6 min read

Microsoft and its pivots

Microsoft and its pivots
Speedboats

When I joined Microsoft in 1988, OS/2 was going to be the next big OS platform, and most people in the Systems Division were working on OS/2 or related projects. I had early versions of OS/2 on my desktop machine, and all the development teams I worked with were targeting OS/2.   Over the next several years, this plan unraveled as it became clear that Windows was a better product.   The small Windows team had made great progress — running a GUI and full MS-DOS virtualization in a much smaller memory footprint than OS/2.  Windows was making progress in the marketplace, more and more Windows software was launching. Internally, you could see the slow erosion of OS/2 commitment; people started installing Windows on their machines. And then, seemingly all at once, the strategy shift was announced, and everyone was full steam ahead on Windows.  Within hours, it seemed like everyone re-imaged their hard disks and reset their team strategy.

Microsoft was a medium-sized company at this point, but it wasn't a monolith. Many people used a "speedboat" metaphor to describe the company: Microsoft wasn’t an aircraft carrier, but was a fleet of speedboats.  And first a few would turn, then a few more, and then, like birds flocking, the whole company would pivot.  Even though the company was no longer small, it still could pivot nimbly.

In the 1994-95 timeframe, I lived through and had a hand in the internet pivot in the company.   And a similar dynamic happened.  A few people started to install IP stacks and internet clients and talk about them, then a few more, then some product teams began to pivot, and then it was just a landslide of changes.  It was chaos internally for a while — multiple teams were building browsers (and more teams wanted to), several teams were building server platforms, and many were building authoring tools.  But it all got sorted out, and the company settled on a coherent strategy.  

Microsoft is midway through its next big pivot, to AI.  Every week, new things are announced.  Satya posted about Microsoft's most recent AI capabilities, which all sound cool:

  • Researcher and Analyst agents to help you dig into topics.  Sounds akin to ChatGPT’s Deep Research
  • Notebooks.  Similar to Google’s NotebookLM, I love the idea of grounding an agent in exactly the context for my project.
  • Search.  Least excited about this.
  • Create.   The one idea mentioned, turning a PPT into an explainer video, I would love that.   

So I clicked through the announcement links to try it all out.  What a bad experience — not because the capabilities are poor, but because finding and using the features is nearly impossible.

Copilot isn't a thing, but is a bag of things

The announcement links to a Microsoft 365 Copilot marketing page, which repeats many of Satya’s details, adds some more, and provides links.  Navigating all the links introduces a flurry of Microsoft projects and concepts:

  • The Microsoft 365 Copilot app (formerly Office). The site points you to the Microsoft macOS downloader, which does not mention Copilot nor install anything Copilot-branded; it just installs all the normal Office apps. I don’t understand the phrasing of this link. Is the Copilot app replacing Office?
  • Eventually, I found a Microsoft Copilot app in the Mac App Store. It’s weird that it is easy to find in the App Store but impossible to find via Microsoft’s Copilot web pages. So I installed it. It seems fine, but where are all the new things Satya talks about?
  • Copilot Chat — is this different than Copilot?   The chat URL takes me to a Copilot page?  But various Microsoft sites seem to imply they are something different — “…With Microsoft 365 Copilot and the new Microsoft 365 Copilot Chat…”?
  • Where are Copilot Notebooks?   Oh, these are apparently not available for personal accounts.
  • Where is the Agent Store, including the Researcher, Analyst, and Skills agents? This is also not available to personal accounts.
  • Copilot Search is just part of Bing, I guess.  Is Copilot Search just the AI search summary answers, or is it a different thing?   And when I click on the Copilot link on the Bing page and am transported to the Copilot site, why is it different from the Copilot page for M365/Office?
  • Satya talks about the new Create features — is this a separate app, or just part of Copilot?  Or just part of the individual office apps?   I am not sure where to turn a PowerPoint deck into an explainer video, as he mentions — there is no Copilot tab or button in PowerPoint, and when I ask Copilot to help me turn a PPT into an explainer video, it was no help.
  • Microsoft Copilot Studio.  I probably can’t get this.  Following the licensing options docs, you are introduced to the Microsoft Power Platform, the US Government Community Cloud, the Microsoft 365 admin center, Copilot Studio for Teams, the Digital Messaging and Chat add-ons for Dynamics 365 Customer Service.  I apparently qualify for none of these.
  • Copilot Control System. This is also hidden behind one of the licensing schemes above.
  • Copilot Analytics, including the Copilot Studio Agents Report in Microsoft Viva Insights.  Also unavailable to me due to licensing.
  • The Frontier Program.  This program grants access to some of these capabilities; there is no way I qualify for this program.

