AI coding productivity gains are 2-3x in practice, not the commonly cited higher figures.

A hidden human labor cost called 'bot sitting', the unglamorous work required to make AI usable, consumes 6.4 hours per week, roughly half the time AI is claimed to save.

“Which the report finds consumes 6.4 hours per week, or roughly half of all the time that AI supposedly saves.”

AI practitioners at a recursive event report getting 2x work done with AI assistance.

“The median answer was basically two. In other words, people felt like they're getting two times as >> work done thanks to AI. But that was also framed in an interesting way where it was like, but note that as of today >> if you were not there, your productivity would drop to close to zero.”

Box does not claim 10x productivity gains from AI for its engineering team.

“We don't claim that it's a 10x productivity game to our engineering team.”

Stripe built global tax filing in approximately one-third of the time it took to build US tax filing, despite greater complexity.

“We built that in about a third of the time that it took us to get to US filing.”

Coding will be a completely solved problem by the end of 2026.

“We're 6 monthsish or towards the end of the year to be completely done with code. Like it's a solved problem and then we'll probably hit some form of, you know, light RSI by end of next year.”

Laguna S outperforms models two to three times its size.

“We are outperforming models two or three times their size.”

Neural Concept AI helps engineers find designs that are 2-5% more aerodynamic than human-generated designs.

“We also help engineers to find designs that are 2, three, 5% more aerodynamic.”

As coding models accelerate implementation speed, compute will again become the primary bottleneck for AI research iteration.

“Now coding models are much more efficient and can help us implement stuff much faster. compute might become a bottleneck again because previously like if you want to train a new model say you want to generate new synthetic data and then or write a new algorithm it might take a few weeks and during that period of time you don't you might not have experiments to run but now you can build that thing within a few hours then you can immediately train a model now you have to have enough compute to try all of the ideas. So compute might be the bottleneck of iterating speed again.”

Entry-level candidates with AI proficiency are 5x-10x more productive than those without, causing Steven Bartlett to pass over non-AI-proficient candidates even for entry-level roles.

“The person that still is on the coldface of hiring in my company is when I see entry- level positions, the first thing I'm looking for is if they have an AI proficiency. And there are candidates now, even for me at entry- level positions that I'm not selecting for because I realize that someone with an AI proficiency in that exact same role is now like a five or 10x person.”