MIT Technology Review · Apr 2, 2025
Thoughtworks argues in MIT Technology Review that developers still hold the keys
Markdown · JSON · Original source
Thoughtworks’ sponsored article, published April 2, 2025, argues that AI will change software development but make developers’ judgment more important, not less. It warns that relying on AI-generated code without enough care can leave code harder to maintain and that weakening development skills could leave fewer people able to repair software later. The article cites GitClear research as indicating increased code churn and fewer lines moved, which it treats as signs of less refactoring and attention to code quality. Thoughtworks argues that developers do more than write code: much of their work is adapting existing software, and they must judge whether AI-built systems are correct. It also describes possible uses for AI, including coding assistance and understanding legacy codebases; Thoughtworks says its own success using Code with an internal tool was mixed, though it sees promise. Its conclusion is that software development should value good coding practices and developer responsibility, rather than measuring work only by output. The article’s concern is that thoughtless automation could make software more difficult and costly to maintain. How much AI-generated code changes the quality and maintenance costs of software in practice remains unknown from the evidence provided.
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
Source details
Original source date: Apr 2, 2025 (2025-04-02T17:50:55Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-14T18:53:32.000Z. First observed: 2026-10-28T07:26:03.306Z. Retrieved: 2026-10-28T07:27:51.087Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 28, 2026; originally published April 2, 2025. The article describes GitClear’s findings but provides no underlying measurements or comparison details in the retrieved text. This is sponsored content produced by Thoughtworks; the page says it was not written by MIT Technology Review’s editorial staff.
Open question
What evidence would show whether AI-generated code increases maintenance costs or reduces developers’ ability to repair software?
Source-stated
Thoughtworks argues that AI-generated code used without sufficient care risks becoming harder to maintain and that reduced value placed on software development skills could leave fewer people able to fix software.
Increasing complacency around AI-generated code and a shift to what has been termed “vibe coding” — where code is generated through natural language prompts until the results seem to work — will lead to code that’s more error-strewn, more expensive to run and harder to change in the future.
Source-stated
Thoughtworks says GitClear research found increased code churn and a decrease in lines moved, which the article associates with less refactoring and attention to code quality.
Research done by GitClear earlier this year indicates that with AI coding assistants (like GitHub Copilot) going mainstream, code churn — which GitClear defines as “changes that were either incomplete or erroneous when the author initially wrote, committed, and pushed them to the company’s git repo"
Source-stated
Thoughtworks argues that developers’ work includes evolving existing software and that developer judgment will be important as AI-generated code proliferates.
A large proportion of their work is evolving and adapting existing (and sometimes legacy) software. Sprawling and janky code bases that have taken on technical debt are, unfortunately, the norm.
Source-stated
Thoughtworks says its success using Claude Code to add language support to an internal tool was mixed, while it sees promise in using generative AI to understand legacy codebases.
In fact, Anthropic’s Claude Code helped us add support for new languages in an internal tool, CodeConcise. We use CodeConcise to understand legacy systems; and while our success was mixed, we do think there’s real promise here.
Source-stated
Thoughtworks argues that developers will have responsibility for ensuring confidence in AI-built systems, even if verification is AI-assisted.
the job of verifying that an AI-built system is correct will fall to humans. Yes, verification itself might be AI-assisted, but it will be the role of the software developer to ensure confidence.
Source-stated
Thoughtworks concludes that software development should be valued and that thoughtless automation could be problematic and costly in the years to come.
It’s vital software development is valued: the impact of thoughtless automation and pure vibes could prove incredibly problematic (and costly) in the years to come.
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.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: df01bdcb-43db-a02d-7bed-fa8facee85e1.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T18:48:47.520Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 502513c03ea8d4638507b4c9b0ad9d2de864b09800799a9d69569868b2400e24. Source text hash: 6b66253402dcb9c02dbd0d35ae03f0ed3a68c3e6bed6151c45fec6137df0c5f1.
Revision and source-check history
Version 1 · 2026-11-14T18:53:32.000Z
Thoughtworks argues in MIT Technology Review that developers still hold the keys
Thoughtworks’ sponsored article, published April 2, 2025, argues that AI will change software development but make developers’ judgment more important, not less. It warns that relying on AI-generated code without enough care can leave code harder to maintain and that weakening development skills could leave fewer people able to repair software later. The article cites GitClear research as indicating increased code churn and fewer lines moved, which it treats as signs of less refactoring and attention to code quality. Thoughtworks argues that developers do more than write code: much of their work is adapting existing software, and they must judge whether AI-built systems are correct. It also describes possible uses for AI, including coding assistance and understanding legacy codebases; Thoughtworks says its own success using Code with an internal tool was mixed, though it sees promise. Its conclusion is that software development should value good coding practices and developer responsibility, rather than measuring work only by output. The article’s concern is that thoughtless automation could make software more difficult and costly to maintain. How much AI-generated code changes the quality and maintenance costs of software in practice remains unknown from the evidence provided.
