Our CEO, Amanda Richardson, and our Head of Product, Frank Hauben, can’t agree on anything else.
We put Amanda and Frank on camera, handed them five spicy hot takes about AI in hiring, and told them not to sit on the fence. They sat on the fence anyway. Here’s what they actually fought about.
Amanda hires for the next ten years. Frank hires for next Tuesday. Both are right, which is exactly why this is hard. Here are the five fights, and the one thing they finally agreed on.
01. Is “AI fluency” a real skill, or a fad we all agreed to take seriously?
The take: if you’re measuring how well someone prompts, you’ve already lost the plot.
Amanda didn’t hedge: AI fluency isn’t really a thing. In a year we’ll be embarrassed that we treated it as a separate skill, like listing “Microsoft PowerPoint” on a resume as though it were a personality trait. Look at whether a candidate’s actual skills shine through, not whether they’re “good at AI.”
Frank took the other side without blinking. It’s the number one thing customers ask about. Not the clean prompt, but whether you understand what context the model even has, how you feed it, and what you know to pull back out. That’s not a party trick, it’s something he wants to hire for.
“It’s not yes-and. It’s necessary but not sufficient.”
The question underneath: is AI a baseline skill like literacy, or the differentiator between good and great? Nobody would commit. Which is the honest answer.
02. Speed is a vanity metric.
The take: everyone bragging about moving faster is optimizing the wrong number.
Frank’s favorite company value and favorite metric are the same: speed. Speed to prototype, to ship, to get something in front of a real user. If he could optimize one number, that’s it.
Amanda’s response: he only says that because he’s faster than me. Then she really dug in – speed is short-term. As models get faster, nobody wins on speed, because everyone has it. You win on good inputs and the right outputs. And then she coined the keeper:
“Bot-shitting is real. Just because you can create it fast doesn’t mean it’s good, and just because you can ship it fast doesn’t mean it’s built well, or that it’s even the right thing.”
Then she argued against herself: slow because you’re thinking critically? Good. Slow because you’re scared to pick a prototype? Build them all fast and let feedback decide. Frank conceded too: five half-built features is worse than one that’s finished.
03. Prompt quality and token usage are noise.
The take: the signal isn’t the clean prompt, it’s the result.
Rare agreement, sort of. Both think token-counting is a distraction. As Amanda put it, unless you’re the CFO, why do you care?
But the token question refuses to die inside a real interview. If two candidates hit the same output and one did it at a fraction of the cost, do you hire the cheaper one? Do you let candidates pick their own model when the pick changes their score? Is model selection a skill, or a constraint we haven’t figured out how to grade? No answers, just a dozen new questions the industry hasn’t solved.
04. Rank them: process, output, speed. No fence-sitting.
The take: you have to choose. Pick a hierarchy and defend it.
Amanda put the process first, output second, speed dead last. Her reasoning is the sharpest idea in the whole conversation, and it’s specifically about hiring:
“If you hire on output alone, you’ll miss great candidates who just got lost in the process. If they have a good process, you can coach them to the right output. It’s the McKinsey case study: miss the question at the top of the branch and the whole answer comes out wrong, even though the person is brilliant.”
Frank, the self-declared speed evangelist, surprised everyone and ranked output first. His logic: speed was never the goal, it was always speed to output, and output is the thing you actually get paid for.
The unlock was the distinction: process wins when you’re hiring, output wins when you’re managing performance. Two jobs, two metrics. Most companies use one yardstick for both and wonder why it doesn’t work.
05. Is AI fluency a real skill, or a category we invented to sell things?
The take: the last one, and the meanest one.
Amanda, who happily admits she’s all about inventing categories and selling things, still called it: AI fluency is the next word processor. Then she declined to fully commit, because plenty of customers are treating AI familiarity as a real culture change. Fence-sitting, and she said so.
Then Frank landed it:
“Maybe we should just call it a mindset. Then it’s something we can all accept.”
And that’s where they met. Not fluency. Not prompting. Mindset. Curiosity, willingness to experiment, willingness to try modern tooling before anyone tells you to. The word will change within months. The thing underneath it won’t.
So what do you actually do with this?
Two aligned leaders at the same company couldn’t agree on how to hire for AI, and they built the platform other companies interview on. If it’s this contested inside CoderPad, it’s a coin flip inside a committee that hasn’t thought about it at all.
That’s not a reason to freeze, it’s a reason to be deliberate. The companies getting this right aren’t the ones with the perfect definition of “AI fluency.” They’re the ones who decided what they value, built an interview that tests for it, and stopped guessing. Choose on purpose, not by accident.
Stop guessing what AI-native talent looks like. Ask the people who measure it.
Book a 15-minute call with a CoderPad AI hiring consultant. Bring a role you’re hiring for. We’ll help you pin down what to test for, what a good process looks like, and how to spot the mindset that separates AI-native from AI-curious.