Recommended Free Tools
Paul Thurrott’s April 15, 2024, premium analysis, “Embrace, Extend, and … Copilot?”, argues that Microsoft’s AI rollout echoes its response to the Internet in the 1990s: adopt a major new technology, integrate it into existing products, and try to help shape what comes next. The comparison is about platform strategy—not proof that Copilot is a deliberate “embrace, extend, and extinguish” scheme.
What Thurrott’s title is asking
The ellipsis invites a comparison between Microsoft’s old Internet strategy and its current use of Copilot. Thurrott’s answer is qualified: the resemblance is real at the level of platform and distribution strategy, but the analogy does not establish that Microsoft is repeating an anticompetitive playbook.
The article is a premium analysis by Paul Thurrott, published on Thurrott.com on April 15, 2024. “Premium” labels access to the commentary; it is not part of the argument and does not refer to a separate product called Copilot Premium.
What “embrace and extend” originally meant
Allard’s 1994 proposal
Thurrott traces the phrase to Microsoft engineer J Allard’s 1994 internal memo, “Windows: The Next Killer Application on the Internet.” In the account presented in the article, Allard urged Microsoft to respond to a fast-growing Internet by embracing its culture, users, technologies, and standards; extending Windows with compatible Internet tools and services; and then innovating toward a leadership role.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match#1 Best Overall
The sequence matters: the memo’s formulation was “embrace, extend, then innovate,” not “embrace, extend, and extinguish.” Thurrott presents it as an effort to help Microsoft catch up with the Internet, not as an inherently destructive plan.
How the phrase acquired a darker meaning
During the late-1990s browser and standards battles, “embrace and extend” became associated with concerns about Microsoft’s integration of Internet Explorer with Windows and the competitive pressure on Netscape. Thurrott notes that public debate also drew on testimony and reported internal language using formulations such as “embrace, extend, and extinguish” and “embrace and smother.” Those hostile formulations should not be retroactively treated as the wording of Allard’s memo.
The distinction is useful beyond Microsoft history. Adopting a technology and adding features is not, by itself, the same thing as excluding competitors. Whether integration becomes anticompetitive depends on conduct and effects, not just a slogan.
Rank #2
Microsoft’s “AI beside, AI inside, AI outside” model
Thurrott connects his comparison to a framework attributed to Microsoft executive Stevie Batiche at the company’s May 2023 Build event. The categories describe ways of integrating AI, not a fixed product roadmap. The examples below are those discussed in Thurrott’s article, not a complete inventory of Microsoft products.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
AI beside
AI is placed alongside familiar software and workflows, so a user can call on it without leaving an existing environment. Thurrott points to Copilot in Windows 11 and Microsoft 365, as well as experiences in or alongside Word and Excel.
AI inside
AI becomes part of how a product is built and used, rather than simply appearing as an added chat window. Clipchamp and Designer are examples cited in the article.
Rank #3
AI outside
AI could act across applications and services, coordinating tools to carry out a user’s goal. Thurrott describes the possibility as a “copilot of copilots.” This is a future-facing direction in the article, not evidence that a mature, broadly available system can already reliably orchestrate every workflow.
How the Internet and AI strategies compare
| Internet-era framing in Allard’s memo | AI-era framing discussed by Thurrott |
|---|---|
| Embrace Internet culture, users, technologies, and standards. | AI beside: put AI alongside applications people already use. |
| Extend Windows with Internet-compatible tools and services. | AI inside: build products such as Clipchamp and Designer around AI capabilities. |
| Innovate toward a leadership role in shaping what comes next. | AI outside: explore AI coordination across applications and services. |
This is a conceptual mapping, not a literal match. Allard’s proposal described phases intended to happen in order. “AI beside,” “AI inside,” and “AI outside” are overlapping implementation patterns; they can coexist rather than replace one another in sequence.
Why the analogy is useful—and where it stops
What carries over
In both cases, Microsoft is responding to a major technology shift and using an established software ecosystem as a route to adoption. Integrating a new capability into familiar products can reduce friction for users and give a platform owner a significant distribution advantage. Thurrott’s point is that Microsoft is not merely adding an AI feature: it is positioning its products and services around a broader change in computing.
Rank #4
What changes with AI
AI that only assists within one application is different from AI that acts across several. Cross-application orchestration depends on permissions, identity, data access, and working integrations with the services involved. It also raises the stakes of mistakes: an incorrect answer is one problem; an incorrect action carried out across systems can have wider consequences.
Thurrott draws on Steven Sinofsky’s Hardcore Software: Inside the Rise and Fall of the PC Revolution as a counterweight to the idea that “embrace and extend” is inherently sinister. As Thurrott recounts Sinofsky’s view, technology companies commonly build on existing standards and products while adding value for customers. Sinofsky’s further point is that the Internet was larger than Windows: Windows could not contain it, and could instead become one application running on the Internet.
The article also invokes Netscape’s development from earlier browser work associated with Mosaic, and Apple’s ability to enter categories after others while differentiating through added value and user experience. These are Thurrott’s illustrative comparisons, not proof that the companies followed identical strategies or faced identical circumstances.
Best Value
Does Copilot amount to “embrace, extend, and extinguish”?
No such conclusion follows from the article’s analogy. Copilot can be understood as one visible way Microsoft is integrating AI into its ecosystem, much as Windows was used to deliver Internet-era capabilities. But integration and exclusion are different claims. Establishing an anticompetitive pattern would require evidence about how Microsoft behaves toward rivals and customers, not simply the presence of Copilot in Microsoft products.
What would matter in evaluating the concern
- Defaults and choice: Can users select another assistant or change relevant defaults?
- Interoperability: Can competing assistants access comparable interfaces and services, or does Microsoft restrict that access?
- Bundling and pricing: Are customers effectively required to pay for AI they do not want, and are rivals disadvantaged by the way products are packaged?
- Data and permissions: Does Microsoft have a privileged data or context advantage, and are access and governance rules clear?
- Switching costs: Do reliance on Microsoft-specific identity, data, APIs, or workflows make it materially harder to change platforms?
- Market effects: Is there evidence of exclusion or impaired competition, rather than simply a successful integrated product?
These questions also matter to buyers, independent of any legal finding. Consumers may value convenience but want control over defaults. Businesses may find an embedded assistant useful while needing to scrutinize data access, permissions, and usage costs. Developers may gain reach through Microsoft’s ecosystem while becoming more dependent on its infrastructure and interfaces.
Why “Copilot” does not name one product
Microsoft uses the Copilot name across offerings with different audiences and capabilities; GitHub Copilot, Microsoft 365 Copilot, and Copilot Studio are not interchangeable subscriptions. The specific products and examples in Thurrott’s 2024 analysis should be read in that historical context, not as a current product or licensing guide.
For example, GitHub documents its own plan structure separately from its usage-based billing rules. Some interactions can involve AI Credits or additional usage costs, so a subscription label alone may not describe total spend. See GitHub’s Copilot plan information and billing documentation for current details. Those GitHub terms do not establish pricing or licensing for Microsoft 365 Copilot or other Microsoft offerings.
What the article ultimately argues
Thurrott uses Copilot to revisit an old Microsoft phrase and ask how companies turn platform shifts into product strategy. His comparison is strongest as a way to see the recurring role of integration and distribution: Microsoft can bring a new technology to users through software they already rely on. It is weakest when treated as a prediction that AI must follow the Internet’s path or as evidence that the company intends to extinguish rivals.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.




