Baba

Founder @amazethu. Less is More

[Baba @babat00nday]

As mentioned earlier, amaspeak is officially in Alpha. 🚀

I am looking for builders and early learners to help us shape this foundation.

Come help us build this right: https://t.co/6wnUJZ7opn

[Baba @babat00nday]

African languages deserve world-class technology.

For too long, the resources available for learning our languages haven't matched the complexity and beauty of the cultures they represent.

That is why I am building amaspeak and it is officially in Alpha. https://t.co/egYDEYXTt8

[Baba @babat00nday]

General apps are great, but conversational fluency in languages like Yoruba, Swahili, and many more requires a specialized approach.

I am crafting a new kind of experience to help the diaspora finally speak their mother tongue and connect with family.

[anshuman @athleticKoder]

She dumped me last night.

Not because I don't listen.
Not because I'm always on my phone.
Not even because I forgot our anniversary (twice).

But because,

in her exact words:

"You only pay attention to the parts of what I say that you think are important."

I stared at her for a moment and realized...

She just perfectly described the attention mechanism in transformers.

Turns out I wasn't being a bad boyfriend. I was being mathematically optimal.

See, in conversations (and transformers), you don't give equal weight to every word. Some words matter more for understanding context. Attention figures out exactly HOW important each word should be.

Here's the beautiful math:

Attention(Q, K, V) = softmax(QK^T / √d_k)V

Breaking it down:

Q (Query): "What am I looking for?"
K (Key): "What info is available?"
V (Value): "What is that info?"
d_k: Key dimension (for scaling)

Think library analogy:

You have a question (Query). Books have titles (Keys) and content (Values). Attention finds which books are most relevant.

Step-by-step with "The cat sat on the mat":

Step 1: Create Q, K, VEach word → three vectors via learned matrices W_Q, W_K, W_V
For "cat":

Query: "What should I attend to when processing 'cat'?"
Key: "I am 'cat'"
Value: "Here's cat info"

Step 2: Calculate scoresQK^T = how much each word should attend to others

Processing "sat"? High similarity with "cat" (cats sit) and "mat" (where sitting happens).

Step 3: Scale by √d_kPrevents dot products from getting too large, keeps softmax balanced.

Step 4: SoftmaxConverts scores to probabilities:

"cat": 0.4 (subject)
"sat": 0.3 (action)
"mat": 0.2 (location)
"on": 0.1 (preposition)
"the": 0.1 (article)

Step 5: Weight valuesMultiply each word's value by attention weight, sum up. Now "sat" knows it's most related to "cat" and "mat".

Multi-Head Magic:Transformers do this multiple times in parallel:

Head 1: Subject-verb relationships
Head 2: Spatial ("on", "in", "under")
Head 3: Temporal ("before", "after")
Head 4: Semantic similarity

Each head learns different relationship types.

Why This Changed Everything:

Before: RNNs = reading with flashlight (one word at a time, forget the beginning)

After: Attention = floodlights on entire sentence with dimmer switches
This is why ChatGPT can:

Remember 50 messages ago

Know "it" refers to something specific
Understand "bank" = money vs river based on context
The Kicker:Models learn these patterns from data alone. Nobody programmed grammar rules. It figured out language structure just by predicting next words.
Attention is how AI learned to read between the lines.

Just like my therapist helped me understand my focus patterns, maybe understanding transformers helps us see how we decide what matters.

Now if only I could implement multi-head attention in dating...

Still waiting for "scaled dot-product listening" to be invented.

[Baba @babat00nday]

@fidexcode QBasic or Java not too sure anymore, it's been a while

[Baba @babat00nday]

I’ve been heads-down working on the next step in this journey, bringing all these lessons together.

It's still early, but I'm finally solving a problem that hits close to home.

I'll be sharing some updates in the upcoming weeks. Stay tuned.

#VoiceAI #AfricanTech #BuildInPublic #FounderJourney #AfricanLanguages

[Baba @babat00nday]

I recently came across some prototype videos that took me down memory lane to how this journey started back in 2021 — right after I left Meta.

This was long before the current AI explosion. 🧵

2021: The Earliest Prototype. A voice-first app for requesting songs via YouTube entirely in Yoruba.

[Baba @babat00nday]

The funny thing? All of this was built pre-LLMs.

We didn't have the "magic" of modern AI models yet. I just had the belief that voice interfaces were the future.

Every experiment was a stepping stone. They helped me understand the market potential and where I could add value.

[Baba @babat00nday]

@CatalinVoss @ReadWithEllo @getbluejay_ai @RohanVasishth @faraz_ml @cartesia @krandiash @jundesai @conduitai @0xpunnk @nke_ise @africadeeptech @JolloftechHQ @tokunboelegbe Ultimately, I left motivated and reminded that I'm "not alone" as a builder.

I wrote up my full, more-nuanced thoughts on this over on LinkedIn. Incredibly excited to see how these ecosystems grow.

https://t.co/FEA6K2qPnV

[Baba @babat00nday]

I've been exploring a nomadic life for a while now, but my recent trip to the Bay Area for SF Tech Week was too energizing not to share.

Left with 3 big takeaways on EdTech, a new wave of "AI", and the power of community.

[Baba @babat00nday]

@CatalinVoss @ReadWithEllo @getbluejay_ai @RohanVasishth @faraz_ml @cartesia @krandiash @jundesai @conduitai @0xpunnk Shout out to (@nke_ise) @africadeeptech, @JolloftechHQ, @tokunboelegbe for nurturing this community of builders.

[Baba @babat00nday]

The stars are aligning!

[Baba @babat00nday]

It is a skill to recognize BOTH talent and skill - @babat00nday

[Baba @babat00nday]

I have to let things flow, I want to be a stream and not a pond otherwise I would go stale and start smelling - @samuelekpe