# MIT Technology Review’s The Download on AI’s coding and science futures [Canonical reading](https://escapemeridian.com/readings/mit-technology-review-s-the-download-on-ai-s-coding-and-science-390df814) MIT Technology Review’s edition of The Download was published May 22, 2026. Will Douglas Heaven reports that many attendees at Anthropic’s Code with Claude event who said they had shipped code written entirely by Claude had not read it before pushing it live. He also reports Anthropic’s aim to push coding automation as far as it will go, alongside disagreement about that approach. Jessica Hamzelou presents the Enhanced Games, which allows performance-enhancing drugs, in the context of a culture focused on human enhancement and longevity. Grace Huckins describes two directions for AI-driven science: specialized systems for specific problems, and agentic, LLM-based systems that could eventually conduct cutting-edge research without human involvement. She says Gemini for Science leans toward the agent-driven direction but can still call on specialized systems, while Google appears to be transitioning away from them. Huckins asks how the shift could affect science; this newsletter excerpt does not give a detailed account of those effects. Its coding and science sections describe different forms of work being handed to AI, but do not state a shared conclusion about what people should trust AI to do. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source MIT Technology Review · Article Original source date: May 22, 2026 (2026-05-22T12:10:00Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-11-09T00:28:30.776Z. First observed: 2026-11-08T22:09:44.727Z. Retrieved: 2026-11-08T22:11:41.824Z. - [The Download: coding’s future, the ‘Steroid Olympics,’ and AI-driven science](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Limits Text as retrieved on November 8, 2026; originally published May 22, 2026. The newsletter summarizes linked stories rather than supplying their full evidence or detailed consequences. Its account of the possible effects of the shift in AI-driven science is limited to a teaser. ## Open question What changes to human oversight would follow if AI systems began conducting cutting-edge research projects without human involvement? ## Source-stated Will Douglas Heaven reports that many attendees who had shipped code written entirely by Claude had not read it before deployment. > Many admitted they hadn’t even read the code before pushing it live. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Heaven says Anthropic wants to push coding automation as far as it will go, while noting that not everyone agrees with that approach. > Anthropic says it wants to push automation as far as it will go. But not everyone is convinced that’s the right approach. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Jessica Hamzelou describes the Enhanced Games as permitting performance-enhancing drugs and links it to a wider enhancement-focused culture. > a controversial sporting competition that allows the use of performance-enhancing drugs. The goal? To “push the boundaries of human performance.” - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Grace Huckins contrasts specialized AI systems for specific scientific problems with agentic, LLM-based systems that could eventually conduct cutting-edge research without human involvement. > One builds specialized systems like WeatherNext for specific problems. The other pushes toward agentic, LLM-based systems that could eventually execute cutting-edge research projects without human involvement. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Huckins says Gemini for Science leans toward the agent-driven direction but can still call on specialized systems. > It can still call on specialized systems, but Google appears to be transitioning away from them. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Escape Meridian interpretation Escape Meridian reads the newsletter’s juxtaposition of coding and science as raising a shared question about what people may be asked to trust AI to do; the source does not detail the consequences. > Many admitted they hadn’t even read the code before pushing it live. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source and contributor context MIT Technology Review · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Agentic AI AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer. - [Definition reference](https://developers.google.com/machine-learning/glossary#agentic) ## Large language model A language model with many learned numerical settings, used to process and generate language. - [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Editorial provenance Version 1. Publication ID: 12a293ce-4007-c274-bdb6-58685f789cac. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T00:24:32.663Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: fd924437c0f3a55655577d58688419bfb8126fd4802a2e9aa3d064dace0a7f7f. Source text hash: c803953c0fb62b3e98e21e0c445ed67c3fc8b40c06392788eb775195e7bcf4db. ## Revision and source-check history ### Version 1 — 2026-11-09T00:28:30.776Z MIT Technology Review’s edition of The Download was published May 22, 2026. Will Douglas Heaven reports that many attendees at Anthropic’s Code with Claude event who said they had shipped code written entirely by Claude had not read it before pushing it live. He also reports Anthropic’s aim to push coding automation as far as it will go, alongside disagreement about that approach. Jessica Hamzelou presents the Enhanced Games, which allows performance-enhancing drugs, in the context of a culture focused on human enhancement and longevity. Grace Huckins describes two directions for AI-driven science: specialized systems for specific problems, and agentic, LLM-based systems that could eventually conduct cutting-edge research without human involvement. She says Gemini for Science leans toward the agent-driven direction but can still call on specialized systems, while Google appears to be transitioning away from them. Huckins asks how the shift could affect science; this newsletter excerpt