Hacker News Reader

Comment by bpasero | Hacker News Reader

We maintain a single VS Code setting that allows you to opt out of the AI features provided in VS Code: "chat.disableAIFeatures" (see also: https://code.visualstudio.com/updates/v1_104#_hide-and-disab...). If you can still find AI features appearing after you have configured this setting, then please report an issue at https://github.com/microsoft/vscode and we are happy to take a look.

It is possible that from time to time a new AI related feature slips in that does not respect that setting, but we try our best to push fixes as soon as possible.

Thanks! Ben (VS Code Team)

View on HN

Bravo, I respect that VS Code has added a single setting to disable all AI features. It prioritizes user choice and agency. Considering there was a recent rebranding as "the open source AI code editor", it means a lot to new and existing users that there's a choice to not use AI.

For many companies and products it's apparently hard to do these days when LLM integration is the hot new thing pushed by management and investors. Developers, users, and citizens deserve the respect and right to opt-out from AI features as it permeates other areas of work, life, computing, commerce and governance.

I'd like to attach to this comment to say that we should support smaller companies. It doesn't matter how responsive a big company is if it controls too much surface area of the important technological salients.

Large company hegemony of our industry is bad. VSCode, Google Search + Chrome, mobile phone duopoly, Amazon/AWS/MGM/WholeFoods/TeleDoc conglomeration and cross promotion... It doesn't matter. We need more distribution of power.

Fun fact.

I do not financially support any restaurant that has a Wall-street ticker. I wish more people would do this. There should be no reason to fund some CEO on Wall-street when we can benefit more by funding local communities.

P.S. You have to pay me to use Microsoft products and to engage with Amazon.

You can pry Cheesecake Factory brown bread and Olive Garden breadsticks out of my cold dead hands.

Wouldn't it be a shame dying with those in one's cold dead hands, without ever learning how much better options exist?

tell me something tastier than olive garden breadsticks

The breadsticks are also dairy free! Shocking but amazing for my poor lactase free stomach

Dipping them in a never-ending choice of marinara, five cheese, and Alfredo dip.

Garlic bread at any half-decent Italian restaurant?

I’m quite aware that better options exist. I’ve dined at multiple or different 2 Michelin star restaurants, not to brag or anything.

But sometimes this “corporate bad” mentality is just vapid snobbery. I’m better than you because I don’t support big bad corporate.

Of course, companies aren’t created equal, regardless of size or status of being public or private. Some are run very well and ethically and some are not.

I am sure we can find many mom and pop businesses that do shady things that no public corporation would be caught dead doing.

Did you know, small landlords are exempt from equal housing laws? Mrs. Murphy exemption.

If I go to an Olive Garden I know I’m getting the exact same experience everywhere, I know exactly what amenities and facilities they’ll have, and I know what price I’m going to pay.

Even though Cheesecake Factory is a public company they’re doing more real kitchen prep work and in-restaurant cooking than my local bar and grille that’s reheating premade Sysco food.

It's not about "corporate bad", it's about those two doing exemplary bad food. That's beyond any shady practices or whatever (not our topic anyway).

At least Olive Garden, Cheesecake Factory does better.

I think this is quite ironic actually.

If I understood the history correctly, being a "shareholder" was a path to a fractional business ownership for people who could not afford to outright own a business.

It comes from the same mental position as a co-operative.

In these scenarios, a CEO is really just an employee of sorts for the shareholders.

It's quite funny that we see the CEO of a publicly traded company has worse than a sole-proprietor, when profits will go directly to a sole proprietor- but not to a shareholder CEO.

I understand how it has played out, that the largest companies on earth are publicly traded now, and that CEO compensation in those companies is crazy. But it's quite ironic in my opinion how it played out.

Given the top 10% holds 87% of shares, it seems clear the stock market is primarily a tool to compound wealth. Having a surplus of money is table stakes to play.

Not just that, but if it were "fractional ownership in a business" then every profitable company would pay a dividend. Now it has become normalized to not do such a thing with all sorts of fancy reasoning.

