zlacker

[parent] [thread] 43 comments
1. bramha+(OP)[view] [source] 2025-05-21 11:53:15
Seeing Microsoft employees argue with an LLM for hours instead of actually just fixing the problem must be a very encouraging sight for businesses that have built their products on top of .NET.
replies(10): >>nashas+F >>ozim+56 >>svick+o6 >>mikrl+w9 >>gwervc+Mf >>lloyda+Rl >>empath+6o >>LunaSe+U21 >>Allege+Mo1 >>pier25+cy1
2. nashas+F[view] [source] 2025-05-21 11:58:47
>>bramha+(OP)
I sometimes feel like that is the right outcome for bad management and bad instructions. Only this time they can’t blame the junior engineer and are left to only blame themselves.
replies(2): >>qoez+25 >>snacke+jA1
◧◩
3. qoez+25[view] [source] [discussion] 2025-05-21 12:31:21
>>nashas+F
They'll probably blame openai/the AI instead.
replies(1): >>nashas+N6
4. ozim+56[view] [source] 2025-05-21 12:39:34
>>bramha+(OP)
That is why they just fired 7k people so they don’t argue with LLM but let it do the work /s
5. svick+o6[view] [source] 2025-05-21 12:41:55
>>bramha+(OP)
You don't want them to experiment with new tools? The main difference now is that the experiment is public.
replies(4): >>gmm199+S7 >>stickf+5f >>flmont+Mg >>PKop+4q
◧◩◪
6. nashas+N6[view] [source] [discussion] 2025-05-21 12:45:35
>>qoez+25
AI has reproducible outcomes. If someone else can make it work, then they should too.
replies(1): >>davegu+7u
◧◩
7. gmm199+S7[view] [source] [discussion] 2025-05-21 12:54:21
>>svick+o6
I wouldn't necessarily call that just an experiment if the same requests aren't being fixed without copilot and the ai changes could get merged.

I would say the copilot system isn't really there yet for these kinds of changes, you don't have to run experiments on a language framework to figure that out.

8. mikrl+w9[view] [source] 2025-05-21 13:08:04
>>bramha+(OP)
I remember before mass LLM adoption, reading an issue on GitHub where an increasingly frustrated user was failing to properly describe a blocking issue, and the increasingly frustrated maintainer was failing to get them to stick to the issue template.

Now you don’t even need the frustrated end user!

replies(1): >>shulta+Pd
◧◩
9. shulta+Pd[view] [source] [discussion] 2025-05-21 13:38:43
>>mikrl+w9
one day both sides will be AI so we can all relax and enjoy our mojitos
replies(2): >>marcos+Ws >>some_r+4w
◧◩
10. stickf+5f[view] [source] [discussion] 2025-05-21 13:45:36
>>svick+o6
It's pretty obviously a failed experiment. Why keep repeating it? Try again in another 3 months.

The answer is probably that the Copilot team is using the rest of the engineering organization as testers. Great for the Copilot team, frustrating for everyone else.

