Software on digikar's microblog

Performance Comparison of Stan on Intel 228V vs Apple Air M4

Fri, 28 Aug 2026 20:09:09 +0200

Apple Silicon is definitely impressive in benchmarks. For real world tasks, however, they recommend running testing against real applications. Now, I’m not running any rigorous tests. Just some informal tests and eye-balling the times.

The most intensive task I’ve been doing this month is building and running stan models. For those not in the know, Stan is a software for bayesian data analysis. The documentation is really good. It essentially provides a rich C-like DSL (minus the memory management) to write bayesian models. It then transpiles this to C++, compiles it, and then runs it. For small datasets, both compilation and runtime become comparable.

As against my earlier announcement about the use of macbook, I eventually returned to Linux as a found a good deal on a T14 Thinkpad with Intel 228V, which has very reasonable battery life to get me through full conference days and long journeys. But I have the M4 Macbook Air lying around that I don’t want to give, because I want to use it for occasionally testing my libraries on MacOS. Cloud-hosted solutions for MacOS cost on the order of 60 USD/month, which basically means you can get a full macbook in a few months. Very unaffordable.

But more recently, I found some time and peace of mind to dive into tunneling and set up a cloudflared tunnel to access my macbook through a public domain. This means I can now access my dormant macbook from anywhere even if I don’t have a public-facing static IP.

But is it any useful? Apple M4 packs a punch. Depending on the metric, it is about 25-50% faster than Intel 228V. I also noticed that while 228V has a peak of 4.5GHz, in reality, it only spends a minute or two above 3GHz, and often settles down to 2.5GHz. I don’t know how all the benchmark numbers compare against thermal throttlings. Benchmark numbers also don’t have strong cross-domain correlations.

So, the question then becomes, for my particular use case, which is compiling and running Stan Models, how do Intel 228V and Apple M4 compare? I also found an officially maintained stan-dev/performance-tests-cmdstan repository that I hope is representative of my usage. So, [depending on how lucky you are], the following should allow you to benchmark stan on your computer. As of this writing, I’m on commit b329e269 of the repository.

1
2


git clone --recursive https://github.com/stan-dev/performance-tests-cmdstan.git
python3 runPerformanceTests.py -j N -runj N stat_comp_benchmarks # N is the number of cores

You might want to prefix that last command with time which is what I do for obtaining the below numbers. They represent duration in minutes; thus lower the better. For whatever reason, Intel 228V did not throttle while running this workload. It consistently had at least one core above 4GHz.

Test \ Platform	Intel 228V	Apple M4 (Powersave)	Apple M4 (Full)
4 cores compile + run	2:25	4:10	2:07
1 core compile + run	2:40	3:55	1:57
4 cores run	0:35	2:10	1:05
1 core run	0:55	1:55	1:00
4 cores compile (diff)	1:50	2:00	1:07
1 core compile (diff)	1:45	2:00	0:57

Well, these were obtained by eye-balling about 2 runs each, just to check if they differ. Often, the difference between subsequent runs was less than 1 second. So, nothing rigorous, but nothing totally off either. It does seem that multiple cores has negligible impact on the runs. If anything, on MacOS, the multiple cores actually degraded the run! It’s also possible that difference in compilation times are due to the difference between g++ on Linux and clang++ on MacOS; I could not trivially get the system to compile with clang++. There could also be a bunch of different optimization flags that are not being used.

That said, it seems this is not the workload where I will get any significant benefits from Apple Silicon in the immediate future. There are similar reports and discussions here, here and here. Most of the reported gains seem to be from testing a 2018-era CPU to the M-series Apple Silicon :/. That’s not to dismiss the energy efficiency of Apple Silicon. M4 gets me about 2 conference days(!) on a single charge. Its standby time is also something that might put smartphones to shame. But at least for this particular workload, it doesn’t seem it’s a great fit.

How do I use AI (LLMs)?

Tue, 17 Feb 2026 00:28:43 +0100

Over the past few days, I have been asked the “How do I use AI?” question a few times. So, here’s an attempt at a common answer along with a few potential FAQs.

Do I think AI will replace us soon (= next 5 years)?

My own long term interest has been in the development of AI that can be at par with humans in certain ways. And to the extent it is not at par with humans, it will remain a topic of research. What certain ways? Answering that can be multiple research programs in itself! One aspect is causality, and perhaps, the most accessible resource is the first chapter of Judea Pearl’s The Book of Why.

AI looks great only at tasks you are bad at, because you don’t even (i) have the basic knowledge to judge it (ii) have not spent enough time to judge its output. Mere readability and plausible sounding is not understanding, remember?

Briefly, no, I do not think AI will replace us in the next 5 years. One reason is causality. But, there are multiple reasons.

Do I think companies will replace human employees because of AI?