I’m not alone in struggling through all this — Ethan Mollick pointed out this week how hard it is to get access to Microsoft’s AI features.  It is chaotic, and I don’t understand the strategy — Microsoft should want every early adopter to try these capabilities out.  It is no wonder ChatGPT is crushing Copilot — it is so much easier to install and try ChatGPT, all the features are right there, and you can easily upgrade or downgrade your license as needed. 

All this confusion is Microsoft’s own creation. Is Copilot a product or a feature of all products?  Every Microsoft group seems to whack on some Copilot goodness — even Microsoft games are getting a Copilot.  Copilot as a feature is confusing — every product group implements it differently, and Copilot features are stuffed into existing products in non-uniform ways. This is compounded by hiding Copilot access behind a thicket of diverse licensing regimes across all the products.

I can only imagine the internal organizational wars around Copilot.   Right now, it seems like Microsoft is showing all the org chart fractures in its Copilot delivery.  I am hopeful that, as in past pivots, the company will straighten it all out.  

As for me, I am heading back to ChatGPT.

Shorts

An interesting thread from Dare on what he wants out of AI tools from Microsoft.  Perhaps Microsoft is not even building the right features.

A Deep Peek Into DeepSeek AI’s Talent And Implications For US Innovation | Hoover Institution A Deep Peek Into DeepSeek AI’s Talent And Implications For US Innovation — another measure of how the US is allowing its educational system and talent pipeline to decay.  This should be a red alarm for policymakers.   We need to invest more in research, invest more in institutions, and win the global immigration war for the best talent.  We are doing none of that — yet another story on how we are tearing apart our research capacity.

A well-argued paper positioning AI as just another technology, not some world-threatening phase change.  This is a very useful lens; it is not productive to push panic and disaster.

The Impossible Four-Hour Test You Need to Pass to Become a Tariff Pencil Pusher.  A lot of work and economic activity are tied up in navigating the tariff system — and injecting uncertainty into the system is not making it any easier.

Five-Minute EV Charging Is Here, but Not for U.S.-Made Cars.  We are so far behind on EVs.

America Underestimates the Difficulty of Bringing Manufacturing Back.   Some sane policy recommendations are at the end here:

  • Invest in education.   The good jobs of the future require an educated workforce.
  • Invest in healthcare.   We need a healthy workforce with less downtime.
  • Focus on high-end manufacturing.
  • Increase requirements for foreign companies to site operations and ownership in the US.
  • Attract great manufacturing immigrants.
  • Phase in tariffs slowly so that companies can react.

Noah Smith on building for the future:  

“Squabbling over whose people “built America”, and trying to keep out anyone who doesn’t spring from that stock, is a fool’s errand; instead, we should be welcoming and encouraging the people who will keep building America tomorrow, no matter what part of the globe they hail from. Fighting over whose ancestors were greater is useless; instead, we should resolve to create a great line of descendants. Rather than restoring our societies to the (real or imagined) glories of the past, we should expect to build our societies to peaks never imagined.
That task is what we should take pride in. That is what we owe to the future, and to those who came before.”

Skyrmions kind of break my brain.  Once upon a time, I knew enough quantum science so that I could just barely wrap my head around synthetic quantum objects like these, but those days are long past.

Should I want an Impulse range?   Seems very cool.

Closing

And I will close with a poem that is insightful and hilarious: We Who Are Your Closest Friends by Phillip Lopate - Poems | Academy of American Poets.