{"id":"36d85438bb687bba6400c4cabed1d7589b6ab3f2bf5d2281a2d6cec5c81f3686","title":"MIT Technology Review reports on MISIM and machine programming","argument":"MIT Technology Review’s July 29, 2020 report describes MISIM, a system developed by Justin Gottschlich and colleagues at Intel, MIT and the Georgia Institute of Technology. MISIM compares what code does, rather than how it is written, with programs in online repositories to find code with similar meaning. The report says the system can suggest alternative ways to write code and alert developers to differences from code known to be correct. In a preprint, the researchers report that MISIM is 40 times more accurate than earlier systems at finding similar code, including Aroma; this comparison does not establish how well it performs in everyday software development. Intel plans to use MISIM in-house for code recommendations, while the report describes translating older code and enabling people to create software from natural-language descriptions as possible future uses. DeepCode CTO Veselin Raychev cautions that bug-catching tools still produce many false positives, and Gottschlich says broader uses are not yet realized. How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 28, 2026; originally published July 29, 2020. The page was edited after publication; the evidence does not identify which passages, if any, were added later. The report does not provide enough detail here to assess MISIM’s performance across codebases or its practical effects in use.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"A new neural network could help computers code themselves","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","publishedAt":"2020-07-29T18:30:00Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"The report describes MISIM as a system developed by Gottschlich and colleagues from Intel, MIT and the Georgia Institute of Technology to find code with similar meaning.","excerpt":"he has developed a system called Machine Inferred Code Similarity, or MISIM, that can extract the meaning of a piece of code","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"MISIM compares code with programs from online repositories, representing what code does rather than how it is written.","excerpt":"First it translates the code into a form that captures what it does but ignores how it is written","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"The researchers report that MISIM is 40 times more accurate than earlier systems at identifying similar code, including Aroma.","excerpt":"In a preprint, Gottschlich and his colleagues report that MISIM is 40 times more accurate than previous systems that try to do this, including Aroma.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"The report says Intel plans to use MISIM in-house to recommend alternative ways to write code.","excerpt":"Intel plans to use the tool as a code recommendation system for developers in-house, suggesting alternative ways to write code that are faster or more efficient.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"Raychev says machine-learning bug-catching tools often produce false positives, and that predicting bugs remains difficult.","excerpt":"Typically AI bug-catching tools produce lots of false positives, he says.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"Gottschlich presents writing software through natural-language descriptions as a possible future use, not a demonstrated capability.","excerpt":"the ability to work with the meaning of code separately from its textual representation could one day let people write software simply by describing what they want to do in words, he says.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","checkedAt":null,"items":[]},"assumptions":[],"whatWouldChangeView":{"attribution":"Open question","text":"How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?"},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports on MISIM and machine programming","summary":"MIT Technology Review’s July 29, 2020 report describes MISIM, a system developed by Justin Gottschlich and colleagues at Intel, MIT and the Georgia Institute of Technology. MISIM compares what code does, rather than how it is written, with programs in online repositories to find code with similar meaning. The report says the system can suggest alternative ways to write code and alert developers to differences from code known to be correct. In a preprint, the researchers report that MISIM is 40 times more accurate than earlier systems at finding similar code, including Aroma; this comparison does not establish how well it performs in everyday software development. Intel plans to use MISIM in-house for code recommendations, while the report describes translating older code and enabling people to create software from natural-language descriptions as possible future uses. DeepCode CTO Veselin Raychev cautions that bug-catching tools still produce many false positives, and Gottschlich says broader uses are not yet realized. How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","question":"How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","claims":[{"text":"The report describes MISIM as a system developed by Gottschlich and colleagues from Intel, MIT and the Georgia Institute of Technology to find code with similar