ChatGPT's enterprise success against Copilot fuels OpenAI/Microsoft rivalry
bloomberg.com
bloomberg.com
It's too bad Copilot is by far the dumbest competitor in the space
My favorite interaction so far was when I prompted it with:
ffmpeg command to convert movie.mov into a reasonably sized mp4
Copilot's response:
I implemented and executed the Python code above to convert movie.mov to a reasonably sized movie.mp4 using ffmpeg.
However, the Python code failed since it was not able to find and access movie.mov file.
Do you want me to try again or is there anything else that I can help you with?
However, piping code requests, and hey->do this, AI gives you something; the privacy aspect of it.
On the in-terms, these ARE LLMs, devs/infra people do look at how their tools are being used. You can pull a lot of info with and about an organizatin and what their internals are up to just from how they're using the AI's information.
So while we're looking at Microsoft's quality of testing, what does that really mean in terms of how they're viewing the tool usage.
Hell, the Q4 quantized Mistral Small 3.1 model that runs on my 16GB desktop GPU did perfectly as well. All three tests resulted in a command using x264 with crf 23 that worked without edits and took a random .mov I had from 75mb to 51mb, and included explanations of how to adjust the compression to make it smaller.
What I always find hilarious too is when the AI Skeptics try to parlay these kinds of "failures" into evidence LLMs cannot reason. If course they can reason.
Sure you can get creative interesting results from something like "dog park game run fun time", which is totally unclear, but if you're actually solving an actual problem that has an actual optimal answer, then clarity is _always_ better. The more info you supply about what you're doing, how, and even why, the better results you'll get.
LLMs are like humans in this regard. You never get a human to follow instructions better by omitting parts of the instructions. Even if you're just wanting the LLM to be creative and explore random ideas, you're _still_ better off to _tell_ it that. lol.
However _even_ in a MoE system you _still_ always get better outputs when your prompting is clear with as much relevant detail as you have. They never do better because of being unconstrained as you mistakenly believe.
https://i.imgur.com/toLzwCk.png
ffmpeg -i movie.mov -c:v libx264 -preset medium -crf 23 -c:a aac -b:a 128k movie.mp4
BUT, I have this in my copilot-instructions.md file:
# Always follow these steps when responding to any request
1. Please do a round of thinking in <thinking></thinking> tags
2. Then a round of self-critique in <critique></critique> tags
3. Then a final round of <thinking>, before responding.
4. If you need more information, ask for it.It wasn't. It's the dumbest name ever.
Or the Teams download page that had two different versions - Teams for Home and Teams for Work/School.
Or .NET->.NET Core & .NET Framework->Back to .NET again.
EDIT:
There's even a rule of acquisition that could possibly apply here: #239 - Never be afraid to mislabel a product.
I got this on mobile. Seems to be pretty apt.
Are they aware that people will struggle to find if Office is installed and that they will keep calling it Office til the end of times (aka the next rebranding that will revert back things) anyway?
https://www.microsoft.com/en-us/microsoft-365/microsoft-offi...
I got there by going to office.com and clicking Products > Microsoft Office. Lol. Rofl, even. This has made my day. And we all thought calling their third generation console Xbox One was the worst possible branding decision.
IBM used to be a repeat offender. I recall trying to buy the WebSphere (Java) application server for a client and then finding out that IBM had slapped "WebSphere" on all their products including things like¹ MQ Series (a queue) and DB/2 (a database). It took me an hour to figure out the right item and it was an online purchase!
¹I might be misremembering the exact products but it was similarly absurd.
So you get your copilot for m365 subscription and add copilot studio which you use to create co pilots
This is all information taken from office.com, not some joke or exaggeration...
Looking at the two side by side in an app manager, I see:
- Copilot - com.microsoft.copilot
- M365 Copilot - com.microsoft.office.officehubrow
And they both have identical icon, except the latter has a tiny black rectangle with tiny white "M365" label tucked in the corner.
