# MIT Technology Review reports that ChatGPT search points toward AI agents

[Canonical reading](https://escapemeridian.com/readings/mit-technology-review-reports-that-chatgpt-search-points-toward-8975be7e)

MIT Technology Review's November 5, 2024, report describes OpenAI's ChatGPT search and Realtime API updates as steps toward AI agents: assistants that can carry out connected tasks using information and software tools. OpenAI's Olivier Godement imagines agents that know people's preferences and work across their emails, apps, and calendars; he says voice could help when people are not typing or looking at a screen.

Godement identifies reasoning and connecting tools as hurdles. OpenAI's o1 uses reinforcement learning and chain of thought, but Godement says models need to become more reliable, faster, and cheaper, and the technique needs to cover more fields. University of Washington professor Chirag Shah disputes that these systems genuinely reason, arguing that they imitate logic learned from training data and can break under scrutiny. On tool use, Godement says o1 is not very reliable, while describing research in the area as promising.

Godement expects adoption for customer support and other assistant-based tasks to grow in the next year, but says use cases are hard to predict and may surprise. The report presents agents as a direction with unresolved reliability and tool-use challenges, not as a demonstrated ability to complete complex tasks dependably.

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: Nov 5, 2024 (2024-11-05T10:58:48Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-16T15:58:24.199Z. First observed: 2026-10-31T04:12:00.473Z. Retrieved: 2026-10-31T04:13:52.757Z.

- [How ChatGPT search paves the way for AI agents](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Limits

Text as retrieved on October 31, 2026; originally published November 5, 2024. The report recounts OpenAI leaders' descriptions and plans, but supplies no measured evidence that agents can complete complex tasks dependably. Chirag Shah's criticism is presented as his view; the report does not resolve the disagreement about whether the models reason.


## Open question

What level of reliability will AI agents need to complete complex tasks safely, and how will that be measured?


## Source-stated

MIT Technology Review reports that OpenAI announced ChatGPT search, which lets users search the internet using a chatbot.

> OpenAI announced it was launching ChatGPT search, which allows users to search the internet using the chatbot.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Olivier Godement describes a future in which an agent knows a person's preferences and can work across their email, apps, and calendars.

> That agent knows you extremely well. It knows your preferences,” Godement says. The agent will have access to your emails, apps, and calendars

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement identifies reasoning and connecting different tools as two hurdles for agents.

> But there are two big hurdles that need to be overcome before agents can become a reality, Godement says.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Chirag Shah argues that large language models imitate reasoning rather than genuinely reason, and may fail when examined closely.

> These models sometimes seem to be really amazing at reasoning, but it’s just like they’re really good at pretending, and it only takes a little bit of picking at them to break them

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement says o1 can use tools only somewhat reliably and calls research on tool use promising.

> Godement says o1 can “sort of” use tools, though not very reliably, and that research on tool use is a “promising development.”

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement expects adoption of AI for customer support and other assistant-based tasks to grow in the next year, while saying future use cases are difficult to predict.

> In the next year, Godemont says, he expects the adoption of AI for customer support and other assistant-based tasks to grow. However, he says that it can be hard to predict how people will adopt and use OpenAI’s technology.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Escape Meridian interpretation

Escape Meridian reads the report as presenting AI agents as a direction whose dependable completion of complex tasks remains unresolved.

> Building AI agents requires us to be able to trust that they will be able to complete complex tasks and do the right things, says Huet.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## 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.


## AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

- [Definition reference](https://www.ibm.com/think/topics/ai-agents)

## API

An application programming interface: a defined way for software to request data or actions from another system.

- [Definition reference](https://developer.mozilla.org/en-US/docs/Glossary/API)

## Reinforcement learning

Training by trial and error, in which a system is rewarded for the outcomes its designers want.

