I think the biggest issue is M365 Copilot was sold as something that would integrate with business data (teams, files, mail, etc.) and that never worked out quite well.

So you end up with a worse ChatGPT that also doesn't have work context.

When you do have that work context MS copilot performs quite well. But outside of that usecase it's easy to see their model is pretty bad.

It absolutely does not perform well with work context.

> and people were buying it! real corporations and governments were buying this crap - it's insane

Anedote: in Wall Street, Global Relay and TeleMessage are the major players when it comes to achieving communication for compliance.

before that wallstreet ran on yahoo messenger! they only stopped because new yahoo brand owners didn't understood the value of this and shut it down because there weren't enough teens signing up.

These days you can 'just' create an S3 tables bucket. https://docs.aws.amazon.com/AmazonS3/latest/userguide/s3-tab...

Athena is really expensive though and you will often run into a hard limit on the size of your query.

Like most things serverless Athena is cheap as long as you don't use it.

My company has 100s of data pipelines that are executed infrequently.

For this use case Athena is ridiculously cheap and easy to use vs most other solutions.

I never found Athena expensive. Compared to employment cost it will be miniscule.

And some times, if your query is CPU extensive but the queried data size is not huge you can get a ridiculous value for money, like many CPU-days in 10 minutes for just $5 if your query covers 1TB after partitioning.

Query size limits are also configurable.

Obviously it depends on what data you are working on, but not having to set up and pay for a computational cluster is a huge cost saving.

Agreed.

A lot of people worry would worry about "vendor lock-in" here, but it's certainly convenient.

Interesting, I had a similar experience as an exchange student in Britain.

I've brought MS Copilot licenses for my company in ~ February 2023. They were sold in a yearly commitment and offered no trials. A bad deal, but I was afraid of missing out AI productivity gains.

I'm definitely not renewing those. Times are hard and the value provided does not justify the cost.

I wonder how many companies will do the same.

I recently personally resubscribed to copilot after cancelling my subscription a couple months ago since it was not providing value. But now with the new/beta “Edit” mode and being able to specify to use o1, o1 mini, and sonnet 3.5, the $10 a month feels a lot more worth it. The edit mode has outperformed aider for me.

Sorry, I think we are talking about different stuff.

I was referring to Copilot for M365, which costs 30 USD / month and targets biz users.

Github Copilot is definitely worth it.

Oh wow, I definitely misunderstood, all the “Copilot” usage by MS/Github got me mixed up

I've found Copliot to be really good at generating funny memes and haikus for specific issues/tasks where I work. The productivity gains come from me not having to use Photoshop and have more time to browse websites like this one.

In some cases it is useful, like in Excel where it can generate formulas or describe an approach to a problem. Not different than GH Copilot. The same for MS Power Automate editor AI assistent.

But in other cases like Word or Outlook it is just a louzy summarizer and does not have much added value.

That's surprising. I haven't met a single user who found Copilot useful in Excel.

I've found some that enjoyed using it a "smart auto complete" both in Word and Outlook, however.

Most of my users are financial analysts.

Interesting because I pay for Copilot and Supermaven, although most of my coworkers don't. To be fair, it was twisting their arms to get them to use linters, formatters, and other tools so I think asking them to use AI auto-complete is a bit much right now.

That particular issue also exists in GitLab. See https://gitlab.com/gitlab-org/gitlab-runner/-/issues/2797

Right, but Gitlab does have the excellent built-in pipeline editor that will visualize and validate your pipelines for you.

It can also render the complete pipeline config (making it easy to run and debug the problematic parts locally just by copying the relevant parts, even if they're hidden in and include somewhere).

That was not my personal experience. CS and Warcraft 3 community lobbies featured rampant cheating. Way more than CS:GO and Dota 2.

I have done the following in the past:

1. pip install libfoo==1.x.x

2. pip install libfoo==2.x.x --target ~/libs/libfoo_v2 # vendor libfoo v2

3.

import sys

import libfoo

original_sys_path = sys.path.copy()

sys.path.insert(0, '~/libs/libfoo_v2')

import libfoo as libfoo_v2

sys.path = original_sys_path

There are caveats of course. But works for simple cases.

RPG Maker brings me so many good memories... highly recommend!

Why would any enterprise use e2e encryption?

In most juridictions and industries you must be able to audit internal and b2b communication.

so that when someone pwns your chat server, they don't walk off with all your communication history.

If you want audit, you then add it on separately, in a separate locked-down deployment, compartmentalised from the rest of your infra and the chat server, so that an attacker would need to pwn an audit client connected to that instead.

Might want to let Google know this

... So that the content of the chats isn't trivially read by unauthorised third parties?