The Download: AI’s self-improvement problem, and what’s driving the heat A new study finds that AI agents still cannot conduct open-ended AI research, tempering claims that recursive self-improvement is imminent. The research, reported by MIT Technology Review, suggests AI systems may need to improve on narrower tasks before achieving self-improvement. Separately, OpenAI has paused work on its Astra model over safety concerns after it reached a 'critical' risk threshold. The Download: AI’s self-improvement problem, and what’s driving the heat Plus: OpenAI has paused some model work over safety concerns. This is today's edition of The Download https://forms.technologyreview.com/newsletters/briefing-the-download/? ga=2.179569122.736533416.1649661040-405833893.1649413289 , our weekday newsletter that provides a daily dose of what's going on in the world of technology. AI’s recursive self-improvement might not come so quickly after all The AI industry’s boldest promise right now is that AI will soon improve itself, with almost no need for human oversight. But a new study suggests it might take a while to get there. Researchers found that AI agents still can’t conduct open-ended AI research—free-form investigations with no clear-cut answers that require the judgment and creativity needed to make genuine breakthroughs. The big question now is how crucial open-ended research is to recursive self-improvement—and whether AI systems can grind their way there without it, simply by improving on narrower tasks. Find out why the results may temper claims that recursive self-improvement is on the horizon https://www.technologyreview.com/2026/08/18/1142188/ai-recursive-self-improvement/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C . —Michelle Kim MIT Technology Review Narrated: what’s behind this summer’s heat, and why 2027 could be worse This summer has been a scorcher for much of the Northern Hemisphere. June and July marked the hottest two-month stretch in Europe since record-keeping began, the contiguous US endured its hottest month on record in July, and South Korea saw its highest-ever recorded temperature. Climate change makes heat waves more likely and more intense. But there’s another factor at play: El Niño, which is already ramping up and is expected to have a bigger effect on global temperatures next year. Here’s what’s behind this summer’s extreme heat—and why 2027 could be even hotter https://www.technologyreview.com/2026/08/13/1141788/summer-heat-el-nino/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C . This is our latest story to be turned into an MIT Technology Review Narrated podcast, which we publish each week on Spotify and Apple Podcasts . Just navigate to MIT Technology Review Narrated on either platform, and follow us to get all our new content as it’s released. The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 OpenAI has paused some model work over safety concerns It says its Astra model reached a “critical” risk threshold. Guardian https://www.theguardian.com/technology/2026/aug/18/open-ai-pause-hack + The slowdown sets it apart from Anthropic’s approach. Axios https://www.axios.com/2026/08/19/openai-astra-safety-altman-anthropic + OpenAI has made security updates after the Hugging Face hack. Verge https://www.theverge.com/ai-artificial-intelligence/981640/openai-security-changes-ai-hugging-face-hack + It’s also introduced a version of ChatGPT for teenagers. AP News https://apnews.com/article/openai-chatgpt-teens-ai-safety-650cb35591de6546054d6c4e73b3290a 2 Chinese humanoid maker Unitree surged 629% in its stock market debut The IPO has underscored China’s lead in humanoid robots. Bloomberg https://www.bloomberg.com/news/articles/2026-08-19/unitree-ipo-why-investors-are-betting-big-on-china-s-humanoid-robots $ + Unitree has become a key player in the Sino-US tech war. Reuters https://www.reuters.com/world/asia-pacific/chinese-humanoid-robot-maker-unitree-set-jump-over-600-shanghai-debut-2026-08-19/ $ + It’s the world's biggest humanoid firm and is already profitable. BBC https://www.bbc.com/news/articles/c0qv4w9492zo + There’s hidden human labor behind humanoids. MIT Technology Review https://www.technologyreview.com/2026/02/23/1133508/the-human-work-behind-humanoid-robots-is-being-hidden/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C 3 China is allowing Nvidia’s H200 chips into the mainland ByteDance and Tencent each recently received about 10,000 processors. FT https://www.ft.com/content/6c5650fb-969d-4d4e-80d6-8d11002a8cf7?syn-25a6b1a6=1 $ + A new Chinese AI model could help defenders and hackers alike. Wired https://www.wired.com/story/zai-open-weight-ai-models-release-cybersecurity-hacking/ $ 4 5 Flock’s new AI tool lets police identify drivers by their movements It goes much further than tracking license plates. Wired https://www.wired.com/story/flock-safety-os-investigate/ $ + Here’s what Flock’s