- [Definition reference](https://developers.google.com/machine-learning/glossary#reinforcement-learning-rl)

## Chain of thought

The step-by-step working a model writes out before giving its final answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#chain-of-thought-prompting)

## Tool use

A model's ability to operate other software, such as a search engine, a calculator or a code runner, while it works.

- [Definition reference](https://docs.anthropic.com/en/docs/build-with-claude/tool-use/overview)

## OpenAI

OpenAI develops AI models and products and publishes research, release announcements and evaluations.

- [Explore this context](/entities/openai)
- [Reference](https://openai.com/news/)

## Editorial provenance

Version 1. Publication ID: 2d2e8347-a1c2-55db-5bc8-afc607788348.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T15:53:43.638Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model.

Observation: 22a907ac534c1a9b9bf67ca70d1bee2af8dd59254cb6c197b23c115f3306d4a2. Source text hash: 86a3fa1fb0e97ed5c3a5077ba82df2b96164ff186bef74628c853a6a85440802.


## Revision and source-check history

### Version 1 — 2026-11-16T15:58:24.199Z

MIT Technology Review's November 5, 2024, report describes OpenAI's ChatGPT search and Realtime API updates as steps toward AI agents: assistants that can carry out connected tasks using information and software tools. OpenAI's Olivier Godement imagines agents that know people's preferences and work across their emails, apps, and calendars; he says voice could help when people are not typing or looking at a screen.

Godement identifies reasoning and connecting tools as hurdles. OpenAI's o1 uses reinforcement learning and chain of thought, but Godement says models need to become more reliable, faster, and cheaper, and the technique needs to cover more fields. University of Washington professor Chirag Shah disputes that these systems genuinely reason, arguing that they imitate logic learned from training data and can break under scrutiny. On tool use, Godement says o1 is not very reliable, while describing research in the area as promising.

Godement expects adoption for customer support and other assistant-based tasks to grow in the next year, but says use cases are hard to predict and may surprise. The report presents agents as a direction with unresolved reliability and tool-use challenges, not as a demonstrated ability to complete complex tasks dependably.

## Original source

MIT Technology Review · Article

Original source date: Nov 5, 2024 (2024-11-05T10:58:48Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-16T15:58:24.199Z. First observed: 2026-10-31T04:12:00.473Z. Retrieved: 2026-10-31T04:13:52.757Z.

- [How ChatGPT search paves the way for AI agents](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Limits

Text as retrieved on October 31, 2026; originally published November 5, 2024. The report recounts OpenAI leaders' descriptions and plans, but supplies no measured evidence that agents can complete complex tasks dependably. Chirag Shah's criticism is presented as his view; the report does not resolve the disagreement about whether the models reason.


## Open question

What level of reliability will AI agents need to complete complex tasks safely, and how will that be measured?


## Source-stated

MIT Technology Review reports that OpenAI announced ChatGPT search, which lets users search the internet using a chatbot.

> OpenAI announced it was launching ChatGPT search, which allows users to search the internet using the chatbot.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Olivier Godement describes a future in which an agent knows a person's preferences and can work across their email, apps, and calendars.

> That agent knows you extremely well. It knows your preferences,” Godement says. The agent will have access to your emails, apps, and calendars

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement identifies reasoning and connecting different tools as two hurdles for agents.

> But there are two big hurdles that need to be overcome before agents can become a reality, Godement says.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Chirag Shah argues that large language models imitate reasoning rather than genuinely reason, and may fail when examined closely.

> These models sometimes seem to be really amazing at reasoning, but it’s just like they’re really good at pretending, and it only takes a little bit of picking at them to break them

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement says o1 can use tools only somewhat reliably and calls research on tool use promising.

> Godement says o1 can “sort of” use tools, though not very reliably, and that research on tool use is a “promising development.”

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Source-stated

Godement expects adoption of AI for customer support and other assistant-based tasks to grow in the next year, while saying future use cases are difficult to predict.