meaning.","excerpt":"he has developed a system called Machine Inferred Code Similarity, or MISIM, that can extract the meaning of a piece of code","support":"source-stated"},{"text":"MISIM compares code with programs from online repositories, representing what code does rather than how it is written.","excerpt":"First it translates the code into a form that captures what it does but ignores how it is written","support":"source-stated"},{"text":"The researchers report that MISIM is 40 times more accurate than earlier systems at identifying similar code, including Aroma.","excerpt":"In a preprint, Gottschlich and his colleagues report that MISIM is 40 times more accurate than previous systems that try to do this, including Aroma.","support":"source-stated"},{"text":"The report says Intel plans to use MISIM in-house to recommend alternative ways to write code.","excerpt":"Intel plans to use the tool as a code recommendation system for developers in-house, suggesting alternative ways to write code that are faster or more efficient.","support":"source-stated"},{"text":"Raychev says machine-learning bug-catching tools often produce false positives, and that predicting bugs remains difficult.","excerpt":"Typically AI bug-catching tools produce lots of false positives, he says.","support":"source-stated"},{"text":"Gottschlich presents writing software through natural-language descriptions as a possible future use, not a demonstrated capability.","excerpt":"the ability to work with the meaning of code separately from its textual representation could one day let people write software simply by describing what they want to do in words, he says.","support":"source-stated"}],"limitations":["Text as retrieved on October 28, 2026; originally published July 29, 2020.","The page was edited after publication; the evidence does not identify which passages, if any, were added later.","The report does not provide enough detail here to assess MISIM’s performance across codebases or its practical effects in use."],"mediaId":null,"correctionReason":null,"id":"36d85438bb687bba6400c4cabed1d7589b6ab3f2bf5d2281a2d6cec5c81f3686","publicationId":"d479f6ad-808a-faef-4492-199dbee36503","version":1,"sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","sourceTitle":"A new neural network could help computers code themselves","publisher":"MIT Technology Review","publishedAt":"2020-07-29T18:30:00Z","originalPublished":{"value":"2020-07-29T18:30:00Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-10-28T08:50:14.637Z","createdAt":"2026-10-29T00:51:38.814Z","historical":true,"entityIds":[],"topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"faf9ca38002729e99b4065df34c051a18d7c33c0ec3d6ff618d57b7a599bca3c","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-28T08:50:11.724Z"},"provenance":{"textHash":"7a06ffd93a75e4ebc99b86d888aeaff618f75cff706f8e68ccfe2808e0764dad","observationId":"e6b6511c12e041b88b9a4456532ba63e0601b8c704b5ff9eb41b9798d4276548","retrievedAt":"2026-10-28T08:51:52.113Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"c4a65e37-f8a5-bbea-e105-f703fc7e8817"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"53512b38-3145-6420-df69-b143ee0f6e40"}},"audit":{"status":"passed","reviewedAt":"2026-10-29T00:50:28.648Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/mit-technology-review-reports-on-misim-and-machine-programming-36d85438","formats":{"html":"/readings/mit-technology-review-reports-on-misim-and-machine-programming-36d85438","markdown":"/readings/mit-technology-review-reports-on-misim-and-machine-programming-36d85438.md","json":"/readings/mit-technology-review-reports-on-misim-and-machine-programming-36d85438.json"},"attribution":"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.","readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[],"entities":[]},"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"36d85438bb687bba6400c4cabed1d7589b6ab3f2bf5d2281a2d6cec5c81f3686","revision":"d479f6ad-808a-faef-4492-199dbee36503","revisionKind":"publication-id","title":"MIT Technology Review reports on MISIM and machine programming","url":"https://escapemeridian.com/readings/mit-technology-review-reports-on-misim-and-machine-programming-36d85438"},"source":{"type":"source","id":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","revision":"7a06ffd93a75e4ebc99b86d888aeaff618f75cff706f8e68ccfe2808e0764dad","revisionKind":"sha256-source-text","title":"A new neural network could help computers code themselves","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/"},"status":"available","items":[],"next":null,"scope":"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."