If you mostly use chatbots to answer questions, this is the part worth paying attention to. The real shift is not that AI writes code faster. It is that coding agents can reopen ideas you already gave up on because the build work felt too annoying, too slow, or too complex.
Most people will read Terry Tao's post as another model demo. That misses the useful decision. A product update matters when it changes your next decision, not when it adds another feature bullet. If you miss that, you waste time, budget, and attention on the wrong comparison.
The sharpest version is simple: coding agents do not replace programmers. They replace 'too hard, maybe later.' [C002]
In Tao's July 11 post, 'Old and new apps, via modern coding agents by Terry Tao' [C001], the key fact is not just that he made a new app. It is that he wanted a relativity visualization tool back in 1999, dropped it because the coding complexity was too high, and then got a working version in a few hours in 2026 with Claude Code.
The public transcript is what makes this more than a slogan. Tao was not sitting back and watching magic. He kept the physics and math constraints straight, corrected mistakes, and checked details while Claude Code handled much of the repetitive build-and-iterate work. So the lesson is not 'anyone can ship complex software in hours.' The lesson is that the overhead that kills a good idea can shrink fast when the person guiding the work knows what must stay true.
Boundary: this is one public Claude Code build story, not a benchmark across tools, teams, hardware, or operating systems. But it is still a useful filter for your own backlog. Start with the project you parked at 'too much hassle,' not the one you already knew how to finish. Share this with the person whose best idea is still sitting in the someday pile.