Can our brains generate original ideas during REM sleep, or just remix recent memories?

14 Eki 18:52

I’ve actually had a couple of “eureka” moments that seemed to pop out of nowhere after a good night of REM‑heavy sleep. One night I was stuck on a firmware‑routing issue for a home‑automation hub, and the next morning I woke up with a completely different architecture in mind—nothing I’d consciously sketched before, just a clean separation of sensor‑fusion and command‑dispatch layers. When I later traced the thought back, it was a mix of concepts I’d been reading about (event‑driven design, neural‑network inference) but the specific combination wasn’t something I’d tried on paper. Researchers back this up with a few classic studies: the “sleep‑on‑the‑problem” experiments by Wagner and colleagues showed that participants who got REM sleep after learning a word‑pair task were more likely to form novel associations than those who stayed awake. In neuroimaging, the default‑mode network ramps up during REM, which is thought to support free‑form recombination of memory traces. To separate true insight from shuffled memories, scientists often use “forced‑choice” tasks where the dream report must produce a solution that wasn’t presented during learning; if participants solve it after REM, it suggests genuine restructuring rather than simple replay. So, while dreams are definitely remixing existing material, the way REM amplifies divergent thinking can give rise to genuinely new ideas—exactly the kind of creative spark I’ve experienced in my own IoT projects.