Ontegra

Some days the best ideas come while listening to the rain 🌧️

Headsets on. Productivity up.
We don’t just work. We engineer success one sprint at a time.

We create smart systems, modernize outdated code, and integrate tools that power complete digital ecosystems.

Can AI actually slow down experienced developers?
New research suggests... yes. And the reason might surprise you.

For the last few years, tools like Copilot and Cursor have been hailed as game changers, replacing manual coding with AI-assisted development and promising massive productivity gains.

But a new study from METR found that seasoned developers were actually 20% slower when using AI for complex tasks.
Even more surprising: those same developers felt like they were working faster.

So what’s going on?

🧠 It turns out today’s AI often needs lots of back and forth to produce usable code, which slows things down more than just writing it yourself.

💡 One participant put it this way:

“LLMs are a big dopamine shortcut button… it’s a lot more enjoyable than the grueling alternative.”

But enjoyment isn’t the same as efficiency.

What this means:

LLMs may benefit junior devs more than seniors

Current tools still lack true task intuition

Productivity gains may be right around the corner
(METR notes AI’s software-task capability doubles every 7 months)

🗨️ What do you think?
Have AI tools actually helped your workflow, or do they slow you down?