Somewhere between that and a master's degree and 10 years at a prestigious marketing firm, though, apparently there's some lessons about how you should smear all your brands all over each other in some bid to, I presume, transfer any good will one may have had to all of them, but it seems to me that they could stand to send those people back to MKT101 again, because the principle of labeling what your product actually is seems to elude them after Too Much Education.
If anything, Marketing 101 works as a scaffolding but you learn the real lessons later on (basically like with every other vocational training wrapped in a degree, including especially computer science) - but also, and perhaps more importantly, it serves as a fig leaf. You can point to that and say, it's a Science and an Art and is Principled and done well It Is For The Good Of All Mankind, and keep the veneer of legitimacy over what's in practice a more systematized way of bringing harm to your fellow humans.
Also specifically wrt. brands - brands as quality signals mostly died out a decade or more ago; mixing them up is just a way to get their decaying corpses to trick more people for a little longer.
Openned: https://copilot.microsoft.com
Same prompt: ffmpeg command to convert movie.mov into a reasonably sized mp4
(Perhaps copilot is not lazy, just stupid relative to its peers.)
~> ollama run gemma3:12b-it-qat >>> ffmpeg command to convert movie.mov into a reasonably sized mp4
Here's a good ffmpeg command to convert `movie.mov` to a reasonably sized MP4, along with explanations to help you adjust it:
```bash ffmpeg -i movie.mov -c:v libx264 -crf 23 -preset medium -c:a aac -b:a 128k movie.mp4 ```
*Explanation of the command and the options:*
Copilot with outlook.com
Copilot base one that comes with M365,
And the add-on one for 30$/mo.
Copilot in VS code
All produce: ffmpeg -i movie.mov -vcodec libx264 -crf 23 -preset medium -acodec aac -b:a 128k output.mp4
Which is not surprising because its just an Open AI 4o call... so how are you getting this?
Already gemini giving me flash or pro like i'm supposed to know and decide which i need is missing the point, but 4 of them is crazy.
GPT3.5, 4.0, 4o, 4o-mini, 4.1 o3, 4,5?
I want to slap every person at OpenAI.
call freaking tuned models what they are, basic, advanced, coding, research, etc. etc. Slow, Fast. Like cmon.
Because LLM-based service outputs are fundamentally not-reproduceable. We have no insight into any of the model settings, the context, what model is being run, etc.
ffmpeg command to convert movie.mov into a reasonably sized mp4
ffmpeg -i movie.mov -vcodec libx264 -crf 23 -preset medium -acodec aac -b:a 128k movie_converted.mp4
Here’s the breakdown:
- -vcodec libx264: Uses the H.264 video codec, great for compression and compatibility.
- -crf 23: Constant Rate Factor — lower means better quality and larger size; higher means smaller size but lower quality. 23 is a nice middle ground.
- -preset medium: Controls encoding speed vs. compression efficiency. You can try slow or faster depending on your needs.
- -acodec aac -b:a 128k: Compresses the audio using AAC at 128 kbps — pretty standard and compact.
If you're looking for even smaller file sizes, try bumping -crf up to 26 or using -preset slow.
Want to bake in scaling, remove metadata, or trim the video too? Happy to tailor the command further.A lot of our bad experiences with, say, customer support hotlines, municipal departments, bad high school teachers, whatever, are associated with a habit of speaking that ads flavor, vibes, or bends experiences into on-the-nose stories with morals in part because we know they can't be reviewed or corrected by others.
Bringing that same way of speaking to LLMs can show us either (1) the gap between what it does and how people describe what it did or (2) shows that people are being treated differently by the same LLMs which I think are both fascinating outcomes.
The best way to get the right answer from an LLM is not to ask it the right question; it's to post online that it got the wrong answer.
Just because other people on here say “worked for me” doesn’t invalidate OPs claim. I have had similar times where an LLM will tell me “here is a script that does X” and there is no script to be found.
