In the age of AI there's ZERO reason for "the global North" to continue using Microsoft. Wean your finance guys off of Excel, somehow. Replace the CTO with somebody that can hire IT that doesn't depend on the crutches of Microsoft.
If people's MS subscription revenue were funneled into on-site devs and devops for open source solutions, we'd all live in a better world. Sure, the money wouldn't be funneled to a single ledger in Redmond, but the overall productivity would be higher and there'd be greater GDP growth in other areas.
I've heard of so many small companies with a slightly technical bent replacing their SalesForce with better CRMs that make everyone happier, thanks to AI. MS is next.
Let's make sure that dollars get funneled away from economic rent and into more productive efforts. Schumpeterian rents might exist for a bit, but they should never exist long. Microsoft's rents have lasted for too long to be economically optimal.
All roads lead to Excel. When you try to make a nice dashboard in Jira that incorporated custom fields and all the data was normalized, nope - they send out an excel spreadsheet every time. They do this to any tool you try to use as a group tool to update things - like Confluence. Before you know it it's back in Excel again and you have multiple sources that keep getting lost.
I think the criticism this post makes applies regardless of how you get your spreadsheets. It would apply equally to Google Workspace.
Plus, I guess there are office 365 outages, but the outage I personally hear more chatter about is the hosted exchange services (outlook), you will still need an email server in a post ai world, right?
But firstly, it's a legitimate choice for normies who just want to run a business, and second, as I said, the criticisms this article brings would be applicable to any other big tech provider who is laying off staff and leaning heavily into LLMs.
It's not possible to just run your own email server anymore. You can try, but nobody will interact with it, no matter how long or how hard you work to build your reputation. Fastmail still goes to spam, and they're very large and dedicated. It's hopeless. You use one of the big five or so providers, or you're simply done.
The problem is never Excel, its user and permissions management. Apart from Google Suite there is not really any other one stop shop for all of that, and definitly nothing open source (for now)...
Office 365 / Azure AD subscription is cheap. It comes with legal certifications that are very expensive to obtain alone and usually shareholders pressure you to get them. It has remote management capabilities and a very alive closed source plugin ecosystem that's battle tested.
How long would it take FOSS to replace this? How long would businesses and their shareholders will wait? Buying Microsoft will GUARANTEE that you'll have the certification you need in weeks, how long would FOSS would ship and PROVE that they obey a non-open "open" standard?
Will FOSS GUARANTEE cross-compatibility of the file format?
What you are proposing isn't stupid if the most important goal is independence. Most businesses don't care about independence. They PREFER dependencies they can blame and sue / settle with.
I think you are massively underestimating the engineering effort of replacing Excel in large organizations. I don’t think the finance guys can reliably vibe-code an alternative to a massive product that’s at the core of their operations.
Handling simultaneous edits, guaranteeing correctness, handling massive data sets… as amazing as AI is, this is a huge can of worms.
You don't need to vibe code an alternative to excel, you need to vibe code a solution to the problem you're solving with an excel sheet.
Any nontrivial excel sheet should be a database backed application, the finance guys who build them are in every sense programmers they're just using the wrong tool by convention.
You're missing the point that the users don't want an application. They want to go off on flights of fancy altering the "solution" as they see fit. Excel lets them do whatever crazy thing they want without constraints. Having a database-backed application is exactly what they don't want because they can't be "creative" with it.
I agree with you, BTW, but I've never seen moving Excel sheets to an application work in practice. You can do it, and it just creates more Excel sheets.
* Redo UI patterns (for old users): Only matter for on-boarding and compatibility, and probably need to redo the anti-patterns, not necessarily a good idea, or at least must be hold before the next:
* Define a much better internal model than excel has (I bet a mixture of relational, array model as core, then a reactive layer on top), including proper support for correct datatypes (not localized!), dates in all shapes, currency, commodities, unit of measure, etc, tables default as fixed type per column instead of free-form, then we have proper "variant" values, use algebraic types, and so on...
* Then redo the UI
* Define how import/export correctly (so not more non-sense about import based on localized dates, currencies...)
* ... and from legacy formats like Excel
P.D: I explore the idea on https://tablam.org and my goal is eventually marry what excel+access do in one app
If organizations can't move off Excel, and outages like this continue to be a problem, then competitors that don't use Excel stand to gain an advantage.
Actually importing gene names was possible without having autoconverted names, you had to make sure you were importing files using an import tool, currently listed under the "Data" tab and then importing data from file/web or data, or use the button specific for .csv/txt. In this way, Excel would open gene data in a table editor without renaming gene names if you made sure, this has been possible since at least 2005 I believe, or maybe even earlier. It was critical to not just auto-click or open a .csv or .txt file with Excel, but specifically open a blank or existing workbook and then use the import tool from inside the application. Opening data table files directly with Excel was/is typically bad practice.
Excel is still the tool in Finance.
It is the database, the visualization, the calculation enginer, the collaboration tool, and many other things all rolled into one.
Edit: actually I think there's got to be a far better and completely different UI that's possible now with AI. There has to be something better. What is it? I don't know. Some clever specialist will invent it.
Just throw the raw data at the LLM and tell it to do the analysis directly on that, and there's a good chance it will use python and some data science libraries for you.
And if you want, it'll turn it into a web page where people can drag and drop more raw data files on to get updated analytics.
No. Excel really doesn't have an equivalent replacement. Other parts of the office suite do although if you're a lawyer (revision marking) or have other specific integration needs, you probably need Word.
If you're just doing lists with maybe some simple calcs (which is what most people use spreadsheets for most of the time), pretty much anything works.
May I ask for a few examples of things you do in Excel that don't have a replacement?
The vast majority of people I have seen using Excel don't know how to write a formula. Now they just started copy-pasting whatever ChatGPT was telling them to copy-paste, but well... and anyway all the alternatives do support formulas of course.