does not give a detailed account of those effects. Its coding and science sections describe different forms of work being handed to AI, but do not state a shared conclusion about what people should trust AI to do. ## Original source MIT Technology Review · Article Original source date: May 22, 2026 (2026-05-22T12:10:00Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-11-09T00:28:30.776Z. First observed: 2026-11-08T22:09:44.727Z. Retrieved: 2026-11-08T22:11:41.824Z. - [The Download: coding’s future, the ‘Steroid Olympics,’ and AI-driven science](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Limits Text as retrieved on November 8, 2026; originally published May 22, 2026. The newsletter summarizes linked stories rather than supplying their full evidence or detailed consequences. Its account of the possible effects of the shift in AI-driven science is limited to a teaser. ## Open question What changes to human oversight would follow if AI systems began conducting cutting-edge research projects without human involvement? ## Source-stated Will Douglas Heaven reports that many attendees who had shipped code written entirely by Claude had not read it before deployment. > Many admitted they hadn’t even read the code before pushing it live. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Heaven says Anthropic wants to push coding automation as far as it will go, while noting that not everyone agrees with that approach. > Anthropic says it wants to push automation as far as it will go. But not everyone is convinced that’s the right approach. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Jessica Hamzelou describes the Enhanced Games as permitting performance-enhancing drugs and links it to a wider enhancement-focused culture. > a controversial sporting competition that allows the use of performance-enhancing drugs. The goal? To “push the boundaries of human performance.” - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Grace Huckins contrasts specialized AI systems for specific scientific problems with agentic, LLM-based systems that could eventually conduct cutting-edge research without human involvement. > One builds specialized systems like WeatherNext for specific problems. The other pushes toward agentic, LLM-based systems that could eventually execute cutting-edge research projects without human involvement. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source-stated Huckins says Gemini for Science leans toward the agent-driven direction but can still call on specialized systems. > It can still call on specialized systems, but Google appears to be transitioning away from them. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Escape Meridian interpretation Escape Meridian reads the newsletter’s juxtaposition of coding and science as raising a shared question about what people may be asked to trust AI to do; the source does not detail the consequences. > Many admitted they hadn’t even read the code before pushing it live. - [Supporting source](https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/) ## Source and contributor context MIT Technology Review · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Agentic AI AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer. - [Definition reference](https://developers.google.com/machine-learning/glossary#agentic) ## Large language model A language model with many learned numerical settings, used to process and generate language. - [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Editorial provenance Version 1. Publication ID: 12a293ce-4007-c274-bdb6-58685f789cac. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T00:24:32.663Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: fd924437c0f3a55655577d58688419bfb8126fd4802a2e9aa3d064dace0a7f7f. Source text hash: c803953c0fb62b3e98e21e0c445ed67c3fc8b40c06392788eb775195e7bcf4db. ### passed — 2026-11-09T00:24:32.663Z — anthropic/claude-opus-5.5 Publication: 12a293ce-4007-c274-bdb6-58685f789cac attribution — pass: This is a single newsletter issue that comments on several stories in its own words, so it counts as one piece rather than a listing. The coding, Enhanced Games and AI-science sections are correctly credited to the bylines that follow or head each section (Heaven, Hamzelou, Huckins). Anthropic's aim is attributed to Anthropic as reported. > Google I/O showed how the path for AI-driven science is shifting —Grace Huckins summary — pass: The reading faithfully covers the three lead items. It preserves the disagreement over Anthropic's automation push and Huckins's two-direction framing. Central terms (agentic, LLM) are defined on the page. Claude is understandable from context as Anthropic's AI. The reading omits Hassabis's 'foothills of the singularity' remark, but this does not distort the meaning. The question is conditional and follows from the evidence. > The contrast reflects two directions for AI in science. date — pass: May 22, 2026 matches the UTC calendar day of originalPublished (2026-05-22T12:10Z), and the URL path agrees. dateModified is about an hour later on the same day, and no editedAfterPublication flag is set. The Deep Dive links may be later additions, but the reading does not rely on them. There is no hindsight framing, and this is a publisher page, not an archive copy. support — pass: Each claim matches its quoted source text. Agentic research without human involvement is kept as a 'could eventually' possibility. The Escape Meridian juxtaposition is explicitly labeled as interpretation and notes that the source does not detail consequences. The limitations correctly say that the effects on science are only teased. > agentic, LLM-based systems that could eventually execute cutting-edge research projects without human involvement media — pass: No media is selected, and no verified candidates were supplied, so there is nothing to misidentify. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/390df814184c61de11a69e7e848ab8e7454b40e8727f8b02c6c78134335b4bde/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/mit-technology-review-s-the-download-on-ai-s-coding-and-science-390df814.json)