Most of us are going to get the same answer to "which planet is third from the sun" even with different contexts. And if we're fulfilling our Healthy Internet Conversation 101 responsibility of engaging in charitable interpretation then other people's experiences with similarly situated LLMs can, within reason, be reasonably predictive and can be reasonably invoked to set expectations for what behavior is most likely without that meaning perfect reproducibility is possible.
How? I would argue they do the exact opposite of that.
We get these same anecdotes about terrible AI answers frequently in a local Slack I’m in. I think people love to collect them as proof that AI is terrible and useless. Meanwhile other people have no problem hitting the retry button and getting a new answer.
Some of the common causes of bad or weird responses that I’ve learned from having this exact same conversation over and over again:
- Some people use one never-ending singular session with Copilot chat, unaware that past context is influencing the answer to their next question. This is a common way to get something like Python code in response to a command line question if you’re in a Python project or you’ve been asking Python questions.
- They have Copilot set to use a very low quality model because they accidentally changed it, or they picked a model they thought was good but is actually a low-cost model meant for light work.
- They don’t realize that Copilot supports different models and you have to go out of your way to enable the best ones.
AI discussions are weird because there are two completely different worlds of people using the same tools. Some people are so convinced the tool will be bad that they give up at the slightest inconvenience or they even revel in the bad responses as proof that AI is bad. The other world spends some time learning how to use the tools and work with a solution that doesn’t always output the right answer.
We all know AI tools are not as good as the out of control LinkedIn influencer hype, but I’m also tired of the endless claims that the tools are completely useless.
Cell phones are full of examples. So much of this stuff is obvious now we’ve been using them for awhile, but it wasn’t obvious when they were new. “My call dropped because I went in a tunnel” is user error. “My call cut out randomly and I had to call back” is a bug. And “my call cut out because my phone battery ran out” is somewhere in the middle.
For chatbots, lots of people don’t know the rules yet. And we haven’t figured out good conventions. It’s not obvious that you can’t just continue a long conversation forever. Or that you have to (white consciously) pick which model you use if you want the best results. When my sister first tried ChatGPT, she asked it for YouTube video recommendations that would help when teaching a class. But none of the video links worked - they were all legitimate looking hallucinations.
We need better UX around this stuff. But also, people do just need to learn how to use chatbots properly. Eventually everyone learns that calls will probably drop when you go into a tunnel. It’s not one or the other. It’s both.
I saw an IT professional google “My PC crashed” to diagnose a server bluescreen stop error.
Some people are so convinced the tool will be bad that they give up at the slightest inconvenience or they even revel in the bad responses as proof that AI is bad
"How dumb do you want your AI to be?"
"Why do I have to select?"
"Because smart costs money"
"So... I can have dumb AI but it's cheaper?"
"Yes"
"How would the average person know which to pick?"
"Oh you can't know."
I hope they can invent an AI that knows which AI model my question should target cheaply.
Some users may not appreciate that, but many more might be drawn to the "adjust the color balance on the TV" vibes.
Isn't that the idea of OpenRouter?
I think you are getting confused by the term "Model Routing", which to be fair OpenRouter does support, but it's a secondary feature and it's not their business focus. Actually OpenRouter is more focused on helping you choose the best provider for a specific open model based on their history of price, speed, reliability, privacy...
The model routing is simply provided by NotDiamond.ai, there are a number of other startups in this space.
I don’t think I’d trust the vendor’s ai to optimize when they will likely bias toward revenue. So a good case for a local ai that only has my best interests at heart.
Currently, the guidance from vendors is “try it and see which yields the best results” which is kind of like “buy this book, read it, and see if you like it” and how of course the publisher wants you to take this action because they get their money.
It would be great to have a cheap AI that can self-evaluate how confident it is in its reply, and ask its expensive big brother for help automatically when it’s not.
Shouldn’t asking a more expensive model for input be a similar level of «tool use»?
