with pioneering the modern smartphone, and in turn, the walled garden that is the App Store. Credit: Apple
It was a bold move, and one that raised eyebrows among developers and technology commentators. But it worked. Consumers loved having access to a library of clean and functional apps, built right into the device. Meanwhile, they didn’t really care that they couldn’t run whatever kooky app some random on the Internet had dreamed up.
Apple sold the walled garden as a feature. It wasn’t ashamed or hiding the fact—it was proud of it. It promised apps with no viruses and no risks; a place where everything was curated and safe. The iPhone’s locked-down nature wasn’t a restriction; it was a selling point.
But it also meant Apple controlled everything. Every app paid Apple’s tax, and every update needed Apple’s permission. You couldn’t run software Apple didn’t approve, full stop. You might have paid for the device in your pocket, but you had no right to run what you wanted on it. Someone in Cupertino had the final say over that, not you.
When Android arrived on the scene, it offered the complete opposite concept to Apple’s control. It was open source, and based on Linux. You could load your own apps, install your own ROMs and even get root access to your device if you wanted. For a certain kind of user, that was appealing. Android would still offer an application catalogue of its own, curated by Google, but there was nothing stopping you just downloading other apps off the web, or running your own code.
Sadly, over the years, Android has been steadily walking back that openness. The justifications are always reasonable on their face. Security updates need to be mandatory because users are terrible at remembering to update. Sideloading apps need to come with warnings because users will absolutely install malware if you let them just click a button. Root access is too dangerous because it puts the security of the whole system and other apps at risk. But inch by inch, it gets harder to run what you want on the device you paid for.
## **Windows Watches and Waits**
The walled garden has since become a contagion, with platforms outside the smartphone space considering the tantalizing possibilities of locking down. Microsoft has been testing the waters with the Microsoft Store for years now, with mixed results. Windows 10 tried to push it, and Windows 11 is trying harder. The store apps are supposedly more secure, sandboxed, easier to manage, and straightforward to install with the click of a button.
Microsoft has tried multiple times to sell versions of Windows that are locked to exclusively run apps from the Microsoft Store. Thus far, these attempts have been commercial failures.
Microsoft hasn’t pulled the trigger on fully locking down Windows. It’s flirted with the idea, but has seen little success. Windows RT and Windows 10 S were both locked to only run software signed by Microsoft—each found few takers. Desktop Windows remains stubbornly open, capable of running whatever executable you throw at it, even if it throws up a few more dialog boxes and question marks with every installer you run these days.
How long can this last? One hopes a great while yet. A great deal of users still expect a _computer—_ a proper one, like a laptop or desktop—to run whatever mad thing they tell it to. However, there is an increasing userbase whose first experience of computing was in these locked-down tablet and smartphone environments. They aren’t so demanding about little things like proper filesystem access or the ability to run unsigned code. They might not blink if that goes away.
For now, desktop computing has the benefit of decades of tradition built in to it. Professional software, development tools, and specialized applications all depend on the ability to install whatever you need. Locking that down would break too many workflows for too many important customers. Masses of scientific users would flee to Linux the moment their obscure datalogger software couldn’t afford an official license to run on Windows;. Industrial users would baulk at having to rely on a clumsy Microsoft application store when bringing up new production lines.
Apple had the benefit that it was launching a new platform with the iPhone; one for which there were minimal expectations. In comparison, Microsoft would be climbing an almighty mountain to make the same move on the PC, where the culture is already so established. Apple could theoretically make moves in that direction with OS X and people would be perhaps less surprised, but it would still be company making a major shift when it comes to customer expectations of the product.
Here’s what bothers me most: we’re losing the idea that you can just try things with computers. That you can experiment. That you can learn by doing. That you can take a risk on some weird little program someone made in their spare time. All that goes away with the walled garden. Your neighbour can’t just whip up some fun gadget and share it with you without signing up for an SDK and paying developer fees. Your obscure game community can’t just write mods and share content because everything’s locked down. So much creativity gets squashed before it even hits the drawing board because it’s just not feasible to do it.
It’s hard to know how to fight this battle. So much ground has been lost already, and big companies are reluctant to listen to the esoteric wishers of the hackers and makers that actually care about the freedom to squirt _whatever_ through their own CPUs. Ultimately, though, you can still vote with your wallet. Don’t let Personal Computing become Consumer Computing, where you’re only allowed to run code that paid the corporate toll. Make sure the computers you’re paying for are doing what you want, not just what the executives approved of for their own gain. It’s your computer, it should run what you want it to!
hackaday.com
October 22, 2025 at 11:33 PM
What Happened To Running What You Wanted On Your Own Machine? | Hackaday https://hackaday.com/2025/10/22/what-happened-to-running-what-you-wanted-on-your-own-machine/
Taking Haiku 4.5 for a spin, this time with a scientific computing codebase.
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
October 22, 2025 at 9:37 AM
Taking Haiku 4.5 for a spin, this time with a scientific computing codebase.
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
Is there something like an “arXiv skill” for LLMs, so they can read up on research methods on the fly?
Reposted by Marco Zocca
I'm of the belief that all robots should have planters for heads merely because it'd be nice, it'd add a fun secondary movement to their walk, and then different robots could have different plants as a sort of "hair style"
October 20, 2025 at 3:33 PM
I'm of the belief that all robots should have planters for heads merely because it'd be nice, it'd add a fun secondary movement to their walk, and then different robots could have different plants as a sort of "hair style"