Weekly Tech - April 28th of 2025

What is CUDA? The web almanac for 2024, what it takes to be compatible with Postgres, a bit of Wasmer and how to host a blog on a Nintendo Wii, yes, this week is intense, ready to start?

Hello everyone! I hope you’re all doing well.

It’s a new week, continuing the momentum from last week with a focus on general knowledge and some exciting developments in the cloud. Just a quick reminder that I’ll be on leave next week, so the schedule might be shifted slightly.

Anyway, let’s get started!

---

Common knowledge

Right, we’re kicking things off with a deep dive into general knowledge, and specifically, an explanation of what CUDA is. We’ve also uncovered an optimization for asset caching through Shared Dictionary Compression and finally, we’ve discovered the web almanac – particularly its report for 2024.

• What is CUDA? - Youtube

• Interpreters vs JIT compilation in Javascript engines

• What is Latent Semantic Indexing and how to use it?

• Optimizing performances thanks to Shared Dictionary Compression - Youtube

• The 2024 state of the web report

• Was exposing SQL to end developers an error? The debate is open

---

Database

Just a slight detour into the database area to explore the prerequisites for making a DB ‘Postgres-compatible’ – and the pitfalls to watch out for.

• What is takes to be PostgreSQL compatible

---

Cloud computing

Let’s shift our focus back to the cloud – we’ve discovered Wasmer version 6.0 (with PHP optimizations) and a real gem – a post from Bearstech’s bears (definitely worth following them) about hosting a blog on a Nintendo Wii!

Seriously, it’s possible. The article doesn’t have any comments, so just read it.

• Stop using Docker like its 2015

• Wasmer 6.0 is out and it's fast, blazing fast

• Could a Nintendo Wii expose a blog? Of course

---

Machine learning

We're veering towards machine learning to discover A2A, Google’s 'protocol' to simplify agent interactions (basically, it's like RPC), and Supermemory MCP, MCP to use Supermemory (which has evolved quite a bit since its v0.1).

• A list of resources / applications built on top of LLMs

• Bitnet.cpp, the inference framework for 1-bit LLMs by Microsoft

• A2A, the Google "solution" for agents communications

• The SuperMemory MCP

• n8n, an automation platform with a bit of "magic" from IA

• Could economic tasks be performed by "IA"? Kind of

---

PHP

Finally, let’s wrap up with PHP and NeuronAI, who’ve explained the concept of the ‘context window’ – absolutely essential for your LLMs/models (so they can truly ‘understand’ and ‘take into account’ the content).

• The context window of AI agents in NeuronAI

• Is it hard to add new features in PHP? Not really

• Could PHP be extremely fast? Yep

So, for this week’s update, feel free to leave a comment or like if you’re feeling it! On my end, I’ll see you in two weeks for another update.

In the meantime, have a wonderful weekend and a very good week.

See you soon!