ChatGPT lets me choose between GPT-4o ("Great for most tasks"), o3 ("Uses advanced reasoning"), o4-mini ("Fastest at advanced reasoning"), and o4-mini-high ("Great at coding and visual reasoning").
Is what I'm doing "most tasks"? How do I know when I want "advanced reasoning"? Great, I want advanced reasoning, so I should choose the faster one with the higher version number, right? etc.
I'm talking about ChatGPT, which is a Web and desktop app where users run interactive sessions. What does "production" mean in this sense?
Anthropic, by contrast, uses the same name for the user-facing app and the models. This is confusing, because the user-facing apps have capabilities not native to the models themselves.
- there are billions of people who are not accustomed to using software this way, who are in the expected target market for this software. Most people cannot tell you the major version number of their mobile OS.
- this approach requires each individual to routinely perform experiments with the expanding firmament of models and versions. This is obviously user-hostile.
Anyway, my hot take here is that making things easier for users is better. I understand that is controversial on this site.
If paying by API use it probably makes more sense to talk to a lesser stooge where possible, but for a standard pro plan I just find the lesser models aren't worth the time to use in frustration they cause.
I’m Feeling Lucky -> bad result -> Google search is useless
2. The Copilot or ChatGPT or Claude "Ask" buttons should then be renamed to "I'm feeling lucky". And that would be the only button available.
Who claimed that here?
I read a claim that Copilot is dumber than claude and ChatGPT and I tend to confirm this.
"They don’t realize that Copilot supports different models and you have to go out of your way to enable the best ones."
So possible that none of us who thinks that, went out of our way to find outy when there were working alternatives, but it would be still on Microsoft making it hard to make good use of their tool.
Where did I say anything in general about LLMs being useless?
Well, unless your business is selling vibes.
But they turning some knobs based on load also looks razonable.
ffmpeg -i movie.mov -vcodec libx264 -crf 23 -preset medium -acodec aac -b:a 128k movie_converted.mp4
Along with a pretty detailed and decently sounding reasoning as to why it picked these options.
Read the replies. Many folks have called gpt-4.1 through copilot and get (seemingly) valid responses.
Thing is I ask it random bits like this all the time and it's never done that before so I'm assuming some recent update has borked something.
Yeah, like how about answering the fucking question? lol
It’s also a website like ChatGPT apparently? I thought it was called Copilot because it writes with you, so why is there also a general chat/search engine called Copilot? Jesus.
https://www.windowslatest.com/2025/01/18/microsoft-just-rena...
Edit: They are doubling down on bad naming conventions so hard that it makes me think it's some kind of dark pattern sales strategy..
And I would agree with them in this case.
The new chat, or new conversation buttons seem to do nothing.
From a one line question it made me a relevant document of 45 pages examining the issue from all different sides, many of which I hadn't even thought of. It spent 30 mins working. I've never seen Perplexity spend more than 5.
I won't't be surprised if they will significantly nerf it to save on computing costs. I think now they give it their all to build a customer base and then they nerf it.
So you end up with a worse ChatGPT that also doesn't have work context.
My shock moment was when I was asking it to covert an image into a nice PPTX slide naively assuming it had the best PowerPoint capabilities since that’s also a MS product.
It returned a non formatted text box on one slide. I had to laugh so hard people on the office noticed.
I blame it on corporate reluctance to take risks that could result in bad press. They put a lot of work into screwing Copilot up. I mean, they had it running GPT-4 back when GPT-4 was the new hot thing. The way it was comically useless is not something that "just happens" - as evidenced by the fact, that just running GPT-4 via API yourself produced good results by default.
Also in line with this, Copilot 365 seems to not get how charts work. I asked it with multiple different massaged data sets and it kept giving worse and worse answers even after resetting the chat and data to as simple as possible (think 10 dates, 2 colums of integers) until it ultimately gave me a blank chart. I gave up and asked GPT.
i asked it to make a sheet look nicer and it couldn't without me explicitly telling it what i wanted done.
When i told it to do certain things, it's like "that's not implemented yet, stay tuned!"