replies(1): >>raydev+6m1
11. gwervc+Mf[view] [source] 2025-05-21 13:51:08
>>bramha+(OP)
Especially painful when one of said employee is Stephen Toub, who is famous for his .net performance blog posts.
replies(1): >>svaha1+Vj
◧◩
12. flmont+Mg[view] [source] [discussion] 2025-05-21 13:56:38
>>svick+o6
By all means. Just not on one of the most popular software development frameworks in the world. Maybe that can wait until after the concept is proven.
replies(1): >>mystif+4F
◧◩
13. svaha1+Vj[view] [source] [discussion] 2025-05-21 14:16:56
>>gwervc+Mf
I was thinking that too. He's a great programmer, and at this point I can't imagine he's having fun 'prompting' an LLM to write correct code.
replies(2): >>davegu+zu >>mock-p+MF
14. lloyda+Rl[view] [source] 2025-05-21 14:29:05
>>bramha+(OP)
That is essentially what I tried to say in my comment there but don't think they wanted to hear it.
15. empath+6o[view] [source] 2025-05-21 14:39:46
>>bramha+(OP)
The point of this exercise for Microsoft isn't to produce usable code right now, but to use and improve copilot.
replies(1): >>saati+GY1
◧◩
16. PKop+4q[view] [source] [discussion] 2025-05-21 14:49:54
>>svick+o6
Nah I'd prefer they focus on writing code themselves to improve .NET not babysitting a spam-machine
◧◩◪
17. marcos+Ws[view] [source] [discussion] 2025-05-21 15:04:45
>>shulta+Pd
Well, people have been putting M-x doctor to talk with M-x eliza for decades.
◧◩◪◨
18. davegu+7u[view] [source] [discussion] 2025-05-21 15:11:16
>>nashas+N6
This is just false. Do these models even have reproducible outcomes with a temperature of 0? Aren't they also severely restricted with a temp of 0?
replies(1): >>nashas+ZJ
◧◩◪
19. davegu+zu[view] [source] [discussion] 2025-05-21 15:13:18
>>svaha1+Vj
I hope he writes a personal essay about the experience after he leaves Microsoft. Not that he will leave anytime soon, but the first hand accounts of how they are talking about these systems internally are going to be even more entertaining than the wtf PRs.
replies(1): >>square+1L
◧◩◪
20. some_r+4w[view] [source] [discussion] 2025-05-21 15:21:11
>>shulta+Pd
when that day arrives we'll won't be relaxing, we will be put through a wood chipper
replies(1): >>disqar+mK1
◧◩◪
21. mystif+4F[view] [source] [discussion] 2025-05-21 16:12:05
>>flmont+Mg
Yeah, seems to me that breaking .NET with this garbage will be, uh, extremely bad
◧◩◪
22. mock-p+MF[view] [source] [discussion] 2025-05-21 16:16:05
>>svaha1+Vj
You don’t think he’s having fun getting laid a ton for playing with computers?
replies(1): >>xeonmc+LP
◧◩◪◨⬒
23. nashas+ZJ[view] [source] [discussion] 2025-05-21 16:42:23
>>davegu+7u
Some randomization is intentionally introduced. We are not accounting for that. Otherwise, it should be able to give you the same information.
◧◩◪◨
24. square+1L[view] [source] [discussion] 2025-05-21 16:48:56
>>davegu+zu
This comment thread is incredible. It's like fanfiction of a real person. Of course this engineer I respect shares my opinion. Not only that, he's obviously going to quit because of this. And then he'll write a blog post I'll get to enjoy.

Anyway, this is his public, stated opinion on this: https://github.com/dotnet/runtime/pull/115762#issuecomment-2...

replies(2): >>svaha1+tl1 >>n144q+gK2
◧◩◪◨
25. xeonmc+LP[view] [source] [discussion] 2025-05-21 17:14:06
>>mock-p+MF
I don’t imagine getting laid with computers are particularly enjoyable for humans.
replies(1): >>rasz+pR1
26. LunaSe+U21[view] [source] 2025-05-21 18:20:38
>>bramha+(OP)
Microsoft closed their recently acquired advertisement buy-side platform Xander Invest because they are replacing it with an AI-only platform.

They only gave their customers 9 months to migrate away.

I'm expecting that Microsoft did this to artificially pump up their AI usage numbers for next year by forcibly removing non-AI alternatives.

This only one example in AdTech but I expect other industries to be hit as well.

◧◩◪◨⬒
27. svaha1+tl1[view] [source] [discussion] 2025-05-21 19:57:54
>>square+1L
If he reiterates that comment to me after two beers in a relaxing bar I might believe him.
replies(1): >>davegu+ls1
◧◩◪
28. raydev+6m1[view] [source] [discussion] 2025-05-21 20:01:31
>>stickf+5f
> It's pretty obviously a failed experiment

For it to be "failed" it would have to also be finished/completed. They are likely continuously making tweaks, this thing was just released.

replies(1): >>vright+qE2
29. Allege+Mo1[view] [source] 2025-05-21 20:16:49
>>bramha+(OP)
Given that Microsoft always decided to Will Not Fix issues because they went "oh this thing is throwing errors? Just ignore them". THey're numbskulls that are high on their own farts just as much as their managers. They deserve everything that's happening to them.
◧◩◪◨⬒⬓
30. davegu+ls1[view] [source] [discussion] 2025-05-21 20:37:01
>>svaha1+tl1
Hahaha. 1000% this. Also, first example from the linked video: a "not vibe coded, promise" example of an ascii space invaders clone... Of all the examples of "has a bunch of training code data since the 80s", this is the best representation of exactly what LLM coding is capable of "in 8 minutes".
31. pier25+cy1[view] [source] 2025-05-21 21:22:47
>>bramha+(OP)
Yeah it's quite disheartening.

I recently spent a couple of months studying C# and .NET and working on my first project with it.

.NET, Blazor, etc are not known for a fast release schedule... but if things are going to become even slower with this AI crap I wonder if I made the right call.

I'm quite happy how things are today for making web APIs but I wish Blazor and other frameworks were in a much better shape.

replies(1): >>Kwpols+pC1
◧◩
32. snacke+jA1[view] [source] [discussion] 2025-05-21 21:40:45
>>nashas+F
I think we all know they won’t.