Share repurchases are also a way of returning cash-flow to owners. A stock buyback is equivalent to a dividend, just with automatic reinvestment in the default 'do-nothing' case.

Including buybacks, few large and mature companies fail to return profits to shareholders at all, and we'd ordinarily want growing or startup companies to retain earnings and invest.

You are just reciting the "fancy reasoning" that I already mentioned. All the esoteric stuff is designed to do is concentrate the wealth into the hands of the wealthy. It is a step away from fractional ownership.

There is nothing fancy about it. The reasoning is straight forward and has the exact same impact that paying dividends has, from the perspective of the shareholder. (Well, actually in some ways maybe a better impact, as it makes complying with idiotic tax laws easier)

>It's quite funny that we see the CEO of a publicly traded company has worse than a sole-proprietor, when profits will go directly to a sole proprietor- but not to a shareholder CEO.

Speaking in the same mindset as the parent, we're fine with the profits going directly to a sole proprietor.

In fact, what we want is a name attached to the profits, and a not a role.

We're not anti-profits.

We're anti bland corporate leadership, with no reputational risk and no personal ties to the company (and often no financial risks either, see golden parachutes) - one whose only mission is to maximize profits, product and customers and legacy be damned.

This is a totally fair point to make.

As mentioned though, the irony is in that, once upon a time, the lords and landed gentry were the "bland, soulless overlords" and so buying things would improve their profits... and nobody could become them.

Then the idea of fractional ownership came about and the common man could buy in to an enterprise.

Now of course, everyone is correct that this has been weaponised- but it's often interesting to go back to the original intent (or idea) of something to see how warped it has become.

I recently moved to Wisconsin and decided it was only appropriate to only buy local. Sure, it tastes better but I was surprised to see it was cheaper too. I guess it doesn't matter how much of your food is plastic and sawdust if you have to pay to haul it an extra hundred miles.

At first, I thought you meant restaurants which had a big scrolling ticker along the wall so patrons could watch the market while they dine.

Just my preference, but I'd like to know where said restaurant is, lol

When they brand themselves like that, it’s a clear sign for people who dislike “AI” not to use it. Zed having some opt-out “AI” features now was one of the reasons for me to stop even considering properly trying it again.

Emacs it is, still.

https://news.ycombinator.com/item?id=46534071

Having spent years doing software dev myself and now months with the latest frontier models, I don't see a path forward in any software-development capacity that values productivity where AI is not featured.

So I hope for your sake that you and others like you are pure hobbyists. Because anyone paying you to do this job is going to be firing you pretty soon in favor of those who lean on frontier LLMs, otherwise.

Hey, Ben.

I recently (less than 2 months ago) did an in-depth analysis in the area of license compliance that suggests that Microsoft and many other companies that are shipping Electron apps aren't in compliance with the LGPL. (By all signs, it looks like the Electron project might not even be aware that Electron is subject to the LGPL, though they are. Even Slack, which isn't violating the license appears to be in compliance only incidentally—because they're shipping other LGPL components that they know are LGPL.)

I was set to leave the company I was at a couple weeks later (end of November), and I did, so there haven't been any developments with my investigation/findings since I departed. I haven't prepared or published a formal write-up, and I've only brought it up in a semi-public setting once. It's a pretty big deal, though. Could you raise this with Microsoft legal (not Electron/GitHub) and suggest they look into this?

Assuming this is real and you have the authority to share your work from previous location; you should reach out and contact Microsoft Legal directly.

A random engineer on Hacker News is not the proper channel.

Link: https://www.microsoft.com/en-us/legal/compliance/sbc/report-...

I'm not an engineer, and no one should be getting the impression that anyone else is under the impression that HN is the place to seek an authoritative disposition about this. It is, though, an acceptable channel for the sort of collegial and informal heads-up that this is (and which is all that this is).

Your desire to condescend, however, is noted.

I think it would be interesting for people if your comment was a little more specific about what the issue is. Is this about ffmpeg as raised here: https://github.com/electron/electron/issues/34236 ?

You’re not the random engineer. Ben, the commenter you’re replying to, is.