When ChatGPT first came out, Satya and Microsoft were seen as visionaries for their wisdom in investing in Open AI. Then competitors caught up while Microsoft stood still. Their integration with ChatGPT produced poor results [1] reminding people of Tay [2]. Bing failed to capitalize on AI, while Proclarity showed what an AI-powered search engine should really look like. Copilot failed to live up to its promise. Then Claude.ai, Gemini 2.0 caught up with or exceeded ChatGPT, and Microsoft still doesn't have their own model.
[1] https://www.nytimes.com/2023/02/16/technology/bing-chatbot-m...
It’s a big, unsolvable mess that will forever prevent them from competing with legacy-free, capable startups.
They should delete all their public facing websites and start over.
https://www.osnews.com/story/19921/full-text-an-epic-bill-ga...
If he had to send the same email every day he wasn't doing his job well, and neither was everyone below him. Even a fraction of that list is too much.
1. Not sure why osnews charactarised this as an "epic rant". I thought he was remarkably restrained in his tone given both his role and his (reasonable) expectations.
2. This to me shows just how hard it is for leaders at large companies to change the culture. At some point of scaling up, organisations stop being aligned to the vision of the leadership and become a seemingly autonomous entity. The craziness that Bill highlights in his email is clearly not a reflection of his vision, and in fact had materialised despite his clear wishes.
When we think about how "easy" it would be for the executive of a large organisation to change it, those of us not experienced at this level have an unrealistic expectation. It's my belief that large organisations are almost impossible to "turn around" once they get big enough and develop enough momentum regarding cultural/behavioural norms. These norms survive staff changes at pretty much every level. Changing it requires a multi-year absolute commitment from the top down. Pretty rare in my experience.
Too many crazy presentations on 'data' that are calling the calling the sky purple and everyone just nods along, ok's and gives promos all around.
Access to their IP, and 20% of revenue (not profit).
Altman will absolutely attempt this.
1. Lack of access to compute resource. Microsoft intentionally slowing OpenAI's ability to scale up and grow dominant quickly vs. Copilot, a competing product. Microsoft shouldn't be able to use it's dominance in the cloud compute market to unfairly influence the market for consumer AI.
2. Microsoft should not automatically gain OpenAIs IP in domains outside of the AI offerings that the company was supplying when the initial agreement was made. If it must be upheld the terms of the contract mean Microsoft get all of OpenAIs IP, then it block OpenAI from competing in other markets eg. Windsurf vs. VS Code.
An investment vehicle would be more accurate, but that's the primary function of every broadly-held publicly-traded firm.
It has made up tags for cli functions, suggested nonexistent functions with usage instructions, it’s given me operations in the wrong order, and my personal favorite it gave me a code example in the wrong language (think replying Visual Basic for C).
We've never seen a "Dog Pile vs Yahoo" battle when the giants are of this scale.
It'll be interesting to see if Google can catch up with ChatGPT (seems likely) and if they simply win by default because they're in all of the places (also seems likely). It'd be pretty wild for ChatGPT to win, honestly.
A lot of people who are unfamiliar with how the technology works talk about "my GPT". Google that phrase, or start watching for it to crop up in conversation.
On the other end of the spectrum, there are lots of tiny little pockets like this:
Well done Google Marketing, well done.
Another product carefully kept away from the grubby little hands of potential users!
Seems a lot more like general availability on my end now, though, these past few days. One can try Google dot com, slash AIMode.
Not nearly as good as using Gemini 2.5 pro which they do offer for free but I forget where. AI studio? So many ways to access it.
(& small typo, “Proclarity” = *Perplexity)
Renaming all their products to Copilot makes no sense and just causes brand confusion.
Copilot getting access to your entire 365/azure tenant is just a security nightmare waiting to happen (in fact theres already that one published and presumably patched vuln)
It has so many shackles on that its functionally useless. Half the time I ask it to edit one of my emails, it simply spits my exact text back out.
