# Ethan Mollick argues in One Useful Thing that GPT-5.5 marks a step in AI progress [Canonical reading](https://escapemeridian.com/readings/ethan-mollick-argues-in-one-useful-thing-that-gpt-5-5-marks-a-027177e1) Ethan Mollick’s piece was published on April 23, 2026. He argues GPT-5.5 matters because AI capabilities continue to improve, while its strengths remain uneven. In one coding challenge, GPT-5.5 Pro modeled an evolving harbor town; Mollick says an earlier model only generated replacement buildings. He also used Codex powered by GPT-5.5 to turn old crowdfunding data into an academic paper. Mollick says its literature review and statistics were real, but he found the hypothesis uninteresting and raised concerns about causation. A generated roleplaying game also impressed him technically, though its fiction showed familiar weaknesses: repetitive dialogue, overworked prose and ideas that did not fully pay off. His argument is about progress across models, apps and harnesses, the tools and connections that let AI work on tasks. Together, he says, they are making AI more effective at solving real problems, but they do not remove the jagged frontier: capable outputs can still miss what makes a hypothesis interesting or fiction compelling. Mollick concludes that GPT-5.5 is a noteworthy step, not the end of the process, and says capability gains appear to be accelerating. How far these combined tools can take on consequential work remains unknown, especially when evaluating the result still calls for human judgment. 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 One Useful Thing · Article Original source date: Apr 23, 2026 (2026-04-23T20:00:38.000Z). Date basis: Date supplied by the publisher’s feed. Added to Escape Meridian: 2026-11-07T07:03:24.343Z. First observed: 2026-09-23T05:35:06.471Z. Retrieved: 2026-10-23T20:47:42.423Z. - [Sign of the future: GPT-5.5](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Limits Text as retrieved on October 23, 2026; originally published April 23, 2026. Mollick says he had early access and notes that he did not know all launch details while writing; the account is his evaluation, not an independent assessment of GPT-5.5. ## Open question How far can these combined tools take on consequential work when evaluating the result still depends on human judgment? ## Source-stated Ethan Mollick says GPT-5.5 is significant because AI is still improving and its abilities remain uneven. > It is a big deal because it indicates that we are not done with the rapid improvement in AI. It is also a big deal because it is just plain good. And it is a big deal because even with all of this, the frontier of AI ability remains jagged. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated In Mollick’s harbor-town coding challenge, GPT-5.5 Pro modeled an evolving town, and completed the task faster than GPT-5.4 Pro. > only GPT-5.5 Pro actually modelled an evolving town, rather than just generating new building replacements over time. GPT-5.5 Pro is also much faster than its previous iteration: GPT-5.4 Pro took 33 minutes to complete the task, GPT-5.5 Pro took 20. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick says the paper produced from his crowdfunding files had real statistics and literature, but he questioned the hypothesis and raised concerns about causation. > the literature review is all real, as are the statistics. Instead, it is because, as an expert, I think the hypothesis is not that interesting and there are some standard concerns about causation - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick describes the generated roleplaying game as technically impressive, while noting that its fiction retained familiar weaknesses. > Every generation of AI models has struggled with actually building long-form fiction. If you are a frequent reader of AI writing you see the same problems here: a love of the uncanny; overly complex ideas that do not fully pay off - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick concludes that GPT-5.5 is a noteworthy step rather than the end of AI progress, and says capability gains appear to be accelerating. > GPT-5.5 is clearly not the end of this process, but it is a noteworthy step along the way. I have been writing this newsletter for over three years now, and the pattern has not changed - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated The piece is dated April 23, 2026. > Ethan Mollick > Apr 23, 2026 > I had early access to GPT-5.51, and I think it is a big deal. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source and contributor context One Useful Thing · 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. ## Ethan Mollick Ethan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making. - [Explore this context](/entities/ethan-mollick) - [Reference](https://mgmt.wharton.upenn.edu/profile/emollick/) - [Reference](https://www.oneusefulthing.org/) ## One Useful Thing One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice. - [Explore this context](/entities/one-useful-thing) - [Reference](https://www.oneusefulthing.org/) ## Editorial provenance Version 1. Publication ID: cf3fac33-8daa-8d17-90ce-1f145666bca4. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-07T06:58:43.971Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: f90c317646ac05162eb28c817d20290ca190139b8893213d4fd38bcd9f0ed271. Source text hash: 81e996b947398dce3c3444f4079ac05a5df03099be21b5d451f22a2c62a19026. ## Revision and source-check history ### Version 1 — 2026-11-07T07:03:24.343Z Ethan Mollick’s piece was published on April 23, 2026. He argues GPT-5.5 matters because AI capabilities continue to improve, while its strengths remain uneven. In one coding challenge, GPT-5.5 Pro modeled an evolving harbor town; Mollick says an earlier model only generated replacement buildings. He also used Codex powered by GPT-5.5 to turn old crowdfunding data into an academic paper. Mollick says its literature review and statistics were real, but he found the hypothesis uninteresting and raised concerns about causation. A generated roleplaying game also impressed him technically, though its fiction showed familiar weaknesses: repetitive dialogue, overworked prose and ideas that did not fully pay off. His argument is about progress across models, apps and harnesses, the tools and connections that let AI work on tasks. Together, he says, they are making AI more effective at solving real problems, but they do not remove the jagged