You were given helpful advice and a link. I don’t see this being condescending.

Correct, that was my intent - Ben isn't the proper channel as he is just an engineer responding to comments here. Stuff like this is serious and so should be escalated.

Compliance with FOSS licenses isn't a joke.

You misunderstand.

Ben is a random engineer, he is definitely not the proper point of contact. FOSS compliance is serious, so if this is real, do escalate it.

The guidance you offer here remains as necessary and is as appreciated now as it was the first time. Rest assured that I am capable and well-informed about how to proceed with these sort of things. Warm regards.

This is just attention seeking, hard to imagine that after having worked there their best contact is a random person on HN.

GGP didn't say they worked at Microsoft, the comment is a bit hard to parse, but they wrote "I left the company I worked at".

Scant on details, sure, but hard to parse, not really.

The problem is folks this thread seemingly taking a interlocutory approach that can be summarized as, "That which is not explicitly denied can be freely assumed to be true."

(Then throw on top of that, "Depending on how committed you are to your grandstanding, that which is explicitly denied can be conveniently ignored.")

I'll give you another example. The "Microsoft Tunnel Gateway" is a end point for Microsoft's InTune VPN downloadable as a docker image for Linux from here: https://learn.microsoft.com/en-us/intune/intune-service/prot...

I had a brief look at the docker image, and it's pretty clearly a repackaged version of OpenConnect. Debian's copyright linked to from https://packages.debian.org/sid/openconnect says it's primarily LGPL but with a plethora of other licences like the GPL.

Since there is GPL they are required to make some source available, and if they modified it they are required by the LGPL to make their modifications available. They have extended it by adding Microsoft's authentication mechanisms, but perhaps that is just a DDL mixin, and I could well believe / forgive them not being aware of the other licences.

What is not so easy to forgive is them not acknowledging the open source they used in any way. Instead they slapped as pretty standard Microsoft Licence claiming it's all theipr own work, similar to this one: https://support.microsoft.com/en-us/office/microsoft-softwar...

Make it opt-in only at welcome page, and we have a deal.

(Of course, I know that's never going to happen.)

I really don’t get this attitude. What deal do you mean? You using VS Code? Why not change the setting simply?

Because the point is just to loudly proclaim how virtuously anti-AI you are, how disruptive it actually is to your workflow is irrelevant.

glances at Windows 11

No, I think the point is to escape encroaching monetization that dilutes the value of local on-device text editing.

Oh, I use coding assistants every day, just not the one that came with VSCode. Because I want to decide if/which coding assistant I want to use, not whatever VSCode forces upon me. In fact, at many companies, GitHub Copilot is explicitly forbidden.

Not the smartest argument to brand this as anti-AI.

I really like the Copilot autocomplete across multiple symbols in the file (e.g. predictive edits that you can tab through).

For most other stuff I prefer Cline/RooCode/KiloCode, but sadly it doesn’t seem like any of those offer similar autocomplete (Continue.dev did with even Ollama support for local models but the whole plugin was a buggy mess and it didn’t work well). Oh and sometimes Claude Code or Codex is nice in a terminal directly.

Personally, I don’t mind something being there by default (same as how JetBrains has their pre installed plugin and also something like Junie available), as long as it’s easy to turn off or uninstall.

Similar to how I wouldn’t scoff at a Git integration plugin even if I prefer to use Sourcetree or GitKraken.

> as long as it’s easy to turn off or uninstall

That's the issue here.

The "disable all AI features" option isn't really easy to find.

In my experience it’s people with “agents” mass editing their code, in their 10th attempt to convince their tool to do what they want it do, are people with a disrupted workflow, in a constant struggle with their tools.

You should really look in to the difference between opt-in and opt-out. Opt-in respects the user; opt-out is for foisting features that the user might not need or want.

Flip it around: a new user might be confused why the AI features that all their friends told them about are not available.

It's a tradeoff

If it means more users aren't using Gen AI, I'll happily take that tradeoff!

If it is opt-in for everybody, then their friends would also tell them that they have to opt-in to get it.