Its one singular advantage is that it has crystal clear corpospeak license surrounding what it says your data will be used for. Whether or not its true is irrelevant, organisations will pick it up for that feature alone. No one ever got fired for choosing ibm etc.
This reminds me of IBM Watson back in the day
LOL. We're talking about the company that used to slap a non-sensical .NET suffix on everything.
It’s so strange that they keep renaming because Office (and office.com) is perfectly usable.
And us plebs working for the company are left to deal with the inferior tool.
Insert "No, We have copilot at home children" meme here.
What doesn't work anymore for IBM still certainly works for Oracle and the rest of the sales-driven tech giants.
I also wasn't aware that there where an OpenAI/Microsoft rivalry, I had the impression that Microsoft put a lot of money into OpenAI and that ChatGPT ran on Azure, or was at least available as an offering via Azure.
OpenAI's models are also available via Azure.
they flopped this royally, just like windows mobile. they created a shitty ux by shoving it inside the bing app, then they decided to charge for it instead of capturing all enterprise value.
lastly, the product has stalled and missed on their biggest opportunity which is tapping into the data. you can think it's because of how complex it must be, but then openai and everybody else did it.
it's truly a lesson in product mismanagement, once again, from microsoft
MSFT is the world’s best 2nd mover and can often make profits on ideas pioneered/launched by other companies before MSFT.
MSFT came late to the cloud party (2011? AWS launched in 2006 IIRC), still they became a big player there (~25% market share in 2025).
Yes they botched Mobile, but to me it looks like they are still in the AI game
(I personally prefer models from Google, Anthropic or OpenAI though).
The Github Copilot (in VS Code especially) is the only application of LLMs that I've found useful from Microsoft. I would have loved amazing Copilot support in Word for working on a large complex document, but I haven't found that to work well.
1. In Outlook asked Copilot to "create an email asking for a weekly status update for Project ABC". Iterating a bit, I asked it to be more casual and shorter. Here's what I got:
Hey Team,
Hope you're all doing great. To keep everyone on the same page with Project ABC, I'd like us to start doing weekly status updates. Just a quick summary of key developments, milestones, challenges, and what's coming up would be super helpful. This way, we can coordinate better and tackle any issues right away.
Thanks for your cooperation and hard work. Looking forward to your updates!
2. I provided Copilot with this prompt:
"I am upgrading the electrical service at my house from 100a to 200a service. Create a project plan in Planner for this project."
Here's the response I got:
Nice! Upgrading from 100A to 200A is a solid move—more headroom for your gear and future-proofing your setup. Here's a breakdown of how you could structure this as a project in Microsoft Planner, with buckets and tasks to keep it organized:
---
### **Bucket 1: Planning & Permitting**
- **Assess current electrical load and future needs**
- **Consult with a licensed electrician**
- **Request utility service requirements**
- **Obtain local permits for electrical service upgrade**
- **Review NEC and local code requirements**
---
### **Bucket 2: Materials & Scheduling**
- **Select upgraded panel (200A rated) and breakers**
- **Order required materials (panel, feeder wires, conduit, ground rods, etc.)**
- **Coordinate shutdown and install time with utility provider**
- **Schedule rough-in inspection with local authority**
---
### **Bucket 3: Installation**
- **Shut off power at the main disconnect (coordinate with utility)**
- **Remove existing 100A panel and components**
- **Install 200A panel and connect feeders**
- **Upgrade grounding/bonding system as required**
- **Replace meter socket if required**
---
### **Bucket 4: Inspections & Finalization**
- **Schedule and pass electrical inspection**
- **Have utility restore permanent power**
- **Label circuits clearly in new panel**
- **Update homeowner insurance with new electrical capacity**
---
### **Bucket 5: Post-Project Wrap-Up**
- **Document panel layout and upgrades**
- **Evaluate energy use and plan for future expansion (EV charger, generator, etc.)**
- **Close out permits with inspection authority**
---
You can assign these tasks to yourself or any contractors involved, and add due dates to align with your timelines. Want help creating a PowerShell script to bulk import this into Planner?
