Looking forward to speaking at the Warsaw IT Days 2026 Conference (@WDI_conference) this week! 🚀🇵🇱
I'll be highlighting how agent-driven architectures are reshaping the way LLMs interact with enterprise data.
Check out the agenda & grab your tickets: https://t.co/ACrt56Qg7d
Finally shipped something I’ve been building for a while now 🚀
Introducing NDArray PHP - high performance n-dimensional arrays for PHP, powered by Rust via FFI.
It features zero copy views, has a memory-safe Rust core, single FFI call per operation and has a Numpy-like API that just works. And of course, it’s blazing fast!!
This is a pretty low level library, and is the first among other utilities I aim to publish to improve ML tooling for PHP. The plan is to build an entire ecosystem -Improve TransformersPHP, new ONNX runtime, audio processing extensions, image analysis utils and more higher level ML abstractions for PHP with this foundation.
So, If you’re working with numerical data in PHP, or you’re just curious about Rust/PHP FFI, check it out.
Repo and documentation links in the comments ⬇️
"I want to become a Reverend Sister" 💀
For what it's worth, "for of all sad words of tongue or pen, The saddest are these: 'It might have been!'" - John Greenleaf Whittier
Just released Tokenizers PHP v1.0.0 🎉
For those that have used TransformersPHP, you’ll know tokenization was a part of the pipeline when performing tasks (before actually running the model). Well, I extracted it into its own standalone library and improved it to be wayyy better now.
Lightweight. Minimal dependencies. Pure PHP. No FFI or binaries. Thoroughly tested across BERT, GPT, Llama, Qwen, Gemma, and more.
So what can you actually do with it?
- Count tokens before hitting LLM APIs (save money)
- Prep text for vector embeddings & semantic search
- Split documents into chunks for RAG systems
- Preprocess data for ML classification
- Analyze if your prompts fit different model context windows
If you're doing anything AI/ML related in PHP, you probably need a tokenizer at some point. This makes it dead simple.
I included real examples too so you're not left guessing how to use it.
Link below 👇
There's only one encryption that's proven to be 100% unbreakable, and it’s not AES-256. It’s the OTP.
No, not the SMS password you get for 2FA. The "real OG" OTP: One-time Pad.
I wrote a breakdown on how "Perfect Secrecy" actually works:
https://t.co/a8El6ZdExa
My heart goes out to anyone in this situation. The economy is hard, opportunities are scarce.
There were times I felt like this in my career.
Try not to feel paralyzed. You can always do something. Action is better than inaction.
But first, what you need is clarity. Give yourself time to be clear on what you (should) want. Don't let social media pressure you, not everyone has it figured out.
Talk to people if you have to. Sometimes clarity takes time. So give it time, if you can.
But when clarity comes, execute like mad. Be disciplined, be consistent.
You also need a community of quality people around you who can provide emotional support. Don't do it alone.
Honored to be a keynote speaker at #FlaskconAfrica
I'll be talking about how the Flask framework is fueling a wave of innovation across the continent. This is based on my firsthand experience in African Fintech and building leading ecosystem open source projects.
Don't miss it
Join us on November 1st for the Port Harcourt DeSci Meetup!
DeSci (Decentralized Science) uses Web3 principles like blockchain, transparency, and tokenization to revolutionize research funding, publishing, and collaboration.
Applied for principal engineer at Microsoft when I was in 200lvl.
Didn’t get it ofcourse but the audacity though😂 >>>>>
I should find it again
🎉 I did a thing!
Earlier this month an expository research paper I have been writing was peer-reviewed, accepted and published. 🤲🏽
“Balancing CAP: Achieving Eventual Consistency in Distributed Systems using CRDTs and KDHT”
IRE Journals, Vol 9 Issue 2 Aug-2025, 1130-1149.