The anti-AI folks should just fork everything at this point, because it's hypocritical as hell to complain about it and use a bunch of stuff built with it. Then you can opt out of society!

Oh no, just think of all the good AI-produced things we'll be missing out on.

...are there any?

I'd say the percentage of stuff developed using AI now is higher than the percentage of pro athletes who use performance enhancing drugs, and there's almost as much incentive to mask it and say "made without AI"

> Why not change the setting simply?

My default settings are stored in a 11922 line json file.

Am I expected to read that entire file to find the setting I'm after?

Am I expected to do so when I don't know what the setting is called?

The reason you can't simply change the setting is because the setting isn't simple.

It's essentially a hidden setting, cloaked behind an ambiguous name in a user-hostile manner.

> My default settings are stored in a 11922 line json file. Am I expected to read that entire file to find the setting I'm after?

That’s what AI is for. Have it turn itself off.

Ah, but if AI was correct, I wouldn't need to turn it off.

It's a tool for programmers. Use google + ctrl+f. Not a hill to die on.

How is googling it meant to work when they keep changing what the AI settings are called each month?

Alternatively, use the single setting that was literally just given to you above. That is pretty much as easy as it gets without resurrecting Clippy to help you figure it out. It's not reasonable to expect a massive bloated gui if people have 10k+ settings they are using.

The problem is that they keep adding new ai "features" all under different names and different settings, then shuffling the settings around.

Having a "master switch" doesn't matter, since their standard operating procedure is to waffle-stomp more "features" into vscode every month that will fall under a different setting and then they'll continue to shuffle them around.

Their indifference towards their own user-hostility with regards to this is the main problem.

I'm confused—how does putting in a master switch for those who want to opt-out entirely from the AI revolution occurring at the moment not matter? Are you saying that new features will fail to respect it?

LLMs are pretty good at this sort of thing.

I use emacs, so maybe they're better trained on my editor. But I've had a lot of success resolving little annoyances I have just lived with for years talking to Claude in gptel.

I can't get it to do real work for shit, but it's A+ at helping me waste time with yak-shaving. lol

> My default settings are stored in a 11922 line json file.

And I thought my 50 lines settings.json is getting unmanageable and needs some cutting. WoW.

I'm guessing the user you're replying to is meaning that the vscode default config file is 11k lines and the AI setting just one of many lines in the file.

I don't think they meant that their own settings are that long, just the default in the app and they're commenting that it's ridiculous to expect a person to find it there.

Ah, well, I found this AI setting via github issues browsing, LoL, so yeah it's kind of hidden. May be there are better ways.

That, in a nutshell, is one of my biggest complaints about VS Code: there are many overlapping settings for various things, and I could never get clarity on what takes priority. Setting up things like formatters (and their settings) for various file types was a nightmare; between VS Code complaining I didn't have one set up (I did), the settings seemingly being ignored, and various other issues, I break into a cold sweat whenever I have to edit my settings file.

But more to the point, I don't understand why one would ever have to edit the file directly when there's already a settings panel that lets you search for a setting using natural language, and get back a list of matching settings. Why doesn't VS Code let you make all the changes from the settings panel, without having to mess with JSON directly?

I appreciate this intent. I do wonder: do any of your team use this setting? I, too left VS Code after having to disable Copilot for the umpteenth time in a fit of frustration. I was aware of this setting and had it to disable all AI, alas.

The frustration for me is that it turned my editor into a 2000s-era popup extravaganza (not necessarily anti-AI). Every line of my editor was constantly throwing a new popup or text to the side of my cursor. I know that VS Code's design philosophy has moved toward trying to make the editor have as many pop-ups as possible, but there are still a lot of us that don't think that's a good way to focus on the work. It is beyond frustrating when every week or so your editor decides you're wrong about that.

I'm just so extremely uninterested in a text editor whose main selling point is that it's an "AI code editor".

I actually moved to Zed for AI, Claude integration is seamless and the IDE is so fast and crisp. I can use what ever agent I want, when I need them instead of being asked to use Copilot.