There are plenty of functional replacements. However there are zero replacements that emulate the same keystrokes and keyboard shortcuts and all other aspects of hard-won muscle memory like function names and quirks of behavior.
When companies don't want their employees to work remotely and force them to come to the office because they think it's more profitable, it's not like they give a shit what the employees think.
Why would they give a shit about the little comfort of the employees who don't want to learn new shortcuts?
It's been years since I've really tried to exercise Excel (and haven't used it--or Libre Office--in a long time). Pivot tables, a number of the other data-related functions... Just a bunch of things I'm not going to spend a couple of days digging into. Google Workspace works fine for my needs.
For my use case, mainly science stuff, works like a charm. I find Calc better because I could never get used to the ribbon, and it's also way faster.
I actually did my whole thesis in LibreOffice, including all text, references and figures. Afterwards, during the PhD, it was replaced mostly by Python and R, but at that point I was really hitting some resource limits with huge CSVs.
maybe this could be used as a base and people could use AI to create extra features they need on top of it if there are things missing that Excel has that they want
This. Judging from replies there are a lot of folks here who imagine Excel usage being something on the order of basic readouts with some SUM()'s and MEAN()'s.
The real grip comes from entire business processes implemented across dozens of worksheets with incredibly complicated cross linked formulae relying on Excel edge case behavior. As soon as LibreOffice et al display the most minute differences they're done and will never be mentioned again.
FP means functional programming and SAMOP means "simply a matter of programming" and the humour appears to be a Whoosh moment. Cell based spreadsheet programming is good, and should be portable. But people code to Microsoft bugs.
This would be an unmitigated disaster - most programmers have trouble with Haskell, I can’t imagine saying hey finance guys here’s an LLM and Haskell figure it out when they’re used to using a full on application with UX. It’s actually so crazy I feel like you’re joking.
The serious answer here is that a lot of MS products have kindof dual purpose: they are 90% used for very simple stuff (collecting data into a spreadsheet, having a text editor that allows basic styling), and 10% stuff that power users rely on (huge excel files with complicated formulas for data transformation and visualization, word / ppt templates with byzantine branding). The answer is that for the simple stuff you rely on open / cheap alternatives, and for the complex stuff (eg visualizing and transforming data for non-programmers, generating presentations and branded text documents) you can just ask ai now.
Big business kind of wants 24/4 support that those will not be providing at all though? Plus the various nonsense certifications. I mean try to convince anyone in corporate to use Libre and they'll probably just laugh.
(not that they would ever use that support, they just want to have someone to scapegoat lol)
Row Zero (now part of Databricks) is by far the best spreadsheet on the market - it's not even close.
Row Zero's fully Excel compatible, including keyboard shortcuts, long-tail pivot table configurability, 500+ built-in formulas like LAMBDA and LET, and real-time collab. But it scales 1000x beyond Excel (2 billion row limit vs 1 million for Excel), has built in connectors to everything (Databricks, Postgres, S3, Snowflake, etc), native Python support, deep AI integration with Databricks Genie, and all the security and data governance features to make big banks and healthcare providers comfortable. It's built by the team who brought Rust into AWS and rewrote S3's file system circa 2018.
We're used by individuals and small businesses all the way up to huge deployments in the Fortune 100.
Source: Was founder and CEO, now leading spreadsheet development at Databricks
R and Python are both obvious replacements for data processing, and having the data itself appear in read-write columns of an Excel file is and has always been ludicrous.
Excel was for a time when programming meant more arcane languages. In 2026, there is no excuse for a business major not to know Python or R. They're both easy.
These products all have super deep feature sets. This might meet basic editing needs. There is zero chance they have replicated the full feature set. Word for example has a ton of functionality for lawyers.
I can tell you all of the Excel alternatives are extremely disappointing.
I do agree with that but I also think ultimately that a copy with all the most basic features and the familiar UI is good enough for most users, certainly for me
You write in the imperative, as if advising a leader.
However, leaders are advised by their underlings, who are swayed by their technical officers, who are charmed by the commercial arm of corporations.
The only effective path to actual change is to charm the technical officers with a product that is strictly superior in the metrics they care about. They'll then work their way up the chain.
> You write in the imperative, as if advising a leader.
We should stop referring to management at companies as "leaders". They don't lead anything. They perform administrative tasks so that their employees can actually do the real work.
> The only effective path to actual change is to charm the technical officers with a product that is strictly superior in the metrics they care about.
Probably not gonna work: those "technical" officers are mouse-click engineers. IOW, they don't know much detail, but they sure as hell know where to click!
With respect: you're a biologist, and your experience with backoffice software requirements is probably the same as my experience with biochemistry: near total ignorance.
There are exabytes of documents in Microsoft formats that can be at best "read" with a significant level of data corruption in other products (formatting errors). Writing them back out with shredding them is hit & miss at best. Entire MS products have no equivalent at all that can open their file formats, such as MS Access and Power BI.
Word isn't just paragraphs, italic, and bold.
Excel has features you probably have never even heard of, but people use at scale. Things like tabular data connections using OData. It's the front-end for SQL Server Analysis Services. It has the Mashup columnar engine from Power BI built in to it. Etc, etc, etc.
These things are massive products, with deep integrations across each other, third-party software, third-party plugins, cloud products, and on and on.
There's easily several million man-years of effort sunk into the Microsoft Office ecosystem.
Switching costs are not a matter of failing to "somehow convince" people to switch.
The switching costs are on the order of trillions of dollars, not a decision to be taken lightly.
Here's some critical points everybody normally ignores in these discussion:
1) Partial switching doesn't help, because now you have two problems!
2) Everyone else has to switch too. Google Workspace people were laughing during the Outlook outage, eeeeexcept... they couldn't send or receive emails either -- because their business partners, vendors, and customers likely used Outlook.