},"revisions":[{"id":"36d85438bb687bba6400c4cabed1d7589b6ab3f2bf5d2281a2d6cec5c81f3686","title":"MIT Technology Review reports on MISIM and machine programming","argument":"MIT Technology Review’s July 29, 2020 report describes MISIM, a system developed by Justin Gottschlich and colleagues at Intel, MIT and the Georgia Institute of Technology. MISIM compares what code does, rather than how it is written, with programs in online repositories to find code with similar meaning. The report says the system can suggest alternative ways to write code and alert developers to differences from code known to be correct. In a preprint, the researchers report that MISIM is 40 times more accurate than earlier systems at finding similar code, including Aroma; this comparison does not establish how well it performs in everyday software development. Intel plans to use MISIM in-house for code recommendations, while the report describes translating older code and enabling people to create software from natural-language descriptions as possible future uses. DeepCode CTO Veselin Raychev cautions that bug-catching tools still produce many false positives, and Gottschlich says broader uses are not yet realized. How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 28, 2026; originally published July 29, 2020. The page was edited after publication; the evidence does not identify which passages, if any, were added later. The report does not provide enough detail here to assess MISIM’s performance across codebases or its practical effects in use.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"A new neural network could help computers code themselves","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","publishedAt":"2020-07-29T18:30:00Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"The report describes MISIM as a system developed by Gottschlich and colleagues from Intel, MIT and the Georgia Institute of Technology to find code with similar meaning.","excerpt":"he has developed a system called Machine Inferred Code Similarity, or MISIM, that can extract the meaning of a piece of code","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"MISIM compares code with programs from online repositories, representing what code does rather than how it is written.","excerpt":"First it translates the code into a form that captures what it does but ignores how it is written","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"The researchers report that MISIM is 40 times more accurate than earlier systems at identifying similar code, including Aroma.","excerpt":"In a preprint, Gottschlich and his colleagues report that MISIM is 40 times more accurate than previous systems that try to do this, including Aroma.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"The report says Intel plans to use MISIM in-house to recommend alternative ways to write code.","excerpt":"Intel plans to use the tool as a code recommendation system for developers in-house, suggesting alternative ways to write code that are faster or more efficient.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"Raychev says machine-learning bug-catching tools often produce false positives, and that predicting bugs remains difficult.","excerpt":"Typically AI bug-catching tools produce lots of false positives, he says.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"},{"text":"Gottschlich presents writing software through natural-language descriptions as a possible future use, not a demonstrated capability.","excerpt":"the ability to work with the meaning of code separately from its textual representation could one day let people write software simply by describing what they want to do in words, he says.","url":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","checkedAt":null,"items":[]},"assumptions":[],"whatWouldChangeView":{"attribution":"Open question","text":"How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?"},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports on MISIM and machine programming","summary":"MIT Technology Review’s July 29, 2020 report describes MISIM, a system developed by Justin Gottschlich and colleagues at Intel, MIT and the Georgia Institute of Technology. MISIM compares what code does, rather than how it is written, with programs in online repositories to find code with similar meaning. The report says the system can suggest alternative ways to write code and alert developers to differences from code known to be correct. In a preprint, the researchers report that MISIM is 40 times more accurate than earlier systems at finding similar code, including Aroma; this comparison does not establish how well it performs in everyday software development. Intel plans to use MISIM in-house for code recommendations, while the report describes translating older code and enabling people to create software from natural-language descriptions as possible future uses. DeepCode CTO Veselin Raychev cautions that bug-catching tools still produce many false positives, and Gottschlich says broader uses are not yet realized. How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","question":"How well would MISIM’s recommendations work on code beyond the repositories and examples used to develop and evaluate it?","claims":[{"text":"The report describes MISIM as a system developed by Gottschlich and colleagues from Intel, MIT and the Georgia Institute of Technology to find code with similar meaning.","excerpt":"he has developed a system called Machine Inferred Code Similarity, or MISIM, that can extract the meaning of a piece of code","support":"source-stated"},{"text":"MISIM compares code with programs from online repositories, representing what code does rather than how it is written.","excerpt":"First it translates the code into a form that captures what it does but ignores how it is written","support":"source-stated"},{"text":"The researchers report that MISIM is 40 times more accurate than earlier systems at identifying similar code, including Aroma.","excerpt":"In a preprint, Gottschlich and his colleagues report