# Transcript: Featured Session: 10 Breakthrough Technologies of 2026 **Date:** March 14, 2026 · 10:30 PM **Session:** [Featured Session: 10 Breakthrough Technologies of 2026](/sessions/2026-03-14/pp1148587-featured-session-10-breakthrough-technologies-of-2026) ## Summary Niall Firth from MIT Technology Review presented the publication's 25th annual list of 10 Breakthrough Technologies for 2026. The session covered sodium-ion batteries, AI coding tools, advanced nuclear reactors, AI marketing, base editing (CRISPR), de-extinction, mechanistic interpretability of AI, commercial space stations, embryo screening with polygenic scores, and AI data centers. Each technology was evaluated not just on innovation but on its potential scale and impact on how people live and work. ## Topics `breakthrough technologies` · `sodium-ion batteries` · `ai coding` · `advanced nuclear reactors` · `base editing` · `de-extinction` · `mechanistic interpretability` · `embryo screening` ## Key Takeaways 1. A breakthrough technology must change both scale and substance — innovations that affect everyone trivially or transform only a handful of specialists don't qualify. 2. Sodium-ion batteries could democratize energy storage by removing dependence on scarce lithium supply chains concentrated in just three countries. 3. AI coding tools have permanently changed software development, with 65% of workers using them weekly and expanding beyond code into research, legal, and finance work. 4. Embryo screening for traits like intelligence is already available at over 170 US clinics, raising profound ethical questions about genetic selection. 5. AI's carbon footprint is enormous — ChatGPT's 2.5 billion daily queries generate emissions equivalent to driving 200+ million miles per year, and this is projected to grow dramatically by 2028. ## Full Transcript Each year, MIT Technology Review reporters and editors assemble a list of the top breakthrough technologies that will change the world. The question is: what will really matter in the long term? It's not just about what's on the radar, but about scale, how many people are affected, and how completely a technology changes how we live or work. A technology that changes everyone but changes nothing isn't really a breakthrough, and a mind-blowing idea that never takes off isn't either. The publication has been doing this for twenty-five years, starting in 2001. Over that time, they've gotten some things right. For example, they highlighted the need for cryptographic protocols protected against attacks from future quantum computers, and in 2024, the U.S. National Institute of Standards and Technology released quantum-proof algorithms. They also put stem cells on the list in 2010, and clinical trials for epilepsy and Parkinson's treatments using that technology followed. Satellite constellations were featured six years ago, and Starlink has since launched enormous numbers of satellites into orbit. Some technologies didn't make this year's final list. GLP-1 weight loss drugs were considered but are still a few years from FDA approval in expanded forms. World models, AI systems that learn underlying rules of the world from observation like a baby learning about gravity, were discussed but deemed too early. Quantum-proof personal identity verification was also considered but no single approach has gained traction yet. The first breakthrough on the list is sodium-ion batteries. Most batteries today are based on lithium, which is expensive, hard to find, and creates supply chain complications. Sodium is everywhere and far cheaper. The batteries work the same way as lithium-ion, shuttling ions between electrodes, but they're much safer. Lithium mostly comes from three countries: China, Chile, and Australia. Sodium-ion batteries won't completely replace lithium-ion since they can't store quite as much energy, but they could be massively transformative for affordable energy storage. Next is AI coding tools, which have become a really major part of the software development workflow. A 2025 survey found that sixty-five percent of workers are using these tools weekly. What started as code completion has expanded into full code generation. The concept of 'vibe coding' emerged, where people build websites and apps using natural language rather than traditional programming. This has expanded beyond coding into research, documents, HR, finance, and legal work. Software development has changed forever. Advanced nuclear reactors are on the list. Conventional nuclear plants are expensive and difficult to build. The last one built in the US, in Georgia in 2024, was nearly a billion dollars over budget and finished seven years behind schedule. New reactor designs use alternative cooling methods like molten salt or liquid metal instead of pressurized water. These reactors can run at much higher temperatures and come in smaller, modular formats. They're being pursued not just for general power but specifically to supply energy-hungry AI data centers. AI marketing is another breakthrough. Companies are increasingly using AI tools like ChatGPT and dedicated platforms for portfolio management, content creation, and marketing. Seventy-two percent of US teams use AI for content creation. This represents an entirely new class of tool changing how businesses communicate and sell. Base editing, a new form of CRISPR gene editing, made the list. Traditional CRISPR can reach or shut off genes that cause problems, but base editing can actually rewrite individual letters of DNA bases in our genomes, restoring gene function. It's also safer than standard CRISPR. A baby named KJ received treatment for a unique genetic condition, and early results have been promising. The treatment cost has come down dramatically, potentially to just two or three dollars per edit. De-extinction technology is advancing with companies working on editing genes into modern animals to recreate traits of extinct species. Researchers have created 'woolly mice' by inserting ancient genes, and work is progressing toward larger animals. Each animal has around twenty genetic changes across fifteen genes associated with traits like larger body size and fur characteristics. Similar work is happening in plants, inserting ancient genes into crops like barley or wheat to make them hardier against pests, climate change, or extreme weather. Mechanistic interpretability is a breakthrough in understanding how AI models actually work internally. Researchers have found that concepts like the Golden Gate Bridge, objects like chairs and tables, and even abstract concepts exist as identifiable features inside AI models. When they amplified the Golden Gate Bridge feature, the model started referencing it in everything. When they examined how a model performed arithmetic, they discovered it used completely unexpected strategies, splitting problems into approximations rather than following human-like mental math. These X-ray techniques for AI are magnificent for understanding what's really happening inside these systems. Commercial space stations are coming as the International Space Station ages. A company called Vast plans to launch a station, Axiom Space is developing an inflatable space station, and Sierra Space has partnered with Blue Origin. These won't just be for science; they could host anyone with money who wants to go up. Russia is making its own station, India is building one too, and China already has a relatively new space station. Embryo screening using polygenic scores is perhaps the most controversial technology on the list. For decades, IVF patients have screened embryos for genetic disorders and chromosomal abnormalities. Now some startups are offering screening for traits related to appearance, behavior, or intelligence. Parents could potentially select embryos based on predicted IQ or physical characteristics. These tests deal in probabilities since most traits involve many different genetic elements, not single genes. A survey found thirty-seven percent of Americans would be comfortable with screening for intelligence. The technology is already offered at over one hundred and seventy clinics in the US. AI data centers represent massive infrastructure investment. Hundreds of billions of dollars are being committed to specialized warehouses with custom chips and new cooling systems for training AI models. OpenAI, Google, Amazon, and Meta are all investing heavily. The largest facilities will require more electricity than a small city. An MIT Technology Review investigation found that ChatGPT's 2.5 billion daily queries could emit carbon equivalent to driving over 200 million miles per year. By 2028, AI electricity usage could generate emissions equivalent to driving over 300 billion miles. AI is emerging as a societal factor with its own significant carbon footprint. --- *Source: stt · Language: en · Model: claude-opus-4-6* [← Back to session](/sessions/2026-03-14/pp1148587-featured-session-10-breakthrough-technologies-of-2026)