3) The alternative with the same features, same level of integration, same commercial support, etc... is the same thing, with the same problems, but just a different brand.
You can't escape the reality of this, it is what it is, there's no nirvana that we could all achieve if we "just" did something "simple".
> Excel has features you probably have never even heard of, but people use at scale.
Yeah I've watched Joel Spolsky's "you suck at excel" tutorial, but I've also created financial models for CFOs and the finance team using Excel instead of doing an RShint App because I wanted them to be able to understand and extend the model.
It's software. It can be recreated, more easily than ever.
I'm saying people should put their MS license fees towards in house devs working on open source projects to achieve the goal, rather than just let MS Hoover it up to waste on Copilot interfaces that nobody ever uses. Let the local devs squander that money, after they have reimplemented all the Excel keyboard shortcuts in OpenOffice and added abhorrent plotting themes that recapitulate 30 years of incompetent Microsoft practices.
One example would modeling the costs and revenue for a business unit, the prices from suppliers, the fixed costs of capital, the number and type of employees, the throughput from sales, etc.
One of Excel's many powers is letting people play with numbers in cells and see the results of long chains of computations. Or create an array of potential values and see the results across a wide variety of inputs. Or even use the builtin solver to find the value that minimizes or maximizes a desired output...
> In the age of AI there's ZERO reason for "the global North" to continue using Microsoft. Wean your finance guys off of Excel, somehow.
What does AI have to do with it? Are you proposing that we replace Excel for accounting purposes with something vibecoded? That sounds a bit nuts to me. You can pay a bit of money for mature software that will probably not fail to add 2 + 2 correctly and will not run away with your wife, so why give that up?
I actually suspect that companies such as Microsoft pushing for AI everywhere may undermine that proposition... but then, the alternative is probably LibreOffice Calc, which is eminently boring and doesn't have Copilot popping up.
After 30+ years of hearing this type of twaddle, it's more likely for Microsoft to keel over because of its own leadership incompetence than to be so much as dented by any of these half-baked community efforts.
Oh please, stop it everyone with this thrite trope of the fabled finance guys with their crazy spreadsheet that holds the whole company together. 99% of Excel use is NOT that!
What MS365 does best (for now) is being a one stop shop for user/permissions/device management and coming with outlook/exchange/Teams - nobody with purchasing power gives a shit if ppl get Excel or LibreCalc as spreadsheet, they get it because it comes with the rest.
Unless you spin up an entire devops team plus an ML team to run inference at scale in-house, you're just changing where you funnel your subscription money.
Also: No home-grown CRM will have, in the long run, uptimes anywhere approaching e.g. Office365. You're just smearing one large outage into a lot more small ones, plus a few data losses.
That rent buys you something. The internal development buys you something very different. You might want to make a reasoned call on what you're buying instead of making GDP predictions.
The ONLY WAY to get rid of MS Excel is to make something better and then literally force feed it to the finance folks who will go kicking and screaming all the way.
> Now, are Microsoft going to offer any compensation for either lost business or additional accrued costs? I highly doubt it, even though it is literally their fault and they do still have the cash reserves, despite increasingly wasting it on loss-making AI adventures.
Aren't they? I know that they, and other big techs, have clauses in their terms that say if they don't meet service requirements they will offer refunds/credits. I've never been in a position where I was the one requesting the refund but I always presumed they actually did it.
Some of us software engineers have been predicting the software apocalypse since over a decade so what's going on now is not surprising.
What else won't surprise me is the government trying to step in and making it their problem. That's what I'm worried about. The only right thing to do is remove regulations, remove government intervention from every aspect of the economy, and let the market fix the mess which the central planners have created.
Designing the entire software industry as a giant government-supported oligopoly was not a good idea. It was unethical and it was not going to work in the long term.
"If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization." - Gerald Weinberg, 1970's
> let the market fix the mess which the central planners have created
Isn't a Microsoft dependency failure market-driven? Compared to the 1960's and 1970's, the present-day software industry is much more driven by the market than it is by the government.
Not only they are concentrating power in the hands of an oligopoly, but they're making the whole country unsafe. If everyone uses the same datacenters, it is enough to shut down a few of these data centers to bring civilization to a halt in this country. With the current AI trends, this risk is getting ever greater.
Yes it seems very dangerous. Also, I think a lot of the software we use today, we don't need or we could make much more efficient. We almost certainly don't need so many data centres. Most businesses could self-host their own systems. Even a lot of large companies.
I feel like we probably only need a tiny fraction of the compute capacity if we let things be decentralized and let the market operate naturally.
This was my thought years ago and so now it looks blatantly obvious to me with AI data centres driving electricity prices. It's pretty clear the industry is being propped up by other parts of the economy.
The real "central planners" are the billionaires who have led the way to oligopoly since the 80s. Not sure how handing them more power is going to help distribute it.
backup plans should be discussed (both backing up the data and alternative services to use when services go down), and also switching to better alternatives to Microsoft when possible
Hot take: most people, in most industries, should be able to tolerate a day of downtime in the sense that there are still useful, productive things to do even when all the computers die.
The fact that the world relies on four 9's of availability and throws a hissy fit whenever there are outages is an antipattern.
And also, people should get a snow day every once in a while.
I've often wondered how much unplanned downtime is caused by the aversion to planned downtime.
Like if Exchange Online went down for an hour every Sunday for maintenance, and they made their network configuration change then, there would have been no cascading failure.
This is how you get enshittification. It'll never stop at a day every once in a while, unless we hold those with billions to a higher standard, instead of giving them excuses while the layoffs continue.