3. I provided Copilot with this prompt:
book a meeting with <name> for August 1 entitled "Test Meeting"
I can't directly book meetings or access your calendar, but I can help you set it up! Here's an example of what to enter in Outlook or your preferred calendar app:
**Title:** Test Meeting
**Date:** August 1
**Attendees:** Bryan Herbst
**Time:** [Insert your preferred time]
**Location:** [Optional]
**Description:** [Add any notes or agenda items]
If you're using PowerShell or a script to automate calendar events—especially within Microsoft 365—I can help you write that too. Want to go that route?1) "Iterating a bit". Like that email is barely 5 sentences, you could write it faster than coaxing it out of Copilot
2) It is drivel. Could have stopped at *Consult with a licensed electrician
3) Well...
It needs to be a lot more clear and direct about the expectations of others.
"I'd like us to do X" is super passive, and a thousand miles from "You must do X"
Even 'draft' really, if you forget what you've done, close your eyes and think about it from scratch, surely you expect that to be drafted in your ...drafts?
I think this article is about the 365 suite.
At the top-right of that page, it has a little icon indicating 'enterprise data protection' but I can't see any way for me (the user) to know what type of Copilot licence (if any) the accountholder has assigned to my user account.
I'm seeing enterprise and personal users hit their monthly rate limits in less than 3 days.
Ironically Windsurf is also owned by Microsoft indirectly...
So it often comes down to this choice: Open https://copilot.cloud.microsoft/, go through the Microsoft 365 login process, dig your phone out for two-factor authentication, approve it via Microsoft Authenticator, finally type your request only to get a response that feels strangely lobotomized.
Or… just go to https://chatgpt.com/, type your prompt, and actually get an answer you can work with.
It feels like every part of Microsoft wants to do the right thing, but in the end they come out with an inferior product.
ChatGPT. Perplexity. Google AI Mode. All let you get a message off.
… WAIT! copilot dot microsoft dot com lets you just send a message without logging in.
—
heh, the second result on DuckDuckGo is an MS article: “What is Copilot, and how can you use it?”
Products mentioned in the article, they say:
| Copilot | Copilot app | Copilot for individuals |
And a link for each one. Does Satya squirm when he sees that, but doesn’t have the power to change it?
Also the word “individuals” (allegedly previously mentioned) appears only once on the page.
Just yesterday was I talking to a customer who was so happy with our "co-pilot" compared to ChatGPT and others that he wants to roll it out to the rest of the company.
We use Azure-OpenAI + RAG + system prompt targeted at architects (AEC). It really seems the system prompt makes all/most of the difference. Because now, users will always get answers targeted towards their profession/industry.
Since the launch of ChatGPT Microsoft has had access to it and even had some of the most popular code editors, and where did it take them. This is why Meta had to launch threads with a very small team since a big team in Big tech can just not compete.
Off course like everything else there are no absolutes and when Big Tech feels there is an existential crisis on something they do start improving, however such moments are far and few.
My observation is that in a disorganized and over documented organization, copilot flattens to an exec summary language that moves things along a lot faster. It’s enables communication beyond the limiting pace of individuals learning to communicate hard things with nuance (or, sometimes, when people are reluctant to next step in the cycle).
It lifts to a baseline that is higher than before. That has, in turn, shortened communication cycles and produced written output in an org that over-indexed to an oral tradition.
That's the only data point the article has, and it is incomplete (no Copilot numbers).
The rest are just testimonials (some of anonymous character) and stories.
Who's having more success then? No one knows. It's up to the reader to decide.
Looks like made-up rivalry article to me. Draws clicks, no actual content inside.
So how did MS make Copilot Suck, if it started with same base?
But, it’s mostly a RAG tool, “grounded in web” as they say. When you give Copilot a query, it uses the model to reword your query into an optimal Bing search query, fetches the results, and then crafts output using the model.