that MISIM is 40 times more accurate than previous systems that try to do this, including Aroma.","support":"source-stated"},{"text":"The report says Intel plans to use MISIM in-house to recommend alternative ways to write code.","excerpt":"Intel plans to use the tool as a code recommendation system for developers in-house, suggesting alternative ways to write code that are faster or more efficient.","support":"source-stated"},{"text":"Raychev says machine-learning bug-catching tools often produce false positives, and that predicting bugs remains difficult.","excerpt":"Typically AI bug-catching tools produce lots of false positives, he says.","support":"source-stated"},{"text":"Gottschlich presents writing software through natural-language descriptions as a possible future use, not a demonstrated capability.","excerpt":"the ability to work with the meaning of code separately from its textual representation could one day let people write software simply by describing what they want to do in words, he says.","support":"source-stated"}],"limitations":["Text as retrieved on October 28, 2026; originally published July 29, 2020.","The page was edited after publication; the evidence does not identify which passages, if any, were added later.","The report does not provide enough detail here to assess MISIM’s performance across codebases or its practical effects in use."],"mediaId":null,"correctionReason":null,"id":"36d85438bb687bba6400c4cabed1d7589b6ab3f2bf5d2281a2d6cec5c81f3686","publicationId":"d479f6ad-808a-faef-4492-199dbee36503","version":1,"sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","sourceTitle":"A new neural network could help computers code themselves","publisher":"MIT Technology Review","publishedAt":"2020-07-29T18:30:00Z","originalPublished":{"value":"2020-07-29T18:30:00Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-10-28T08:50:14.637Z","createdAt":"2026-10-29T00:51:38.814Z","historical":true,"entityIds":[],"topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/29/1005768/neural-network-similarities-between-programs-help-computers-code-themselves-ai-intel/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"faf9ca38002729e99b4065df34c051a18d7c33c0ec3d6ff618d57b7a599bca3c","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-28T08:50:11.724Z"},"provenance":{"textHash":"7a06ffd93a75e4ebc99b86d888aeaff618f75cff706f8e68ccfe2808e0764dad","observationId":"e6b6511c12e041b88b9a4456532ba63e0601b8c704b5ff9eb41b9798d4276548","retrievedAt":"2026-10-28T08:51:52.113Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"c4a65e37-f8a5-bbea-e105-f703fc7e8817"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"53512b38-3145-6420-df69-b143ee0f6e40"}},"audit":{"status":"passed","reviewedAt":"2026-10-29T00:50:28.648Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"d479f6ad-808a-faef-4492-199dbee36503","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the report to MIT Technology Review and attributes MISIM to Gottschlich and colleagues at Intel, MIT and Georgia Tech. It attributes the 40x figure to the researchers' preprint and the false-positive caution to DeepCode CTO Veselin Raychev. These all match the source.","excerpt":"Working with a team from Intel, MIT and the Georgia Institute of Technology in Atlanta, he has developed a system called Machine Inferred Code Similarity, or MISIM"},{"dimension":"summary","status":"pass","explanation":"The summary leads with the actor and the development. It explains in plain language that MISIM compares what code does rather than how it is written. It separates planned in-house use from speculative future uses and preserves Raychev's caution. Describing Gottschlich's 'could one day' vision as not yet realized is a fair paraphrase. The question follows from the evidence and is not leading. A general reader can follow the reading.","excerpt":"Intel plans to use the tool as a code recommendation system for developers in-house"},{"dimension":"date","status":"pass","explanation":"July 29, 2020 matches the UTC day of originalPublished (2020-07-29T18:30Z). The evidence kind is publisher-page, so no archive disclosure is needed. Later Deep Dive sidebar items, such as those about AlphaGo's ten-year anniversary, are not used. Plans and visions are reported as forecasts. The page was edited in December 2020, but the cited passages are the article body, and the limitations disclose the edit.","excerpt":"could one day let people write software simply by describing what they want to do in words, he says."},{"dimension":"support","status":"pass","explanation":"Each claim's excerpt is a verbatim passage from the article body and supports the claim. The reading explicitly says the 40x preprint comparison does not establish everyday performance, and it does not overstate benefits. The post-publication edit is disclosed, and the cited passages are plainly part of the original article text rather than later sidebar additions.","excerpt":"In a preprint, Gottschlich and his colleagues report that MISIM is 40 times more accurate than previous systems that try to do this, including Aroma."},{"dimension":"media","status":"pass","explanation":"No media is selected, and no verified candidate media was supplied. Selecting none is appropriate.","excerpt":null}],"createdAt":"2026-10-29T00:50:28.648Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}