This is a concern I share. To the extent that companies want to cut down costs and generate plausible looking progress, AI seems like a great deal, unfortunately. Crony capitalism at its best. I’d guess the only good way out is unionizing. Employee unions are also helpful in other ways beyond the employment threats that AI supposedly places.

In fact, I now fear, that before the apocalypse due to climate change, we will end up in a situation, where many critical softwares are written by AI which has no causal understanding or any of the other forms of understanding evolution has endowed us with. The critical softwares written by AI and unverified by humans will have many edge cases which would have been detected by human teams. These critical softwares will then trigger system failures affecting the lives of many. I can only hope there are enough safeguards in place before AI written softwares reach hospitals, medical devices, nuclear systems, power grids, flight systems, car softwares, and unimaginably many others. If you are using Windows, you might already have received a taste of things to come. Also see other posts like these.

Do I think AI is totally useless? Do I use AI? What do I use it for?

At its core, generative AI – the LLMs powering ChatGPT – are based on word/picture embeddings generated using “fill in the missing word/token/picture given the context”. In other words, these tools are amazing at finding statistical regularities in the text. And to some extent, these also approximate semantic understanding. They key word is some. Given a question or query, these tools can allow you to easily narrow down your search to a 100 documents in a million. Which of those documents are most relevant is for you to find out. In this sense, they are amazing.

In other words, I think AI works great as glorified search engines. However, a critical step is to not trust the output it generates, but, depending on the stakes involved, go and check the references it cites.

Many times the AI will say something that the citation does not say at all! Additionally, in an attempt to summarize the reference, it loses the nuance. For instance, on stackexchange websites, if an answer is bad, someone will point it out in the comments. Amongst two answers, if one answer has 900 upvotes and the other has 10, it’s more plausible that the one with the 900 upvotes is better in some way. (In what way? For that you might need to dig around more.) AI can lose out this context. It can also lose out the context that someone answered, someone countered, the answerer countered and clarified, and therefore, the answer is actually good and not bad as the first comment might make one think.

I’m also aghast to say, AI can sometimes provide references that even my supervisors might not have felt relevant. Instead of asking AI the question, “How do I …” or “What is the …”, ask it “Who has written an article or resource on how do I …” or “Who has written an article or resource on what is the …”. Then check out the references it cites, and use your own brains.

I don’t need to write repetitive emails often. And it is also acceptable (and preferred) to reply to emails briefly in one sentence. So, that, hasn’t been my common use.

I think there are better tools for spelling and grammar than general purpose LLMs. (Looks at Grammarly.) I still think it is better than nothing to ask them to review short sections of your work. Again, take their answers with skepticism. As you learn more about different kinds of writing, you will discover new rules of thumbs to keep in mind.

I use AI to generate code templates as well as for “function finding”: Quick, How do I remove whitespaces from the end of a string in Python? In Javascript? In Common Lisp?

Based on examples I see in the wild (eg. this in Julia, the above in Python, this in Common Lisp, I don’t trust AI to generate good code. And no, it is not a matter of giving more specific instructions – it is an open issue that coding agents do not obey your instructions. Perhaps, because they do not have any notion of an “instruction” at all. All they are generating is plausible looking text given the context. Additionally, writing code is not the hard part. Reading, understanding, maintaining, and writing code that is easy to read, understand and maintain is hard. Better to do it on your own than read through someone’s shit.

The developer of Claude Code – Anthropic itself – highlights that the benefits from AI might be just about 10% even while the impact on understanding and skill formation can be larger.

That said, given some possibilities highlighted above, I have also been experimenting with ollama and llama.cpp. I thought I wasn’t ever going to need more than 32GB RAM. Boy, I am wrong, I’m going to need 64GB RAM, particularly with modern chips coming with powerful GPUs and NPUs.

I also tried to think up a method by which AI can be made to respect the possibilities it is provided with. Unfortunately, it seems that unless one constrains those possibilities externally, there is no way to constrain them internally. But as soon as one imposes external constraints, one loses out on the composability that AI seemingly provides. In other words, human-like intelligence is an unsolved problem.

Lastly, Character AI can be surprisingly good for general chat as well as if you enjoy roleplays but haven’t formed your own roleplay-group. Still, they can get quite unreasonable at times.

Briefly, AI is useful for querying its vast knowledge base to find references that you can judge for yourself. And also for some creative tasks.

Bonus: Noam Chomsky: The False Promise of ChatGPT

The role of Open Source and Data Privacy for Security

Tue, 13 Jan 2026 18:08:32 +0100

Internet has been amazing. You send a message. And it reaches the other end of the Earth in seconds. Even more, you can communicate with live video and audio! It seems magical.

Yet it is not. All communication is transmission of energy. Energy travels through space – which may be occupied by matter or may not. Most space through which communications travel is openly available to everyone. The same space is used by multiple communicators. Does that mean everyone can access the messages I send over the internet?