Or maybe, if everyone was willing to tolerate occasional downtime, they'd compensate for it by not having single-sources of failure on the things that go down, and thereby fixing the structural issues that allow enshittification (dependence that the companies running the services get to use to squeeze users)
When you say "they'd compensate" are you referring to us end users trying to do banking when MS or CloudFlare go down? Or are you saying the big corps themselves would compensate? Enshittification stems from there being no incentive for them to do so, because they answer to shareholders not customers or competition. If you actually want structural issues fixed, I'd implore you to please hold these organisations to a higher standard of quality. I don't understand how average folks tolerating downtime leads to structural issues being fixed.
It's a single Point If failure. Microsoft makes Sure that everything is running smoothly. But in the end, if 100 companies use Microsoft, 100 companies are affected if Microsoft has an outage.
Trust in others, across the board, is tanking. I expect in the not to distant future every company will have a completely proprietary stack, or they will be facing competition circling them while they can't keep their shit running. It is the mono culture that is too easily attacked.
It is entirely likely that with AI, you could just spin up the whole 'stack' as you need it without depending on external services. AI will truly drive the cost of software generation to zero.
Well, no you have it completely backwards. The scenario where every large institution generates their own software stack using AI is one where the sale price of software is zero (there is no demand) but there is a cost (it's whatever the AI companies want to charge, with almost infinite leverage since this organization can't do it alone)
Nope, it's you that is completely wrong. I can already run Qwen-3.8-27B on my 2021 64GB M1 Max Macbook Pro at a totally acceptable tok/s. The hardest part was waiting for the download to complete. It's free modulo cost of electricity.
I can build almost any piece of software I can dream up, the data never leaves my machine, and there's literally no way for anyone to charge me any price ever.
Yes, but can you verify that it is what you wanted it to be for free? If you're making a framework out of thin air it'll still be an undertaking to check that.
Yes? My time is a cost, yes. But otherwise, not really an issue? Does that change my answer to the previous question? Not at all?
I'm struggling to understand the point of your comment. Why would that affect whether or not I am over the pricing barrel with a frontier lab? Like is it easier to verify Opus code than Qwen code? The code is the code.
That would hold only if they are creating their stack over and over again. For the vast majority of what people want to do, make their purpose bound app once and go to work. Advancing tech for the sake of a new font loading do-dad or whatever is pointless. Look at all this crap we made over the last 30 years, and here we are back to preferring text.
Programmers are "program reviewers" now, more and more.
But the tools we have to do this suck. We don't have much beyond "here's the source code as text, look at it, and think really hard".
Another comment[0] suggested having AI generate "code walks" through the code. This seems better than our current approach of "my PR contains a diff, good luck".
I'm working on a tool in Emacs that allows me to mark certain lines, or entire projects, as needing review, and then helps me track which parts have been reviewed. I need something to help me find parts of the code I have never reviewed before.
As PRs get beefier I find myself using the LLM to ask pointed questions and do general surveys of the diff to flag architectural violations, smells, etc. before I read through the entire thing by hand. If you put all of your architectural and business decisions in the repo via spec-driven tools or otherwise then this becomes quite useful. It also helps if you have a vision in your head of what the system is supposed to look like structurally so you can ask the LLM specifically how the PR is handling things that you'd be on the lookout for anyway.
I've been thinking about a workflow where PRs just get automerged into a staging environment of some kind, and then tooling helps ensure that all code, ALL the code in the entire project, is regularly reviewed together.
When lines change, they are put up for review. When lines haven't been reviewed for awhile, they are put up for review. Lines are ranked based upon the complexity of the context they are in, and complex areas of the code are reviewed more often.
Basically, stop focusing on individual PRs and instead focus on continuous review of the entire system with tooling that helps identify those parts of the code that are most in need of review.
Reviewing PRs is wasteful because the optimal way to review a system does not match the way a system naturally grows through individual commits.
> Programmers are "program reviewers" now, more and more.
Our job, much like a locomotive operator, is to make sure the agents don't go off the rails. We call ourselves software engineers in recognition of that.
Big tech isn't here to let anyone down, it is your company executives and administrators who are here to let you down.
I have never understood why companies chose to standardize on Microsoft everything then switch the businesses to Microsoft Cloud everything in the belief that it was cheaper and better, simply because they don't want to pay money for onsite administrators or local IT guys for their IT. I mean if you want to use Microsoft Office then by regular licenses and run them on premises, but everything in the cloud who thought of that?
Many employees also don't know that their bosses or bosses bosses were also given kickbacks to switch their IT to Microsoft/Google/Oracle centralized services. For some obscure reason company bosses and government institutions still switch to Oracle for their services which don't work, wind up with Oracle negotiating with them for money for services which don't work, then they pay termination fees to Oracle if they bail out early.
Another factor I believe to be at work here is large investment firms like Blackrock, State Street etc pressuring the executives of companies which they are major shareholders in to switch their IT to the IT companies they are major shareholders in, therefore boosting the values of their stocks. For me it is the main reason for this centralization of IT services into Microsoft, Amazon and Google.
As someone who started using AWS from the time it went beta, I could see the value in it, but I couldn't see why companies had to switch their systems wholesale to it. I could see the benefits for new startups which had to move quickly and early and perhaps have a need for unexpected traffic, but even after some time the stuff would have to move back onsite once things settled down.
Other the end of the day the situation is entirely down to greed and corruption masking as incompetence.
A week or 2 ago in the UK virtually all the major banks went down due to some problems at Google, Microsoft or one of the other companies. So many banks had their IT hosted with them so everything went down simultaneously.
It is time the governments passed legislation forcing these banks not to centralize their IT with these companies and ensure they have their own independent redundant systems. Maybe they should go back to IBM mainframes or something.
PRs without problems don't really grant any credit against those with problems. I see it it a fraction of PRs. Seeing it sometimes is enough to establish evidence.
Indeed, given that there was a post this spring by ex-Microsoft employee on the poor state of Azure.
"The team had reached a point where it was too risky to make any code refactoring or engineering improvements. I submitted several bug fixes and refactoring, notably using smart pointers, but they were rejected for fear of breaking something."
