Dike E. Kalu

Fascinated by how software can impact businesses.

[Jason Cohen @asmartbear]

Most engineers I know love coding and hate management.

With AI, they’re managers, not doing what they love.

They’re writing specs (which they’ve always hated), then doing code reviews of sub-par code.

It’s efficient they say!

[UserCanDo @UserCanDo]

Happy New Year from everyone at UCD! 🎉💃🏾🥳

[UserCanDo @UserCanDo]

Merry Christmas from everyone at UCD 🎄🎉❤️

[Dike E. Kalu @dikekalu]

@Niklas_Sikorra UserCanDo

https://t.co/46akSM2lhz

[UserCanDo @UserCanDo]

Every benefits team has the same 3 spreadsheets killing productivity:

the Benchmark Matrix (50+ columns, 15-30% error rate)

the Frankenstein Renewal Calculator (broken VLOOKUPs & hidden tabs)
and the Vendor Comparison Grid.

Sound familiar?

[Dike E. Kalu @dikekalu]

@digitalocean Let it go

[Dike E. Kalu @dikekalu]

That feature everyone complains about? Stop wasting time trying to make it more intuitive. Cut it. Sometimes ruthless deletion > endless iteration."

[Olumide Adesina @olumidecapital]

The last decade has shown, it's a waste of time waiting for economic revival, self education and remote learning must be encouraged, because millions of kids in 🇳🇬 deserve to be educated

[Dike E. Kalu @dikekalu]

Dear API Devs,

Our front-end engineer overlords changed the terms of our friendship agreement. Going forward:

1) we must consume our own APIs. No work commitment without proof of test. 2) APIs must come with a client library for: Dart, Typescript, Swift, JavaScript, Kotlin, Rust.

They will graciously delay enforcement until 2025.

I suggest we all adopt inconsistent response payload structures to force them to the bargaining table.

Thoughts?

[Nnamdi “Enyi” Azodo 🐘 @SimplyAzodo]

PM when the frontend dev & designer are sorting things out 😁

[𝐀𝐔𝐍𝐓𝐘 𝐀𝐃𝐀 @Auntyadaa]

Nigerians just want to survive and not die of starvation. #EndBadGovernaceInNigeria 🇳🇬✊🏼

[Dike E. Kalu @dikekalu]

So much truth in this.

[I Am Devloper @iamdevloper]

When the client calls the prototype an MVP, remind them that those letters also stand for "Many Vexing Problems". Turns out words, like code, can be multifunctional.

[Dike E. Kalu @dikekalu]

Every once in a while you meet a hilariously caustic but inseparable front end and back end developer pair who should really have a sitcom developed around them.

Real Developers Of Lagos would be a hit.

[Python Nigeria @PythonNigeria]

#PyConNG Day 3

How has PyConNG 2024 been so far? Share with us in one word 🚀🚀🚀

#PyConNG2024

[Dike E. Kalu @dikekalu]

The thread below was actually triggered by a real experience.

We were battling with the Rust compiler to write a custom extractor for Axum (Rust) whereas spending 30 minutes reading the docs could've saved us hours of time.

It's a lesson that that is easily forgotten especially with so much AI/LLMs information now available at the finger tips.

On one hand ChatGPT led us on a chase down the wrong path.

But, ChatGPT later proved it's worth helping to write the quoted thread.

Here's to bringing our human ingenuity, creativity, and direction to make the most out of our newly found told.

If you're curious, here are the prompts that wrote that thread.

https://t.co/yGZvLGoLvg

[Dike E. Kalu @dikekalu]

Are you a software developer frustrated by code that just won't work? 🧵

Do you find yourself battling endless error messages and mysterious bugs?🤔

It's a common struggle. You've probably tried Stack Overflow, debugging, and even asked ChatGPT for help but still no luck?😤

Well, we have a solution to your problems!

[Dike E. Kalu @dikekalu]

Want more tips like these? Subscribe to my newsletter on LinkedIn https://t.co/G5ELd0hwbr

[Dike E. Kalu @dikekalu]

Introducing "Read the Docs"! The revolutionary technique that helps you understand your code's functionality.📚✨

And best of all, it costs less than a cup of overpriced coffee!
"Read the Docs" is comprehensive and FREE. Yes, you read that right! No more guesswork.🚀

[Dike E. Kalu @dikekalu]

Ask your inner programmer about "Read the Docs" today.

It's the ultimate guide to making your code work. Your future self will thank you.

Side effects may include sudden enlightenment, decreased reliance on trial and error, an uncontrollable urge to refactor everything, and spontaneous complaints about bad documentation.

[Dike E. Kalu @dikekalu]

Are you a software developer frustrated by code that just won't work? 🧵

Do you find yourself battling endless error messages and mysterious bugs?🤔

It's a common struggle. You've probably tried Stack Overflow, debugging, and even asked ChatGPT for help but still no luck?😤

Well, we have a solution to your problems!

[Dike E. Kalu @dikekalu]

AI is making software architects’ hair turn gray quicker.

In addition to drawing fancy diagrams with databases, application servers, front end frameworks, backend services, software architects now need to add in LLM models, prediction APIs, training pipelines, vector databases, agent orchestration, and more.

Here are some of the decisions I’ve had to contend with in the last few months.

I’ve ranked them on a scale of 1 to 5, with 5 being the toughest.

AI model choice (5 - so many choices out there)
What’s the performance, accuracy, cost, and privacy policy of the model?

Model fine-tuning (2 - general models were good enough)
Fine-tune or not to fine-tune? Do the general models do a great job without fine-tuning? What is the cost of errors?

Data management (5 - tons of small decisions with big impacts)
Do we need a vector store? What goes in there? What tokenizer do we use? What is the lifecycle of data stored in the vector database? How do we bring context into the queries by accessing outside data?

Cost Management (4 - money doesn’t grow on trees)
$5 per 1 million input token + $15 per 1 million output tokens (Open API - ChatGPT 4o) doesn’t sound like much. But it adds up quickly. Expecting a large ROI can eliminate this problem.

Overall, I’m grateful for the availability of metered prediction models + LLMs. It’s amazing what can be done without needing specialized AI/ML knowledge.

We’re still on the proof on concept stage on this project but curious to see how building an AI stack is working for people.