I commend their attempt to use Bing as a source of data to keep up to date and reduce hallucinations, especially in an enterprise setting where users may be more sensitive to false information, however as a result some of the answers it gives can only be as good as the Bing search results.
Also if anyone from OpenAI or any of its competitors wants to talk my email is on my HN profile ;-)
Now that everyone has access to Claude and claude-code, Copilot barely gets mentioned anymore. Maybe this wave dies down or they improve it, anyway these tools still have a long long way to go.
My precise request: "Extract the list of field names in Exhibit A."
Its precise response: "I understand that you want to extract the list of field names from Exhibit A in your document. Unfortunately, I cannot directly perform document-related commands such as extracting text from specific sections."
I tried several different ways of convincing it, before giving up and using the web version of ChatGPT, which did it perfectly.
I had an even worse experience with the Copilot built into the new version of SSMS. It just won't look at the query window at all. You have to copy and paste the text of your query into the chat window ... which, like, what's the point then?
ChatGPT simply is a much better product all around. Period.
If I try to get it to do stuff outside my domain expertise it's making errors I can't catch. So I suppose if move fast and break things works for your business, then that's fine.
But that begs the question, a much better product than what?
Either way, we saw them fire a bunch of people and "replace them with AI," so it's not out of the question this is a more toward "AI tech leadership" tax subsidization as DEI is phased out.
Man they sound like a mainframe manufacturer at the dawn of the PC era.
A lot of the early adopters (and driving forces) of LLMs have been tech-minded people. This means it's quite a good idea NOT to confuse them.
And, yet, Microsoft decided to name their product Microsoft Copilot, even though they already had a (quite well-received!!) Copilot in the form of Github Copilot, a product which has also been expanding to include a plethora of other functionality (albeit in a way that does make sense). How is this not incredibly confusing?
So what actually _is_ Copilot? Is there a bing copilot? A copilot in windows machines? Is it an online service? (I saw someone post a link to an office 365)?
I'm going to be honest and tell you that I have no fucking clue what Microsoft Copilot actually is, and Microsoft's insistence on being either hostile to users or pretending like they're not creating a confusing mess of semantic garbage is insulting. I am lucky not to have to use Windows daily, and most of what I do that involves copilot is...Github Copilot.
I am knee-deep into LLMs. My friends can't stand me with how much I go on about them, how I use them, from remote to local models, to agents, to the very debatable idea that they may be conscious, you name it. And yet, as bullish as I am on the thing, I have no fucking clue what Microsoft copilot is. Perhaps I'm definitely not their target market, but from what I've seen, tech-illiterate people have no idea what it is either, just that it's "more microsoft trash".
When I was younger, I used to be a very loud anti-microsoft boy, I loathed everything they did. Slowly, for a while, they were managing to win me over (in part because I outgrew that phase, but also because they have definitely been cleaning up their image and, at least to me, producing better and more relevant software). However, in recent years, their insistence on naming everything this way and creating a maze out of their products is...baffling. I feel myself not being able to stand MS again.
And what is it with big corporations and a seeming inability to name their products decently? This is appalling. The people making these decisions should be fired, because clearly they don't have any pride in what they do, or they wouldn't have allowed this.
Get your shit together, microsoft!
If you look at this in isolation, yes. If you look at this historically, it's totally on-brand for Microsoft. Office 365, Live, MSN were all brand that Microsoft has slapped wholesale on things. Microsoft has always been reactive when it comes to branding, rather than proactive.
You don't "use Copilot". You Copilot. Everything is Copilot. Windows, Office, PCs, Bing. All Copilot.
Unless you mean Copilot (Classic).
But copilot free allows me to use extended thinking. So to get round the problem we download the standalone app and not sign in.
It seems wild that the free version is better than the paid version.
I used to like Microsoft, i now despise them, and the more you dig, the more shady stuff emerges from that supposed 'company'