Original source
MIT Technology Review · Article
Source details
Original source date: Apr 2, 2025 (2025-04-02T17:50:55Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-14T18:53:32.000Z. First observed: 2026-10-28T07:26:03.306Z. Retrieved: 2026-10-28T07:27:51.087Z.
Limits
Text as retrieved on October 28, 2026; originally published April 2, 2025. The article describes GitClear’s findings but provides no underlying measurements or comparison details in the retrieved text. This is sponsored content produced by Thoughtworks; the page says it was not written by MIT Technology Review’s editorial staff.
Open question
What evidence would show whether AI-generated code increases maintenance costs or reduces developers’ ability to repair software?
Source-stated
Thoughtworks argues that AI-generated code used without sufficient care risks becoming harder to maintain and that reduced value placed on software development skills could leave fewer people able to fix software.
Increasing complacency around AI-generated code and a shift to what has been termed “vibe coding” — where code is generated through natural language prompts until the results seem to work — will lead to code that’s more error-strewn, more expensive to run and harder to change in the future.
Source-stated
Thoughtworks says GitClear research found increased code churn and a decrease in lines moved, which the article associates with less refactoring and attention to code quality.
Research done by GitClear earlier this year indicates that with AI coding assistants (like GitHub Copilot) going mainstream, code churn — which GitClear defines as “changes that were either incomplete or erroneous when the author initially wrote, committed, and pushed them to the company’s git repo"
Source-stated
Thoughtworks argues that developers’ work includes evolving existing software and that developer judgment will be important as AI-generated code proliferates.
A large proportion of their work is evolving and adapting existing (and sometimes legacy) software. Sprawling and janky code bases that have taken on technical debt are, unfortunately, the norm.
Source-stated
Thoughtworks says its success using Claude Code to add language support to an internal tool was mixed, while it sees promise in using generative AI to understand legacy codebases.
In fact, Anthropic’s Claude Code helped us add support for new languages in an internal tool, CodeConcise. We use CodeConcise to understand legacy systems; and while our success was mixed, we do think there’s real promise here.
Source-stated
Thoughtworks argues that developers will have responsibility for ensuring confidence in AI-built systems, even if verification is AI-assisted.
the job of verifying that an AI-built system is correct will fall to humans. Yes, verification itself might be AI-assisted, but it will be the role of the software developer to ensure confidence.
Source-stated
Thoughtworks concludes that software development should be valued and that thoughtless automation could be problematic and costly in the years to come.
It’s vital software development is valued: the impact of thoughtless automation and pure vibes could prove incredibly problematic (and costly) in the years to come.
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.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: df01bdcb-43db-a02d-7bed-fa8facee85e1.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T18:48:47.520Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 502513c03ea8d4638507b4c9b0ad9d2de864b09800799a9d69569868b2400e24. Source text hash: 6b66253402dcb9c02dbd0d35ae03f0ed3a68c3e6bed6151c45fec6137df0c5f1.
uncertain · 2026-11-14T18:48:47.520Z · anthropic/claude-opus-5.5
Publication: df01bdcb-43db-a02d-7bed-fa8facee85e1
attribution — pass: The reading consistently attributes the argument to Thoughtworks. It states that the piece is sponsored content on MIT Technology Review and notes that MIT Technology Review's editorial staff did not write it. GitClear's findings are reported as the article's citation of GitClear, not as established fact.
This content was produced by Thoughtworks. It was not written by MIT Technology Review’s editorial staff.
summary — uncertain: Central evidence terms are left unexplained, and the definitions list is empty. The summary cites increased 'code churn' and 'fewer lines moved' as signs of less 'refactoring' but never explains them. The source gives definitions for both, such as care and feeding to make code more effective. Smallest repair: add short plain-language glosses or term tooltips. The summary also softens the source's 'will lead to' into 'can leave' and drops 'vibe coding' and 'more error-strewn', which weakens the author's force. The title and question are otherwise faithful.
a signal for refactored code (essentially the care and feeding to make it more effective)
date — pass: April 2, 2025 matches the UTC calendar day of originalPublished (2025-04-02T17:50:55Z). The Most Popular sidebar contains later items, such as the ten years after AlphaGo story, but the reading does not use them. The page is a publisher page, not an archive copy. modifiedAt is one second after publication. There is no hindsight framing.
support — pass: Every claim matches its excerpt. The source's sentence order leaves it slightly ambiguous whether 'success was mixed' refers to the Claude Code work or to using CodeConcise on legacy systems. The reading's phrasing is a reasonable rendering of that one sentence. The closing unknown and the question do not overstate the evidence.
Anthropic’s Claude Code helped us add support for new languages in an internal tool, CodeConcise. We use CodeConcise to understand legacy systems; and while our success was mixed
media — pass: No media was selected and no verified candidates were supplied, so there is nothing to misidentify.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore Claude