[mlavrent]

I was at Microsoft until July of this year until I left for an SF-based company (not AI though).

The difference between the two with regards to AI tool usage couldn’t be more different- at Microsoft, they had started penalizing you in perf if you didn’t use the AI tools, which often were under par and you didn’t have a choice in. At the new place, perf doesn’t care if you use AI or not- just what you actually deliver. And, shocker, turns out they actually spend a lot building and getting feedback on internal AI tooling and so it gets a lot of use!

The Microsoft culture is a sort of toxic “get AI usage by forcing it down the engineer throats” vs the new “make it actually useful and win users” approach at that new place. The Microsoft approach builds resentment in the engineering base, but I’m convinced it’s the only way leadership there knows how to drive initiatives.

[stickfigure]

Microsoft is forcing the company to dogfood their own tools. They do this because they need the feedback so they can improve their tools, and they think these tools are a critical part of their future.

Presumably your new company isn't building AI tools, so they don't care what you use.

Imagine a developer in 1990s Microsoft saying "I want to use Borland C++ because it's better than the Microsoft IDE". Maybe it is, maybe it isn't, but that's not the point.

[sedawk]

This is true, but is effective only if the dogfooders' feedback is accepted and worked upon. Which is not the case, I can tell you this from first-hand experience (currently work at Msft). Also unleashing and forcing tech savvy people to use immature tools is only asking for trouble when you don't have allocated enough manpower to deal with the fallout (that is, incessant downpour of improvement feedback). One cannot just force engineers to dogfood tools and then ignore them. This is precisely like Win-8 era mania, only this time it's infecting the whole company not just single org!

[taurath]

Not only are you discouraged from criticizing half baked manager-metric led implementations, you’re deeply incentivized to openly praise it if you want to be considered for the next well-funded initiative.

People with fiefdoms don’t like criticism. Microsoft pays their vassal dependent companies to use their products, no users actually like or would choose the products (Teams? 365 copilot? Azure?), and the whole enclosed ecosystem is pretty awful.

[soco]

Count me as that weird user that would choose Azure any second over AWS. The integration and interface stability they offer is simply better. Teams sucks indeed, but as I don't know any less-suck alternative I'll have to trust you, and the with Copilot I never bothered much so again can't tell.

[sedawk]

> ...deeply incentivized to openly praise... Sadly, you're not far off from describing the reality in many cases!

[mlavrent]

They do actually build internal tooling! They key is that it’s actually good enough that feedback to the limited, targeted, and quickly actionable. Microsoft’s internal was immature enough that the general feedback you’d always have is “this is unusable”, which is something the teams building the tools could probably figure out themselves before making the whole company spend time beta testing the tools.

The main point is that the tools need to be of a certain quality/maturity for dogfooding to be effective.

[capr]

Maybe if they would've let their engineers use Borland C++ they would've learned a thing or two for their own product.

[pjmlp]

I keep telling to the WinUI marketing team that instead of talking about how "great" doing XAML C++ is, they should actually buy a copy of C++ Builder.

[roncesvalles]

Microsoft either doesn't care about feedback or doesn't have the engineering ability to act on them, otherwise Visual Studio and Microsoft Teams wouldn't be such terrible pieces of software despite tens of thousands of Microsoft employees using them daily.

[pjmlp]

In 2025, C++ Builder is still better than Visual C++, in what concerns doing Windows GUI development in C++, some things never change, and management keeps being blind to them.

Regarding dog fooding, Project Reunion was also a victim of all engines AI, now the damage is done and only the Windows team cares, because their job depends on using it.

[vkou]

Look, if you want the people to dig trenches with spoons, you can expect them to do it. But if all you're giving them is spoons, you're going to need to give them a lot of slack on the expected digging schedules.

Being forced to use a shit tool because <some other department somewhere in the company wants your feedback>, while your deadlines haven't been adjusted for all this wasted time is not acceptable behaviour. It's the kind of authoritarian horseshit that's that's so often pushed by unproductive parasites onto people who do actual work.