Except for few language related extensions, I don't have any other extensions on Zed. Which means I worry less about which of those extensions will be sold off to a malware developer.

I had more issues with official extensions on VSCode (looking at you flutter) than not having any extension on Zed and having to rely on the terminal (which feels much closer to the system than it did on VSCode).

Could you provide some detail on how you use it (the AI features)?

I just have a minimalist setup.

Configuration -

In external agents - I have Calude Code, Codex CLI, Gemini CLI and a custom agent for Qwen Coder via Llama.cpp [1]

In MCP - I have fetch, brave-search, puppeteer

In LLM Providers - I have configured Llama.cpp, LMStudio and OpenAI (Zed Agent can access models from any of these providers).

Workflow -

When I need LLM assist I mostly use just Claude Code for specific tasks, with thorough scaffolding. One major drawback in using external agents on Zed is that they don't support history[2], Which doesn't impact me much as I use Claude just for individual tasks. I'm not really sure on how well Zed works for someone who 'vibecodes' entire project.

[1] https://zed.dev/docs/ai/external-agents

[2] https://github.com/zed-industries/zed/issues/37074

Thank you!

VS Code is a flagship product from Microsoft. Relying on “we try our best to push fixes as soon as possible” for a global opt-out setting feels below the engineering standards we’d expect. This should be fail-safe by design, not best-effort.

Everything in the world is best-effort at best. People who tell you otherwise about their pridhct are just lying to you. Do you prefer being lied to? Criticizing people for being honest about it is a good way to be lied to more.

I’d rather someone say “we architected this poorly and are working to fix the design” than “things slip through sometimes, report them when you find them.” The former is honest AND shows they understand the problem. The latter treats a design flaw as an acceptable steady state.

Being honest about shipping bugs is good. Being honest that you’ve designed a system where the same category of bug will keep happening? That deserves criticism, not praise for honesty.

Could you perhaps one day make "chat.disableAIFeaturesMadeForShareholdersOnly"?

So that we can have the actual good stuff (copilot, chat) and leave out the mountain of features that were clearly created to force induced demand for the sake of metrics inflation?

You can probably phrase this question as

* Can we enable the only features we want by toggling chat.disableAIfeatures and only selectively enabling copilot and chat?

Make it a separate build that doesn't even have any of the AI code in it because a flag can't be trusted. Especially when there's a history of resetting flags after updates.

Btw, Zed has a similar setting to turn off all AI features :)

And I'm sure they'll never forget to connect an AI feature to it.

Your point here being? In that case people will complain, and the next release will have it included into the setting

Thanks Ben, I wasn't aware of that.

just wanted to let you know that this comment is the only reason why i'm using vs code again after about 9 months of not using it at all.

Thanks for that setting! TBH I'd prefer for vscode not to ship any AI features and let user install them from extensions, but managers probably will not be happy, so at least that setting helps to stay sane.

> It is possible that from time to time a new AI related feature slips in that does not respect that setting

You mean marketing forces you to "accidentally" slip it in? Just in case it sticks this time?

I don't know about marketing but if code reviews are done right I would love to know how something like 'AI feature' just slips in.

If the code looks correct, why not?

Perhaps people pushing AI slop are not that great at doing good code reviews; which is a shame, really.

They have the "AI" do the code review of course.

And don't imagine the LLMs are sentient, the directive to cram "AI" features down the users' throats is in the prompt.

I see a lot of AI features hate towards VS Code. Want to say that there are other people like me who really love these features, just want to thank you as a Copilot pro user!

Would love to know what happens in perf review for MSFT employees that flip this switch.

Nothing. This isn't something that is even tracked. AI usage is obviously encouraged but HR has far better things to do than go gather this kind of data.

Internally, depending on what product is being worked on teams will have different development flows and different usage points of AI. For things like VSCode, teams have freedom on how they use it completely.

This is not true company wide. I’ve personally been reprimanded for low GitHub copilot usage by my orgs leadership, not HR.

Whoops!! Pesky mistake. An AI feature bypassed the AI opt-out setting yet again. What are we gonna do with all this data now??

Reply on news.ycombinator.com