> In the next year, Godemont says, he expects the adoption of AI for customer support and other assistant-based tasks to grow. However, he says that it can be hard to predict how people will adopt and use OpenAI’s technology.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## Escape Meridian interpretation

Escape Meridian reads the report as presenting AI agents as a direction whose dependable completion of complex tasks remains unresolved.

> Building AI agents requires us to be able to trust that they will be able to complete complex tasks and do the right things, says Huet.

- [Supporting source](https://www.technologyreview.com/2024/11/05/1106603/how-chatgpt-search-paves-the-way-for-ai-agents/)

## 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.


## AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

- [Definition reference](https://www.ibm.com/think/topics/ai-agents)

## API

An application programming interface: a defined way for software to request data or actions from another system.

- [Definition reference](https://developer.mozilla.org/en-US/docs/Glossary/API)

## Reinforcement learning

Training by trial and error, in which a system is rewarded for the outcomes its designers want.

- [Definition reference](https://developers.google.com/machine-learning/glossary#reinforcement-learning-rl)

## Chain of thought

The step-by-step working a model writes out before giving its final answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#chain-of-thought-prompting)

## Tool use

A model's ability to operate other software, such as a search engine, a calculator or a code runner, while it works.

- [Definition reference](https://docs.anthropic.com/en/docs/build-with-claude/tool-use/overview)

## OpenAI

OpenAI develops AI models and products and publishes research, release announcements and evaluations.

- [Explore this context](/entities/openai)
- [Reference](https://openai.com/news/)

## Editorial provenance

Version 1. Publication ID: 2d2e8347-a1c2-55db-5bc8-afc607788348.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T15:53:43.638Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model.

Observation: 22a907ac534c1a9b9bf67ca70d1bee2af8dd59254cb6c197b23c115f3306d4a2. Source text hash: 86a3fa1fb0e97ed5c3a5077ba82df2b96164ff186bef74628c853a6a85440802.


### passed — 2026-11-16T15:53:43.638Z — anthropic/claude-opus-5.5

Publication: 2d2e8347-a1c2-55db-5bc8-afc607788348

attribution — pass: The piece is attributed to MIT Technology Review, which matches the publisher and the newsletter framing. Statements are correctly assigned: the vision, hurdles, tool-use comments and adoption forecast to Godement, and the reasoning critique to Chirag Shah. The page is one article plus a newsletter tail, not a bare listing.

> OpenAI’s strategy is to both build agents itself and allow developers to use its software to build their own agents, says Godement.

summary — pass: The reading leads with the report's argument that ChatGPT search and the Realtime API point toward agents, and explains agents in plain terms. It conveys the two hurdles, Shah's forceful dissent and Godement's admitted uncertainty. AI agent, API, reinforcement learning, chain of thought and tool use are covered by page definitions. The question about required reliability and how to measure it follows from the trust hurdle the source raises.

> There are some use cases where you’re not typing, not even looking at the screen, and so voice essentially has a much better modality for that

date — pass: November 5, 2024 matches the UTC day of originalPublished, and the reading notes the later retrieval. The kind is publisher-page, so no archive disclosure is required. The adoption expectation is kept as Godement's forecast. Later sidebar items, such as the AlphaGo ten-year piece, are possible later additions; the reading neither uses nor relies on them.

> In the next year, Godemont says, he expects the adoption of AI for customer support and other assistant-based tasks to grow.

support — pass: Each claim is supported by a matching excerpt. Godement's caveat that o1 is not very reliable at tool use is preserved, and the reading does not assert any demonstrated agent benefit. The interpretation is labeled as Escape Meridian's and grounded in Huet's trust requirement. The question is open rather than leading.

> Godement says o1 can “sort of” use tools, though not very reliably, and that research on tool use is a “promising development.”

media — pass: No media was selected and no verified media candidates were supplied, so nothing could be misidentified.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/8975be7e956a1d8fff9c0b83fa4b43f5c71efce658392c41482d432d110cf092/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-reports-that-chatgpt-search-points-toward-8975be7e.json)
