Giacomo Balli profile picture
Giacomo Balli
The Mobile Guy

For founders and teams whose growth depends on mobile.
Clear judgment when AI, vendors, and product choices muddy the roadmap.

Find the Right Move LinkedIn

The 10x productivity claim with AI Is still mis

The 10x productivity claim with AI Is still misleading but in the way you're thinking.

Saying AI makes your devs "10x engineers" hides the real variable: what you’re comparing against.

Here’s the corrected model.

Known Stack vs Unknown Stack Changes the Baseline 1. Known Stack (You are already fluent) • AI accelerates execution: boilerplate, syntax recall, refactors. • Gains are real but bounded by review, debugging, edge cases, and architecture thinking. • Net result is often 1.5–3x, sometimes more, rarely 10x. Why: You were already productive. AI removes friction, not thinking.

2. Unknown Stack (You have zero prior knowledge) This is where most takes get it wrong. In an unknown stack, AI gains can be enormous because the baseline is not “slow productivity” but near-zero productivity.

If the alternative was: • 3–6 months to reach basic competence • Or never attempting the project at all

Then AI does not make you 10x faster. It makes the task possible.

Example: I built and programmed an Arduino-based device in a couple of days. Without AI, that project was effectively impossible for me. Not slower. Not harder. Impossible.

Why: AI collapses the learning curve by translating intent into executable steps, bypassing months of ramp-up.

Task Type Still Matters

AI shines when: • You can describe the goal clearly • The domain has existing patterns • The feedback loop is fast (compile, run, test)

AI struggles when: • Requirements are ambiguous • Systems are tightly coupled • Failure modes are subtle or high-risk

Unknown-stack gains are highest in contained systems like hardware prototypes, scripts, glue code, or one-off tools.

Debugging and Edge Cases Still Tax the Gains Even in unknown stacks: • You still pay for validation • You still debug misunderstood constraints • You still need to reason about edge cases AI will miss

The difference is that you are debugging something that exists, instead of studying for months before writing line one.

The Real Rule

AI does not deliver a fixed multiplier.

It reshapes the feasibility frontier: • Small gains where you were already strong • Massive gains where the alternative was “don’t even try”

So the correct framing is not “AI makes everyone 10x.” It’s:

AI turns many previously non-viable projects into viable ones... modestly accelerates the rest.

That distinction matters, especially for founders, operators, and anyone choosing what to build next.

Discuss on LinkedIn



Published: Fri, Jan 2 2026 @ 10:16:26
Back to Blog