Sure, those takes are mostly lazy (some of my comments here included TBH). Although I have very weird feeling, that current zeitgeist is about "coding is solved", etc and the same time we have evidence that didn't help much regarding boring thing such as reliability
> You have to wait for Microsoft to find a competent employee who wasn't let go to try and fix what was broken by some inept leftover who was deluded enough to think that Copilot can do everything.
AI does what you ask of it, so, just like when humans wrote the code by hand, some will take time to carefully build systems to validate their work and others will do anything to get the job done as quickly and lazily as possible. It is all but certain that some people will have AIs producing good, reliable code, with others producing mountains of garbage.
In the olden days some never learned, but those that did tried things, failed, and then build systems to prevent making the same mistakes again.
I don't see what's different today. Some will never learn, but of those who do they'll try things, fail, and then build systems to prevent making those same mistakes again.
MSFT stated that they went all in on AI and had a key metric of generating more than a million LOC per developer at MSFT.
Now they have an outage that should honestly result in people going to prison, at some point we have to question leadership's ability to... well lead because unless something drastically changes I don't think society wants to deal with a massive transnational trillion+ corporation with various monopolies that controls many aspects of our lives if they're going to continue to give us shit service.
I'm probably going to get a lot of flak here, but if the damages are that all emails have to wait for a day twice in a year, then that's quite a good product.
While we can engineer a system with better availability, seems the market is quite ok with one or two nines for this.
So the trick would be not to make anything truly critical depend on such a service. And no, payroll is not enough, I'm talking OT.
I'm going to double down on this. You can live 24 hours without logging into company SharePoint or without checking a business dashboard. Truly important systems should have a break glass account for these reasons.
> how indebted the Global North has become to Microsoft and its enshittified services
I wish it’d be only global north boomers but India and much of South East Asia also runs on freaking M$ and not unlikely that excel is big in China too.
That's kind of a partial misstatement. There's a huge problem in some parts of the stack, and there's massive (entirely legal) inducements to reattach to the Redmond teat.
There's only warm fuzzies for helping fix or find OSS but the drive for resilient systems hasn't died.
Other bits of the stack are fine. It's not a flat world.
Germany and Sweden are not China. They have extremely federalized governments. They are also part of the US-based economic system with a lot of neoliberal principles. Their voters pay a lot of tax and they do not really want to pay the price of going fully independent. Microsoft as many US companies got an edge in the 90s due to DoD contracts and corporate contracts. There are many European companies whose systems are deeply tied into Microsoft (one example, the whole restricted info labeling system of Siemens was deeply integrated with Office 365, SharePoint and Exchange when I worked there)
Basically migrating away from Microsoft has always a significant cost even for partial replacements. The democratic systems and shareholder pressure makes it hard to keep long-term projects going in the Western countries and companies. Choosing Microsoft has never costed an election nor caused significant share price but going against it has economic and legal risks.
So without a dictating government with nuclear reasons, you will not get enough pressure to move away.
I do hope for a future where LLMs are so good at patching zero-days that the NSA and all other nation-state 'cyberwarfare' departments (ie software privateers) lose most their capability.
On the other hand, nation-states will be able to run massive GPU clusters full of LLMs dedicated to cyberwarfare. They certainly have the resources to do it.
Imagine the Huggingface incident, but with an agent swarm that's deliberately malicious instead of just misaligned.
Cyberwarfare isn't generally about hacking or stealing data. It's more like a footrace to a chessboard.
Most of the analysis and summary that demands LLMs is public data... massive amounts of data. Otherwise, it wouldn't be "warfare". It would just be spying.
yes, that's why US is so keen on preventing China getting the gpus.
And Israel is so bent on all AI. Trumendous advantages over the countries like Iran who will have to basically become hermit kingdoms to protect themselves.
> The NSA was aware of this vulnerability but did not disclose it to Microsoft for several years, as it intended to use the exploit
This is why agents discovering vulns en masse is pretty great in the long run, it gets ahead of these sort of malicious hoarding shitheads with more resources than ethics.
Unless those who have access to the agents discovering the advance vulns are few, and those who have access to the agents writing vulnerable code are many, right?
That's the same situation as before AI: many developers don't have time/resources/expertise to prevent the advanced exploits, and few very advanced players do.
Feels like we are entering the fall of this crop of mega corps. Between Apple’s absolutely garbage software, Microsoft’s complete incompetence, and Meta’s endless lawsuits… finally time for a little creative destruction?
When your necessary compute is controlled by rentier capitalists, and you’ve had the option to switch to compute not controlled by rentier capitalists for years, who do you really have to blame but yourself?
I’ve been enjoying a piping hot cup of “I told you so” every morning this week, it feels like.
“Microsoft is too big to go down, why do we need a redundancy?” Another round, barkeep.
“Amazon won’t ever have downtime, they’re huge! Our workloads are much safer there.” Make that two, what with MEC1-AZ2 being a smoking hole.
“Why would we need backups when these providers offer us a five-nines SLA?” Sorry, make that three, since Google’s EU-WEST-2-B AZ lost power and its redundancies failed to start.
Now don’t get me wrong, nothing is perfect or bulletproof. That being said, the more eggs you put into a single basket, the more of them get damaged if something happens to said basket.
Everyone lost the plot on backups, redundancy, and diversification in favor of single, “do it all” providers and hyperscalers.
I hope any Microsoft infra greyboards who were canned during these recent rounds of layoffs are feeling deservedly content as they watch what's left of Microsoft hoisting itself by its own petard.
It's clear from the wording of the updates that Microsoft took the better part of a day to even figure out what was happening, and that they're throwing everything at the wall to see what sticks, instead of following the kind of precise recovery plan that they would have had, had they understood their own systems.
Oh, I don’t expect anything from Amazon after their first such wartime incident.
