AI ENGINEER | Fashion Designer | Sound Engineer | Arsenal Fan | Dataguy @accionmfb
Joined August 2019
- Tweets8.6K
- Following1.3K
- Followers2.1K
- Likes103K
Pinned Tweet
I was in class last week. One of my students raised their hand mid-lecture and asked something that stopped me for a second.
"Why is every AI tool built on Python? C++ is faster. Rust is faster. Even Java is faster. So why Python?"
Honestly it’s a fair question. And the answer reveals something really interesting about how the AI industry actually works.
Let me explain this properly. 🧵
All models are wrong, but some are useful.
George Box
No dashboard, forecast or AI model is a perfect picture of reality. The question isn't "is it right?" It's "is it useful enough to make a better decision?"
An AI agent never sees your code.
It only sees a name, a description and a list of inputs. That's the tool schema, and it decides whether your agent works or guesses.
Here's how to write one properly.
AI Agents from Scratch 🧵
How to test it:
Give the agent 10 real questions. For each one, check which tool it picked and what inputs it sent.
If it picks the wrong tool, fix the description before you blame the model.
Amazing one Eureka!
Introducing Eureka Studios. 🚀
eurekastudios.co
We make games that feel like home: our roads, our jokes, our people.
Danfo Run was the first.
@eurekastudioshq
Joseph Pajos retweeted
Good morning
New day, fresh commit.
Yesterday’s bugs don’t define today’s build.
Learn fast. Keep going.
Let us not be weary in well doing: for in due season we shall reap, if we faint not.
Good morning
New day, fresh commit.
Yesterday’s bugs don’t define today’s build.
Learn fast. Keep going.
Let us not be weary in well doing: for in due season we shall reap, if we faint not.
Joseph Pajos retweeted
One thing every engineer learns sooner or later:
You can do everything right and still have things break.
What matters is how you respond when they do.
Do you panic, blame someone, and rush into changes?
Or do you slow down, check the evidence, and work through it?
How you handle the unexpected says a lot about the engineer you are.
Joseph Pajos retweeted
Pastor @nathanielblow o!
I want to scream😭My faith project has started pouring in!
I’ve been chasing some projects for 9 months & I said they must all be approved & funds released THIS YEAR! Two of many just got in TODAY!!😭💃🏽One of it, it’s been silence for one yr, but TODAY!
SQL interview challenge:
Tables:
customers (customer_id, name)
transactions (transaction_id, customer_id, amount, txn_date)
Find every customer who has made no transactions in the last 30 days.
The catch: customers who have never made a transaction at all must also appear in your result.
Note: solve it without using NOT IN.
Drop your answer below
Trace vs span: two words you need if you build or monitor AI agents
When something goes wrong in a system, the first question is: where exactly did it break? Traces and spans answer that.
Span
One single step of work. It has a name, a start time, an end time and details about what happened.
"Check account balance" is a span. "Call the fraud model" is a span. "Send SMS" is a span.
Trace
The full journey of one request, made up of all its spans, linked together in order.
"Customer sends ₦50,000" is one trace. Inside it: verify PIN, check balance, run fraud check, debit account, send SMS. Each one is a span.
The easy way to remember it:
A trace is the whole trip. A span is one stop on the way.
Why it matters for AI agents:
One question to an agent can trigger several model calls, three tool calls and a database search. If the answer comes back slow or wrong, the trace shows you every step, and the spans tell you which one took 8 seconds or returned rubbish.
Without traces, you're guessing. With them, you can point to the exact step.
Tools like OpenTelemetry, LangSmith and Langfuse record traces for you, so you don't have to build this yourself.