frontier: capable outputs can still miss what makes a hypothesis interesting or fiction compelling. Mollick concludes that GPT-5.5 is a noteworthy step, not the end of the process, and says capability gains appear to be accelerating. How far these combined tools can take on consequential work remains unknown, especially when evaluating the result still calls for human judgment. ## Original source One Useful Thing · Article Original source date: Apr 23, 2026 (2026-04-23T20:00:38.000Z). Date basis: Date supplied by the publisher’s feed. Added to Escape Meridian: 2026-11-07T07:03:24.343Z. First observed: 2026-09-23T05:35:06.471Z. Retrieved: 2026-10-23T20:47:42.423Z. - [Sign of the future: GPT-5.5](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Limits Text as retrieved on October 23, 2026; originally published April 23, 2026. Mollick says he had early access and notes that he did not know all launch details while writing; the account is his evaluation, not an independent assessment of GPT-5.5. ## Open question How far can these combined tools take on consequential work when evaluating the result still depends on human judgment? ## Source-stated Ethan Mollick says GPT-5.5 is significant because AI is still improving and its abilities remain uneven. > It is a big deal because it indicates that we are not done with the rapid improvement in AI. It is also a big deal because it is just plain good. And it is a big deal because even with all of this, the frontier of AI ability remains jagged. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated In Mollick’s harbor-town coding challenge, GPT-5.5 Pro modeled an evolving town, and completed the task faster than GPT-5.4 Pro. > only GPT-5.5 Pro actually modelled an evolving town, rather than just generating new building replacements over time. GPT-5.5 Pro is also much faster than its previous iteration: GPT-5.4 Pro took 33 minutes to complete the task, GPT-5.5 Pro took 20. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick says the paper produced from his crowdfunding files had real statistics and literature, but he questioned the hypothesis and raised concerns about causation. > the literature review is all real, as are the statistics. Instead, it is because, as an expert, I think the hypothesis is not that interesting and there are some standard concerns about causation - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick describes the generated roleplaying game as technically impressive, while noting that its fiction retained familiar weaknesses. > Every generation of AI models has struggled with actually building long-form fiction. If you are a frequent reader of AI writing you see the same problems here: a love of the uncanny; overly complex ideas that do not fully pay off - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated Mollick concludes that GPT-5.5 is a noteworthy step rather than the end of AI progress, and says capability gains appear to be accelerating. > GPT-5.5 is clearly not the end of this process, but it is a noteworthy step along the way. I have been writing this newsletter for over three years now, and the pattern has not changed - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source-stated The piece is dated April 23, 2026. > Ethan Mollick > Apr 23, 2026 > I had early access to GPT-5.51, and I think it is a big deal. - [Supporting source](https://www.oneusefulthing.org/p/sign-of-the-future-gpt-55) ## Source and contributor context One Useful Thing · 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. ## Ethan Mollick Ethan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making. - [Explore this context](/entities/ethan-mollick) - [Reference](https://mgmt.wharton.upenn.edu/profile/emollick/) - [Reference](https://www.oneusefulthing.org/) ## One Useful Thing One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice. - [Explore this context](/entities/one-useful-thing) - [Reference](https://www.oneusefulthing.org/) ## Editorial provenance Version 1. Publication ID: cf3fac33-8daa-8d17-90ce-1f145666bca4. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-07T06:58:43.971Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: f90c317646ac05162eb28c817d20290ca190139b8893213d4fd38bcd9f0ed271. Source text hash: 81e996b947398dce3c3444f4079ac05a5df03099be21b5d451f22a2c62a19026. ### uncertain — 2026-11-07T06:58:43.971Z — anthropic/claude-opus-5.5 Publication: cf3fac33-8daa-8d17-90ce-1f145666bca4 attribution — pass: This is a single essay by Ethan Mollick in One Useful Thing. The reading attributes the assessments, examples and conclusions to him and frames them as his evaluation, including his early-access disclosure. > I had early access to GPT-5.51, and I think it is a big deal. summary — uncertain: The paper passage keeps only the caveats about the uninteresting hypothesis and causation. It drops Mollick's strongest verdict: he would have been very happy with the paper as a 2nd-year PhD project, and it took only four prompts. This tilts his assessment more negative and flattens his force. Smallest repair: add his PhD-project comparison and the four-prompt detail. Minor clarity gap: Codex and apps are not explained as desktop AI products, though the essential meaning is still readable. > In short, I would have been very happy if this paper was the outcome of a 2nd year PhD project. And I just gave it four prompts, without ever touching the text myself. date — pass: originalPublished is 2026-04-23T20:00:38Z, which is April 23, 2026 in UTC and matches the dateline. The limitations note the later October 2026 retrieval. The reading uses no hindsight framing, and the acceleration claim stays attributed to Mollick. The evidence is a publisher page, so no archive disclosure is needed. > Ethan Mollick Apr 23, 2026 support — pass: Each claim is grounded in the source: the harbor-town comparison and its timings, the paper's real literature review and statistics alongside his caveats, the fiction weaknesses, and the noteworthy-step and accelerating conclusion. The closing question about human judgment follows from his expert critique and is framed as unknown, not asserted. > with the latest releases, capability gains appear to be accelerating. media — pass: No media was 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/027177e1d4a24237d863202369ac28ebcbdd8fa2da7d47fec223b81e86e92c6b/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/ethan-mollick-argues-in-one-useful-thing-that-gpt-5-5-marks-a-027177e1.json)