That’s a scary thought. All your bank account details, passwords, private messages, your voice, your tone, your looks*, all travel over the internet. Are these all available to the public at large? (*You know now with generative AI, your voice and looks are sufficient to prepare videos as if you had done or said things you hadn’t!)

Unfortunately, in an unencrypted world, it would be true that everyone can see everyone’s messages. You can read malicious users’ messages, and they can read yours. Furthermore, what even determines that a message your messaging app says was sent by your friend was really sent by your friend? And not someone else disguised as your friend?

Humans, some of them, are smart. These problems have been known since the dawn of the internet in the 20th century, possibly before. Solutions have been devised against them. Today, for most communication problems in the general, we know that

If communications satisfy certain assumptions, then certain particular security benefits are guaranteed.

In other words,

If we build communication softwares that satisfy certain assumptions, then we get certain particular security benefits.

These benefits include:

Knowing that the message your messaging app says was sent by a friend was really sent by your friend’s device.
Ensuring that only you and your friend can read messages sent to each other. That is, your messages are inaccessible to anyone else.
… and many more …

In general, the “friend” may be any other communication end-point, such as your bank, where security is just as important if not more. However,

How do we know that communication softwares actually satisfy the assumptions required for security?

Three methods:

We trust the word of the person or institute who developed the software and make profits from the software. In general, the greater motive here is making monetary profits, since that is a very essential aspect of livelihood of the person or institute.
We trust the word of the person or institute who is motivated to study and develop the software and ensure the software satisfies the assumptions required for security. In general, monetary profit as a motive is either absent or very much secondary.
We study the workings of the software ourselves.

Both 2nd and 3rd methods require that how the software works is publicly accessible. This means the source code (but also the infrastructure details) are publicly available. In other words,

Trusting the claim that a particular software is secure becomes easier when the workings of the software are publicly known, that is, it is open source.

How to find open source softwares?

An easy method to find the open source alternative for a software is to simply google "open source alternative to software that interests you". You can also append the search by "site:reddit.com" or "site:stackexchange.com" or you can also consult one of your tech-geek friends or family-members.

It is trivial to see that between two people or institutes:

One who is motivated by monetary profits
Another who is motivated by software security

The second category of people or institutes would end up with more secure softwares in any particular class of softwares in the longer run. In other words, in the longer run, the first category of people or institutes are likely to make mistakes even if, besides making money, they too want to make their softwares secure. This is simply because, for the software developers, motivation for software security drives learning and brings in the time to spend on the software and make changes that may go against the monetary motivation.

That people and software developers can make mistakes brings us to the second point of this blog post. Your data includes - your name, phone numbers, email addresses, passwords, bank account numbers, people or institutes you send messages to or receive messages from, what messages you send, which websites you access, how frequently, at what time, for how long, your phone location, when do you leave your home, office or visit a friend, and much more.

You would WISH that the banks, institutes, and messaging softwares that have access to your data are actually storing them securely, in a manner that does not permit thieves and malicious actors from misusing your data.

This is also called a data breach. In 2025 alone, there were at least 5 data breaches with banking softwares resulting in millions of individuals losing their banking details to malicious hackers. It is not just banking softwares. Data breaches can happen with car companies, education providers, beer companies, sim providers, luxury brands, and pretty much everything that is connected to internet.

If the softwares were open source, it would be easier to ensure security. However, the next best thing to open source might be data privacy. How do you prevent your data from being misused?

To prevent your data from being misused, you restrict your data from being accessible. That is, you keep your data private.

How do you make your data more private? This is a bit like asking How do I make myself more healthy? There’s no simple answer, although one can try some guidelines! See here or here for a detailed guide. Here’s an attempt at some guidelines:

Reduce reliance on closed source softwares for information processing. This includes Microsoft Word, Powerpoint, Excel, as well as Google Docs, Slides, and Sheets, and also Apple. Of course, you may be forced to use these apps because your colleague uses them. In that case, either convince them to shift to open source alternatives or use the open source alternatives for your personal use cases. Simple alternatives include markdown, latex, and open document formats.
Try out Kubuntu on newer laptops or Xubuntu to renew old laptops. Be sure to use the “Long-Term Support (LTS)” versions, currently 24.04. You can also try it as a virtual machine.
Switch to secure open-source messaging softwares:
- Signal or Molly in addition to Whatsapp, Messenger, Instagram.
- Jitsi in addition to Zoom or Google Meet or Teams
Use password managers like Bitwarden.
… You are motivated enough. Read one of those detailed guides now :)

To sum up

Open source aids security. You can find open source alternatives by simply googling “open source alternatives to <app you want>”. Optionally, you can append the search with “site:reddit.com” or “site:stackexchange.com”.
Privacy aids security. Privacy is a bit like moving towards a healthy lifestyle. There are lots of small changes in the way. See this or this for a detailed guide.