I am genuinely curious though to see the strategies they employ to absolve themselves of guilt and foolishness.

Is there precedent for the entire exec and management class embracing a new trend to this kind of extent, then it blowing up in their faces?

◧◩
33. Kwpols+pC1[view] [source] [discussion] 2025-05-21 21:56:32
>>pier25+cy1
.NET has major releases every year. How is that slow for a programming platform/framework?
replies(2): >>crater+hJ1 >>pier25+OJ1
◧◩◪
34. crater+hJ1[view] [source] [discussion] 2025-05-21 23:01:12
>>Kwpols+pC1
Go has a six month release cycle. Rust releases a new stable every six weeks.
replies(1): >>Kwpols+yu9
◧◩◪
35. pier25+OJ1[view] [source] [discussion] 2025-05-21 23:06:36
>>Kwpols+pC1
Yes but the improvements are very gradual. It takes years for something to reach maturity. At least for the web stuff which is what I know of.

Eg:

Minimal APIs were released in 2021 but it won't be until .NET 10 that they will have validation. Amazing that validation was not a day one priority for an API. I'm not certain if even in .NET 10 Minimal APIs will have full parity of features with MVC.

Minification of static assets didn't come until .NET 9 released in 2024. This was already commonplace in the JS world a decade earlier. It could have been a quick win so long ago for .NET web apps.

Blazor was released in 2018. 7 years later they still haven't fixed plenty of circuit reconnection issues. They are working on it but progress is also quite slow. Supposedly with .NET 10 session state will be able to be persist etc but it remains to be seen.

OpenAPI is also hit and miss. Spec v3.1 released in 2021 is still not supported. Supposedly it will come with .NET 10.

Not from .NET but they have a project called Kiota for generating clients from OpenAPI specs. It's unusable because of this huge issue that makes all properties in a type nullable. It's been open since 2023. [1]

Etc.

[1] https://github.com/microsoft/kiota/issues/3911

◧◩◪◨
36. disqar+mK1[view] [source] [discussion] 2025-05-21 23:11:54
>>some_r+4w
...to turn us into soylent-flavored mojitos?
◧◩◪◨⬒
37. rasz+pR1[view] [source] [discussion] 2025-05-22 00:33:36
>>xeonmc+LP
You havent met Gwendolyn bot.
◧◩
38. saati+GY1[view] [source] [discussion] 2025-05-22 02:04:20
>>empath+6o
They can do that in private repos just as easily, this a pr stunt that backfired very badly.
◧◩◪◨
39. vright+qE2[view] [source] [discussion] 2025-05-22 10:35:51
>>raydev+6m1
"This thing has just released"

"It would have to be finished/completed"

Do you honestly not see a problem with those two statements in such close proximity? Is it finished or is it released? The former is supposed to be a prerequisite for the latter.

replies(2): >>tremon+CZ2 >>raydev+8i3
◧◩◪◨⬒
40. n144q+gK2[view] [source] [discussion] 2025-05-22 11:38:11
>>square+1L
Of course that is what he says publicly. Can you imagine him saying anything different on this already very heated PR comment section? Those would be quoted in a headline in a news article the next second.
◧◩◪◨⬒
41. tremon+CZ2[view] [source] [discussion] 2025-05-22 13:42:22
>>vright+qE2
What bliss it must be, to never have encountered Microsoft software before.
replies(1): >>vright+u13
◧◩◪◨⬒⬓
42. vright+u13[view] [source] [discussion] 2025-05-22 13:56:08
>>tremon+CZ2
oh, how I wish you were right... I had to look deep inside some microsoft software, and I think it actually shortened my lifespan
◧◩◪◨⬒
43. raydev+8i3[view] [source] [discussion] 2025-05-22 15:38:28
>>vright+qE2
It's unfinished and it's in the public's hands. I don't see these as opposing ideas.

We can debate whether they should have called this an experiment or an alpha or beta or whatever, but that's a different discussion.

The fact that people are using it currently does not make it a failure. When MS shuts it down, or Copilot is wildly unprofitable for multiple quarters, team behind it quits, etc, etc, then we can determine whether it has failed or not.

But if they continue to have paying customers and users are finding some benefits over not having Copilot, and MS continues to improve it (doesn't let it rot), then you'd have to provide some evidence of its failure that isn't "look at Copilot being stupid sometimes". Especially when stupidity is expected of it.

◧◩◪◨
44. Kwpols+yu9[view] [source] [discussion] 2025-05-25 11:00:01
>>crater+hJ1
Six weeks works for a young language like Rust. .NET releases also include application development frameworks in addition to the language itself.
[go to top]