These Were SingularityHub’s Top 10 Stories in 2025

“Medications can keep some [Parkinson’s] symptoms at bay, but eventually, their effects wear off. For nearly half a century, scientists have been exploring an alternative solution: replacing dying dopamine neurons with new ones. [This year], two studies of nearly two dozen people with Parkinson’s showed the strategy is safe. A single transplant boosted dopamine levels for 18 months without notable side effects. Patients had few motor symptoms, even when they stopped taking regular medications.”

“A team from the IRCCS San Raffaele Scientific Institute in Italy treated infant mice for three blood-related genetic diseases with a custom gene-editing shot that directly edited cells in the mice’s blood. …The edits were long-lasting and survived when transplanted into mice who had not been given the therapy. A dose of ‘mobilizing agents’—chemicals that stimulate cells in the blood and immune system—further boosted the effect in young adult mice.”

“[Unitree’s R1 is] a humanoid robot priced at under $6,000. That’s not pocket change, but it’s orders of magnitude cheaper than most robots in its class, which can run into tens or even hundreds of thousands of dollars. The R1 packs serious mobility, sensors, and AI potential into a package that could fit in a university lab, a workspace—or even, if you’re adventurous, your living room.”

“Dubbed deep tissue in vivo sound printing (DISP), the system uses an injectable bioink that’s liquid at body temperature but solidifies into structures when blasted with ultrasound. A monitoring molecule, also sensitive to ultrasound, tracks tissue printing in real time. Excess bioink is safely broken down by the body.”

“Whereas answers can take minutes to complete on other hardware, Cerebras said that its version of DeepSeek knocked out some coding tasks in as little as 1.5 seconds. According to Artificial Analysis, the company’s wafer-scale chips were 57 times faster than competitors running the AI on GPUs and hands down the fastest. That was last week. Yesterday, Groq overtook Cerebras at the top with a new offering.”

“Using a voice synthesizer, the system translates words spoken in 101 languages into 36 others—not just into English, which tends to dominate current AI interpreters. In a head-to-head evaluation, the algorithm is 23 percent more accurate than today’s top models—and nearly as fast as expert human interpreters. It can also translate text into text, text into speech, and vice versa.”

“A video shows the automated equine galloping through valleys, crossing rivers, climbing mountains, and jumping over crevasses. …Kawasaki’s current motorbikes are constrained to roads, paths, and trails, but a machine with legs has no boundaries—it can reach places no other vehicles can go.”

“Many [mitochondrial] diseases are inherited. But none are treatable. …The new study, published in Science Translational Medicine, took a new approach [to treatment]—gene therapy. Using a genetic tool called base editing to target mitochondrial DNA, the team successfully rewrote damaged sections to overcome deadly mutations in mice.”

“CRISPR has a hefty problem: The system is too large, making it difficult to deliver the gene editor to cells in muscle, brain, heart, and other tissues. Now, a team at Mammoth Biosciences has a potential solution. …Their new iteration, dubbed NanoCas, slashed the size of one key component, Cas9, to roughly one-third of the original. …The compact NanoCas ‘opens the door’ for editing tissues inside the body.”

“The new study aimed to treat solid tumors like blood cancer—with a single injection into a patient’s vein. The team engineered CAR T cells that could hunt down metastasized cancer cells. When infused into the veins of mice they found the engineered cells rapidly shrank tumors in the liver and large intestines without causing dangerous immune side effects. The results ‘pave the way for a…clinical trial,’ wrote the team.”

“[Transmons, the type of qubit favored by the likes of Google and IBM,] have advantages such as faster operation speeds, but their short shelf life [known as coherence] remains a major disadvantage. Now a team from Princeton has designed novel transmon qubits with coherence times of up to 1.6 milliseconds—15 times longer than those used in industry and three times longer than the best lab experiment.”