That being said, I do distinctly remember being called an alarmist for warning that if a plurality of compute is in one physical area, and said compute serves a large amount of customers, then that space is a gigantic soft target.
“Nobody is going to attack a datacenter, stego-tech, you’re being completely unreasonable.”
"It seemed like a good idea at the time", especially when the people holding the purse strings think it's cheaper that way. So we drive without a safety belt because a safety belt would cost money.
In the age of AI there's ZERO reason for "the global North" to continue using Microsoft. Wean your finance guys off of Excel, somehow. Replace the CTO with somebody that can hire IT that doesn't depend on the crutches of Microsoft.
If people's MS subscription revenue were funneled into on-site devs and devops for open source solutions, we'd all live in a better world. Sure, the money wouldn't be funneled to a single ledger in Redmond, but the overall productivity would be higher and there'd be greater GDP growth in other areas.
I've heard of so many small companies with a slightly technical bent replacing their SalesForce with better CRMs that make everyone happier, thanks to AI. MS is next.
Let's make sure that dollars get funneled away from economic rent and into more productive efforts. Schumpeterian rents might exist for a bit, but they should never exist long. Microsoft's rents have lasted for too long to be economically optimal.
All roads lead to Excel. When you try to make a nice dashboard in Jira that incorporated custom fields and all the data was normalized, nope - they send out an excel spreadsheet every time. They do this to any tool you try to use as a group tool to update things - like Confluence. Before you know it it's back in Excel again and you have multiple sources that keep getting lost.
If data is structured enough to be put in a dashboard, it is structured enough to drive automated tools that make the dashboard redundant.
Excel is the modern equivalent of the back of a napkin. It is inescapable.
> can I get these results in a CSV format
And then they import to excel.
At least when I get a CSV I can open it in Calc.
I like CSV as a format!
I think the criticism this post makes applies regardless of how you get your spreadsheets. It would apply equally to Google Workspace.
Plus, I guess there are office 365 outages, but the outage I personally hear more chatter about is the hosted exchange services (outlook), you will still need an email server in a post ai world, right?
Sure you'll need an email server. It just doesn't have to be Exchange.
I wouldn't run Exchange either.
But firstly, it's a legitimate choice for normies who just want to run a business, and second, as I said, the criticisms this article brings would be applicable to any other big tech provider who is laying off staff and leaning heavily into LLMs.
It's not possible to just run your own email server anymore. You can try, but nobody will interact with it, no matter how long or how hard you work to build your reputation. Fastmail still goes to spam, and they're very large and dedicated. It's hopeless. You use one of the big five or so providers, or you're simply done.
I don't think this is entirely true, i have my own self hosted email server and it does alright.
What you're describing is the whole reason we need to ditch them, not reasons to use them.
The problem is never Excel, its user and permissions management. Apart from Google Suite there is not really any other one stop shop for all of that, and definitly nothing open source (for now)...
The idea that small companies would have fewer problems if they had to debug vibecoded software is badly misguided.
I'm advocating for spending money on software devs to open source spreadsheets instead of subscriptions.
Office 365 / Azure AD subscription is cheap. It comes with legal certifications that are very expensive to obtain alone and usually shareholders pressure you to get them. It has remote management capabilities and a very alive closed source plugin ecosystem that's battle tested.
How long would it take FOSS to replace this? How long would businesses and their shareholders will wait? Buying Microsoft will GUARANTEE that you'll have the certification you need in weeks, how long would FOSS would ship and PROVE that they obey a non-open "open" standard?
Will FOSS GUARANTEE cross-compatibility of the file format?
What you are proposing isn't stupid if the most important goal is independence. Most businesses don't care about independence. They PREFER dependencies they can blame and sue / settle with.
Yes, but it's easy to replace Word and Excel with OpenOffice.
What’s your Excel replacement solution?
Excel was always an antipattern for me, I work with biological data, which Excel until recently couldn't even import without data corruption [1]
If it's necessary to recreate all the keyboard shortcuts for Excel like Emacs has done with evil-mode for vim, maybe that works, I don't know.
Hell the finance guys can almost make their own replacement on their own today, with AI.
[1] https://www.science.org/content/article/one-five-genetics-pa...
I think you are massively underestimating the engineering effort of replacing Excel in large organizations. I don’t think the finance guys can reliably vibe-code an alternative to a massive product that’s at the core of their operations.
Handling simultaneous edits, guaranteeing correctness, handling massive data sets… as amazing as AI is, this is a huge can of worms.
You don't need to vibe code an alternative to excel, you need to vibe code a solution to the problem you're solving with an excel sheet.
Any nontrivial excel sheet should be a database backed application, the finance guys who build them are in every sense programmers they're just using the wrong tool by convention.
An agent, sqite, and some html. Are what they want.
You're missing the point that the users don't want an application. They want to go off on flights of fancy altering the "solution" as they see fit. Excel lets them do whatever crazy thing they want without constraints. Having a database-backed application is exactly what they don't want because they can't be "creative" with it.
I agree with you, BTW, but I've never seen moving Excel sheets to an application work in practice. You can do it, and it just creates more Excel sheets.
This is in fact several major things:
* Redo UI patterns (for old users): Only matter for on-boarding and compatibility, and probably need to redo the anti-patterns, not necessarily a good idea, or at least must be hold before the next:
* Define a much better internal model than excel has (I bet a mixture of relational, array model as core, then a reactive layer on top), including proper support for correct datatypes (not localized!), dates in all shapes, currency, commodities, unit of measure, etc, tables default as fixed type per column instead of free-form, then we have proper "variant" values, use algebraic types, and so on...
* Then redo the UI
* Define how import/export correctly (so not more non-sense about import based on localized dates, currencies...)
