Kevin Sivabalan

Question from an ex-Uber engineer:

"I got this reachout from recruiting Uber. I responded that I'm happy to discuss why I left (so Uber can learn from it) but not planning to return.

I got ghosted. Why? They asked, after all!"

Here is exactly why (continued):

AGI is going to have an anxiety moment when it realizes all these tech CEOs are asking it to run their companies.

Will probably realize humans running the company are problematic and not logical.

Maybe it becomes the people’s champion instead

AI tools won't replace good developers—they'll amplify them. The most effective teams use AI to handle routine tasks while focusing human creativity on solving core business problems.

The right startup opportunity at the wrong life stage is still the wrong opportunity. Match your career decisions to your personal timeline.

The best teams are filled with singularly productive developers who are also great teammates. You have to hire them, nurture them, and make space for them to thrive.

@rfkuang.bsky.social I’m v excited you are coming to Toronto! And a new book!

Your standard 'build a to-do list' interview process might be bringing in the wrong people. Tailor your hiring process to reflect the actual work, not theoretical exercises.

Time, Money, Quality. In every project, all three form a sacred trinity. The one you must never sacrifice is quality.

The real success of projects comes from continuous interaction between developers and users. The idea of developers working in isolation for months and emerging with a perfect product is a dangerous fantasy.

If every person on a software project doesn't understand the 'why' behind it—from developers to designers to managers to clients—the project will inevitably fail.

Software projects don't fail due to bad code, but because of something far more fundamental: unclear purpose, poor communication, and inconsistent standards.

The most important AI metric isn't technical accuracy—it's user trust. If your team doesn't trust the AI's recommendations, they won't use them, regardless of how sophisticated your model is.

Always set realistic upper bounds with the AI you vibe code. Don’t let it order a couple thousand tshirts accidentally…

In my first lead role, I learned a crucial lesson: stakeholders care about business solutions, not technical details. Start with outcomes, adapt your communication. Technical depth matters, but impact matters more. #LeadershipLessons #TechCommunication

AI drive-thru attendants could hit 90% accuracy (better than humans at 86%) with 5 key components: Speech-to-Text, LLM, Text-to-Speech, Order Processing, and Data Warehouse. Annual cost? ~$11M. What technical challenge would be hardest to solve?

I've made a few perfect(as good as I can) codebases that are useless, because the business missed the mark on the product.

Sometimes you do everything right, but will fail. That's just life I guess

What did I do throughout the whole debacle? Kept on telling my director and hire ups this was an unclear scope, kept talking to the PMs and devs trying to help them navigate and build an impossible situation.

Needless to say, we were left holding the bag, and today that project no longer is running.

We only had budget for 5 months.... the project carried on for 1.5 years.

The scope just kept ballooning, but my director assured us, to keep building the form

Now anyone who knows anything about building a product from ground up, you know that when you try to tackle two different users as your primary user base, you're going to have a messy situation.

Then we started asking who the end user was. First it was the dispatcher... then we heard it was the truckers...

At first, they wanted just a simple form to replace the paper process, emails, phone calls that their dispatcher was getting. Sounded simple, but they also wanted scheduling!

There was a logistics application that my company was contracted to build. Immediately on day 1, the whole thing sounded really fishy.

I once had a software project fail, because a stakeholder sabotaged the communications

McDonald’s ends AI drive-thru trial as fast-food industry tests automation Companies have touted AI as future of the industry, but technology has also resulted in viral videos of wrong orders

McDonalds... I don't think you should axe your AI drive thru attendant... but does look like you are looking for a new partner to make it! www.theguardian.com/business/art...

Also I heard they have an open source version....

Anyone use Gumroad? I'm trying to start written content on substack, and thinking about various other avenues to start getting my thoughts out (maybe get paid). Wondering what other people have thought about it.

Have you used pi-hole? How did you set it up?