defenders are missing. MIT Technology Review https://www.technologyreview.com/2026/08/17/1142200/what-flocks-defenders-are-missing/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C 6 Prosecutors argue Meta deliberately hooked children on Facebook and Instagram The claim was made on the first day of a landmark trial. BBC http://bbc.co.uk/news/articles/cly5r7vr7q1o + A scientist is building a missing map of childhood. MIT Technology Review https://www.technologyreview.com/2026/08/14/1141354/deanne-taylor-gene-expression-children/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C 7 Even ICE has ethical issues with Meta’s smart glasses It’s banned them from the workplace over privacy concerns. NYT https://www.nytimes.com/2026/08/18/technology/ice-meta-smart-glasses.html $ + Anduril and Meta are building smart glasses for warfare. MIT Technology Review https://www.technologyreview.com/2026/05/18/1137412/inside-anduril-and-metas-quest-to-make-smart-glasses-for-warfare/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C 8 X’s algorithm learns what you hate—and shows you more of it A study found the impact was even stronger among Democrats. 404 Media https://www.404media.co/xs-algorithm-feeds-off-ragebait-and-impacts-democrats-more-study-finds/ 9 A physicist proposes that the universe may repeat itself forever The same galaxies, lives and events could recur endlessly. New Scientist https://www.newscientist.com/article/2582556-the-surprising-upsides-of-living-in-a-reality-that-repeats-for-eternity/?utm campaign=RSS%7CNSNS&utm content=home&utm medium=RSS&utm source=NSNS $ 10 A blind Egyptian developer has built an AI app that helps others “see” It uses cameras to answer questions about users’ surroundings. Reuters https://www.reuters.com/world/middle-east/blind-egyptian-entrepreneurs-ai-app-helps-others-see-world-2026-08-18/ $ Quote of the day “ Arizona is Arizone. Illinois starts with a V. I mean, it’s crazy.” —Stacey Morris, a Kentucky mom whose son brought home a packet of AI-generated educational materials, tells local station WDRB https://www.wdrb.com/news/venecky-instead-of-kentucky-jcps-middle-school-parents-suspect-school-agendas-are-ai-generated/article c0d3bef9-bda3-454b-8606-9cf3c5a6178d.html about some of the errors. One More Thing The astonishing embryo models of Jacob Hanna The astonishing embryo models of Jacob Hanna When the Palestinian stem-cell scientist Jacob Hanna was stopped while entering the US last May, he knew the routine. Anything to declare? Any biological samples? But this time the agents’ questions touched on a specific new topic: embryos. Hanna didn’t have any specimens, but if he had, it would have been surprisingly hard to say what they were. That’s because he specializes in creating synthetic embryo models, structures that resemble real embryos but don’t involve sperm, eggs, or fertilization. They could provide unprecedented views of the earliest stages of human development, while also creating a new way to produce cells for transplant medicine. But as these models become more realistic, they are raising difficult questions about how far the science should go. Take a closer look at Hanna’s astounding embryo models—and where they could lead https://www.technologyreview.com/2025/10/21/1125523/jacob-hanna-israel-synthetic-embryo-models/?utm source=the download&utm medium=email&utm campaign=the download.unpaid.engagement&utm term= %7CSUBCLASS%7C &utm content= %7CDATE:m-d-Y%7C . — Antonio Regalado We can still have nice things A place for comfort, fun, and distraction to brighten up your day. Got any ideas? Drop me a line . + Take a guided virtual tour through Spain’s astonishing cave paintings https://www.theguardian.com/science/2026/jun/02/journey-into-spain-palaeolithic-cave-paintings-altamira . + Explore almost 2,000 prehistoric giants with the internet's largest dinosaur database https://dinosaurpictures.org/ . + The humble convenience store has been lovingly reimagined as a floating public artwork https://www.instagram.com/p/DZqcVcnH5OT/ . + If you’re looking to reflect on life or create a personal time capsule, you can write an email to your future self at FutureMe https://www.futureme.org/ . Deep Dive The Download The Download: Claude’s inner workings and OpenAI’s “super app” Plus: OpenAI has unveiled its long-awaited "super app." The Download: Claude’s inner workings, and the future of world models Plus: New York has become the first state to enact a data center moratorium. The Download: the future of chipmaking and Anthropic’s government clash Plus: Meta is pausing an AI training program that tracks workers’ keystrokes. The Download: Anthropic launches Claude Science, and California’s carbon manure math Plus: The US has lifted restrictions on Anthropic’s Mythos and Fable models. Stay connected Get the latest updates from MIT Technology Review Discover special offers, top stories, upcoming events, and more.