Martin Perez

Martin Perez

@martin.arume.cloud

Semi-retired til next thing. 100% certified human being. Also biker. I was once at twitter @mpermar. Then decided to check this AT protocol thingie.

81 followers102 following128 posts

Imagine. Your house construction works are about to finish. You visit the architect and she tells you: "The house is pretty much done. We understand how it was done but I don't know all details, neither builders do. But we verified it and it will be fine." Welcome to AI-age software engineering.

Martin Perez reposted

Replying to @igalia.com

Twenty-five years ago ten people set out to find whether a company owned by its workers could do serious engineering entirely in the open. We think the answer is in the commit logs. All of it is upstream, where anyone can use it. Here is to the next twenty-five.

On another garden update, this season I made a funny mistake. I actually asked my shopkeeper to pick me some peanut plants which I had seen around. But when he went to the nursery shelves picked something else... This is the lucky outcome: 50 kg of "peanuts" 😅 (not bad for a 2 EUR investment)

AI contrary take: Established software companies will be strongest than ever. The thesis is simple. An established vendor has the MOAT, the contacts and the relations while AI has wiped competitors' competitive advantages like speed, research or talent.

Similarly, AI workhorses don't avail of that MOAT, contacts or relations. It is easy to edit an image with an AI prompt. It's not that simple to become Adobe the enterprise. Adobe sure got a hole in their revenue but in exchange AI has pretty much killed Figma and any other competition.

Zwift buying Rouvy feels a lot like OpenAI buying Anthropic. I hope they do not kill it. I really do not fancy the artificial virtual experience at all.

Martin Perez reposted

There are no technical or compliance reasons to double the size of symmetric keys in response to the threat of quantum computers. This common misunderstanding of Grover's algorithm risks wasting limited resources that should go towards deploying actually urgent post-quantum algorithms.

words.filippo.ioQuantum Computers Are Not a Threat to 128-bit Symmetric KeysThere is no need to update symmetric key sizes as part of the post-quantum transition, due to the details of how Grover's algorithm scales. Most authorities agree.

I've been thinking a lot about how AI could have made our startup successful 15 years ago. Answer: No. It could not. Network, contacts, influence, reach, team, human persistence, long sales cycles, ... If anything, AI would have brought us tons of competitors more. AI won't kill MOATs.

Someday, in a not so distant future, mobile phones will include a light LLM-powered phone call assistant that will handle marketing calls and make support calls efficiently in the background. For those building. That's a product I would buy. For us, consumers. That's the final revenge.

Martin Perez reposted

Amongst other things, there's an open source software supply chain story here. This Android library with 174 stars and one maintainer has taken down many banks in India (in 2023), and "80% of the food delivery market" in Brazil (right now).

Martin Perez reposted

The new @golang.org 1.24.4 is already available in the Bitnami catalog — just 2 hours after the upstream release! ⚡️ This quick release helps ensure important vulnerabilities are addressed without delay. Need FedRAMP compliance with FIPS, STIG? Check www.vmware.com/products/app...

vmware.comVMware Tanzu Application Catalog | VMware TanzuVMware Tanzu Application Catalog is a curated image database of continuously maintained and trusted, pre-packaged application components.
Go@golang.org

🎆 Go 1.24.4 and 1.23.10 are released! 🔐 Security: Includes security fixes for CVE-2025-4673, CVE-2025-0913, and CVE-2025-22874 in net/http, os, and crypto/x509. 📰 Announcement: groups.google.com/g/golang-ann... 📦 Download: go.dev/dl/#go1.24.4 #golang

I've been vibe coding some data processing Python and heck, the amount of crappy code that we will have to manage during the next 5 to 10 years is going to be insane. Now, once these bots are polished and we have them well trained... Well. It's game over. There will be software engs but other type.

There is some awkward satisfaction in correcting your AI coding peer. It feels good because you still can see yourself as smarter, but it's just an optical illusion. You're doing the captcha thing. We all are.