* ... and from legacy formats like Excel
P.D: I explore the idea on https://tablam.org and my goal is eventually marry what excel+access do in one app
> guaranteeing correctness
So long as you don't have things which aren't dates but which Excel insists totally are. (Meme: https://knowyourmeme.com/photos/2429759-involuntary-celibacy...)
But yes, cans of worms are many, and Tardis-like they are larger inside than out.
If organizations can't move off Excel, and outages like this continue to be a problem, then competitors that don't use Excel stand to gain an advantage.
Let the dinosaurs go extinct!
The majority of finance guys can't even vibe code a markdown editor, let's be honest here.
Yeah, we wont be one-shotting stuff like this for a least another year or maybe even two! (This is a risky prediction, but I made it.)
Nobody needs to vibe-code OpenOffice Calc, it has existed for decades and works well.
Actually importing gene names was possible without having autoconverted names, you had to make sure you were importing files using an import tool, currently listed under the "Data" tab and then importing data from file/web or data, or use the button specific for .csv/txt. In this way, Excel would open gene data in a table editor without renaming gene names if you made sure, this has been possible since at least 2005 I believe, or maybe even earlier. It was critical to not just auto-click or open a .csv or .txt file with Excel, but specifically open a blank or existing workbook and then use the import tool from inside the application. Opening data table files directly with Excel was/is typically bad practice.
Excel is still the tool in Finance. It is the database, the visualization, the calculation enginer, the collaboration tool, and many other things all rolled into one.
So your suggestion is to make an Excel clone with AI?
What other option is there?
Edit: actually I think there's got to be a far better and completely different UI that's possible now with AI. There has to be something better. What is it? I don't know. Some clever specialist will invent it.
Python.
Just throw the raw data at the LLM and tell it to do the analysis directly on that, and there's a good chance it will use python and some data science libraries for you.
And if you want, it'll turn it into a web page where people can drag and drop more raw data files on to get updated analytics.
But then you’re still just using a vibe coded Excel. Maybe you save a little bit on subscription costs, but that’s probably swamped by other costs.
What about libreoffice calc?
That is the only legit solution I’ve heard so far. What is your experience with it? A worthy replacement?
No. Excel really doesn't have an equivalent replacement. Other parts of the office suite do although if you're a lawyer (revision marking) or have other specific integration needs, you probably need Word.
If you're just doing lists with maybe some simple calcs (which is what most people use spreadsheets for most of the time), pretty much anything works.
May I ask for a few examples of things you do in Excel that don't have a replacement?
The vast majority of people I have seen using Excel don't know how to write a formula. Now they just started copy-pasting whatever ChatGPT was telling them to copy-paste, but well... and anyway all the alternatives do support formulas of course.
There are plenty of functional replacements. However there are zero replacements that emulate the same keystrokes and keyboard shortcuts and all other aspects of hard-won muscle memory like function names and quirks of behavior.
When companies don't want their employees to work remotely and force them to come to the office because they think it's more profitable, it's not like they give a shit what the employees think.
Why would they give a shit about the little comfort of the employees who don't want to learn new shortcuts?
It's been years since I've really tried to exercise Excel (and haven't used it--or Libre Office--in a long time). Pivot tables, a number of the other data-related functions... Just a bunch of things I'm not going to spend a couple of days digging into. Google Workspace works fine for my needs.
Pivot tables have been in Calc for years. Do you have any current, specific examples of things you need from Calc that it can't do?
But then it really sounds like "yeah, Excel does have alternatives"?
For my use case, mainly science stuff, works like a charm. I find Calc better because I could never get used to the ribbon, and it's also way faster. I actually did my whole thesis in LibreOffice, including all text, references and figures. Afterwards, during the PhD, it was replaced mostly by Python and R, but at that point I was really hitting some resource limits with huge CSVs.
The main issue is that excel has some proprietary features that make it hard to migrate off.
Otherwise calc has everything for most people. Macros work different a little.
I sometimes use it for personal use and it seems all right, but I'm not a power user.
I've been finding myself using Google Sheets more, which has the potential for the same problems in this article.
It might be a worthy replacement for Excel 95, but not really for anything modern.
It works with csv which is totally broken in excel
maybe this could be used as a base and people could use AI to create extra features they need on top of it if there are things missing that Excel has that they want
This. Judging from replies there are a lot of folks here who imagine Excel usage being something on the order of basic readouts with some SUM()'s and MEAN()'s.
The real grip comes from entire business processes implemented across dozens of worksheets with incredibly complicated cross linked formulae relying on Excel edge case behavior. As soon as LibreOffice et al display the most minute differences they're done and will never be mentioned again.
Peyton-Jones says excel is demonstrably FP so UI aside (which is SAMOP) Haskell seems sensible especially for the finance team.
FP and SAMOP mean what? Your suggestion is to use a programming language to replace excel?
FP means functional programming and SAMOP means "simply a matter of programming" and the humour appears to be a Whoosh moment. Cell based spreadsheet programming is good, and should be portable. But people code to Microsoft bugs.
Sorry so many people making unrealistic solutions, I missed that this was humor.
This would be an unmitigated disaster - most programmers have trouble with Haskell, I can’t imagine saying hey finance guys here’s an LLM and Haskell figure it out when they’re used to using a full on application with UX. It’s actually so crazy I feel like you’re joking.
UI is the entire point for an excel alternative, as I understand it.
The computation side is trivial in comparison. But with AI, emulating the UI becomes far easier.
That's like saying, "UI aside, a car is just wheels and an engine, so drivers should break out the welders instead of relying on big auto"
The serious answer here is that a lot of MS products have kindof dual purpose: they are 90% used for very simple stuff (collecting data into a spreadsheet, having a text editor that allows basic styling), and 10% stuff that power users rely on (huge excel files with complicated formulas for data transformation and visualization, word / ppt templates with byzantine branding). The answer is that for the simple stuff you rely on open / cheap alternatives, and for the complex stuff (eg visualizing and transforming data for non-programmers, generating presentations and branded text documents) you can just ask ai now.