[OhMeadhbh]

Dogfooding is great and all, but if you're forcing your engineering staff to dogfood something, you should make sure you're in the same industry as your customers. I've always had a bit of respect for MSFT products in the "I'm a company with about 5 reasonable, but not stellar developers" space. Do I want to build an operating system with Visual Basic? No. Do I force C++ on our loading dock foreman who upskilled to a VB4 dev 'cause he knows the problem domain inside and out? Also no. MSFT traditionally attracted "above average" devs who had the support to work on big projects for a (comparatively) long time.

As J. R. "Bob" Dobbs once said, "I don't practice what I preach because I'm not the kind of person I'm preaching to." ( see https://en.wikiquote.org/wiki/J._R._%22Bob%22_Dobbs )

Maybe the engineers complaining about dogfooding vibe-coding tools aren't the kind of developers you should have vibe-coding.

[etruong42]

> At the new place, perf doesn’t care if you use AI or not- just what you actually deliver

I work at Google, and I am of the overall opinion that it doesn't matter what you deliver from an engineering perspective. I've seen launches that changed some behavior from opt-in to opt-out get lauded as worth engineering-years of investment. I've seen demos that were 1-2 years ahead of our current product performance get buried under bureaucracy and nitpicking while the public product languishes with nearly no usage. The point being, what you objectively deliver doesn't matter, but what ends up mattering is how the people in your orbit weave the narrative about what you built.

So if "leadership" wants something concretely done, they must mandate it in such a way that cuts through all the spin that each layer of bureaucracy adds before presenting it to the next layer of bureaucracy. And "leadership" isn't a single person, so you might describe leaders as individual vectors in a vector space, and a clear eigenvector in this space of leadership decisions in many companies is the vector of "increase employee usage of AI tools".

[wseqyrku]

> they had started penalizing you in perf if you didn’t use the AI tools

That is kind of insane right? They are practically mining their own people for data, one wonders what they would not do to their customers.

[indrora]

I too was at MSFT until the July layoffs.

Hang around old Microsofties and you'll encounter a phrase: "The Deal." The Deal is this informal agreement: Microsoft doesn't pay amazingly but you're given the time to have work-life balance, you can be relatively assured that upper leadership gives a shit about the ICs, there's space for "... So I was thinking..." to become real "... and that's our next product" discussions and that it's okay to fall so long as you can get back up and keep walking afterwards.

The Deal is dead.

People fired for performance after a bad review their manager didn't give them. The constant slimming of orgs and the relentless gnawing at budgets. I watched as a team went from reasonable to gutted because it got the short straw in "unregretted attrition quotas"

AI is driving this, and I want to see the chat logs between executives and copilot. What sycophantic shit is it producing that is driving them to make horrible decisions?

[roncesvalles]

The Deal died when Microsoft got on the layoff bandwagon in Q1 2023 for no good reason and became very aggressive with perf after that. If Microsoft is just as toxic and unstable as Meta, why not just work at Meta for double the money?

Funnily, Apple also has an unspoken "deal" (pay a bit low but treat really well) and they stuck to it even through the layoff era.

[taurath]

AI is busy quietly convincing every executive that uses it that they have no use for people to work out the details of their ideas anymore. It’s so frustrating to have these drive by executives come into a space you’re working in, drop in a 15 page deep think report they got from a 2 sentence prompt and call that contributing. Bonus points if the report is from an AI platform your company doesn’t have approved so you as a line employee can get written up for.

[HacklesRaised]

Is it AI, or is it being run by people entirely divorced from the founder's vision?

[balamatom]

I think it's being compliant with the founder's vision entirely.

[pjmlp]

From the outside this is clearly visible how Project Reunion crashed, C++/WinRT went into maintenance, VC++ losing steam to ISO compliance after bosting about C++20 and C11/17, .NET focus on Aspire/Blazor all AI in detriment of the rest,....

Thankfully I am technology mercenary, polyglot, and use whatever the clients need, regardless of my point of view, but it is sad to see the human part behind those decisions being affected.

[joshwa]

URA quotas—I see the Amazon infection has spread from Seattle to Redmond.

[zeristor]

Perhaps the managers performance goals are linked to take up, this sometimes happens and it all becomes too blunt.