For complex stuff you cannot just always ask ai now.
I have mixed feelings about ai, but I think that word and powerpoint, as tools, are undeniably awful.
> you rely on open / cheap alternatives
Big business kind of wants 24/4 support that those will not be providing at all though? Plus the various nonsense certifications. I mean try to convince anyone in corporate to use Libre and they'll probably just laugh.
(not that they would ever use that support, they just want to have someone to scapegoat lol)
So your suggestion is a text editor? And ask AI for anything complex?
Row Zero (now part of Databricks) is by far the best spreadsheet on the market - it's not even close.
Row Zero's fully Excel compatible, including keyboard shortcuts, long-tail pivot table configurability, 500+ built-in formulas like LAMBDA and LET, and real-time collab. But it scales 1000x beyond Excel (2 billion row limit vs 1 million for Excel), has built in connectors to everything (Databricks, Postgres, S3, Snowflake, etc), native Python support, deep AI integration with Databricks Genie, and all the security and data governance features to make big banks and healthcare providers comfortable. It's built by the team who brought Rust into AWS and rewrote S3's file system circa 2018.
We're used by individuals and small businesses all the way up to huge deployments in the Fortune 100.
Source: Was founder and CEO, now leading spreadsheet development at Databricks
> fully Excel compatible
But is it 1-2-3 compatible?
R and Python are both obvious replacements for data processing, and having the data itself appear in read-write columns of an Excel file is and has always been ludicrous.
Excel was for a time when programming meant more arcane languages. In 2026, there is no excuse for a business major not to know Python or R. They're both easy.
They are not nearly as easy as Excel, even for people who know those other languages well.
I don't think you understand what Excel is and what it does.
Do tell!
So your suggestion is replace excel with python and R? So when I send someone a spreadsheet instead I send them… a python script?
I feel some of these people posting have not thought too hard and/or don’t understand the actual requirements here.
Also Excel is probably the best thing MS produces. There is lower hanging fruit elsewhere
Excel is the best product I would never want to work with.
Unfortunately I have to.
I'm not an Excel expert so I can't judge but https://getartcraft.com/apps/gridcraft is in heavy development at the moment
Why are you recommending unfinished software that you don’t even use… that’s not a legitimate solution.
I'm already using the Word one and I feel like it's better than the original so I'm guessing Excel is at a similar level of quality.
These products all have super deep feature sets. This might meet basic editing needs. There is zero chance they have replicated the full feature set. Word for example has a ton of functionality for lawyers.
I can tell you all of the Excel alternatives are extremely disappointing.
I guess you only need to get a list of the feature sets and then have an AI implement them (and follow up to test them and such)
I do agree with that but I also think ultimately that a copy with all the most basic features and the familiar UI is good enough for most users, certainly for me
Ok fair. Anyone here have experience with the excel clone?
What do you mean? This has been in development since October 2026. That's a whole month. In the age of AI that's pretty solid.
And the 3k commits in a month, all good signs
/s
Other cool thing they can run as headless mcp servers
> In the age of AI there's ZERO reason for "the global North" to continue using Microsoft.
They own 27% of OpenAI. I don't think that having all of them swap to ChatGPT is necessarily going to have the effect you're hoping.
You write in the imperative, as if advising a leader.
However, leaders are advised by their underlings, who are swayed by their technical officers, who are charmed by the commercial arm of corporations.
The only effective path to actual change is to charm the technical officers with a product that is strictly superior in the metrics they care about. They'll then work their way up the chain.
> You write in the imperative, as if advising a leader.
We should stop referring to management at companies as "leaders". They don't lead anything. They perform administrative tasks so that their employees can actually do the real work.
> The only effective path to actual change is to charm the technical officers with a product that is strictly superior in the metrics they care about.
Probably not gonna work: those "technical" officers are mouse-click engineers. IOW, they don't know much detail, but they sure as hell know where to click!
With respect: you're a biologist, and your experience with backoffice software requirements is probably the same as my experience with biochemistry: near total ignorance.
There are exabytes of documents in Microsoft formats that can be at best "read" with a significant level of data corruption in other products (formatting errors). Writing them back out with shredding them is hit & miss at best. Entire MS products have no equivalent at all that can open their file formats, such as MS Access and Power BI.
Word isn't just paragraphs, italic, and bold.
Excel has features you probably have never even heard of, but people use at scale. Things like tabular data connections using OData. It's the front-end for SQL Server Analysis Services. It has the Mashup columnar engine from Power BI built in to it. Etc, etc, etc.
These things are massive products, with deep integrations across each other, third-party software, third-party plugins, cloud products, and on and on.
There's easily several million man-years of effort sunk into the Microsoft Office ecosystem.
Switching costs are not a matter of failing to "somehow convince" people to switch.
The switching costs are on the order of trillions of dollars, not a decision to be taken lightly.
Here's some critical points everybody normally ignores in these discussion:
1) Partial switching doesn't help, because now you have two problems!
2) Everyone else has to switch too. Google Workspace people were laughing during the Outlook outage, eeeeexcept... they couldn't send or receive emails either -- because their business partners, vendors, and customers likely used Outlook.
3) The alternative with the same features, same level of integration, same commercial support, etc... is the same thing, with the same problems, but just a different brand.
You can't escape the reality of this, it is what it is, there's no nirvana that we could all achieve if we "just" did something "simple".
> Excel has features you probably have never even heard of, but people use at scale.
Yeah I've watched Joel Spolsky's "you suck at excel" tutorial, but I've also created financial models for CFOs and the finance team using Excel instead of doing an RShint App because I wanted them to be able to understand and extend the model.
It's software. It can be recreated, more easily than ever.
I'm saying people should put their MS license fees towards in house devs working on open source projects to achieve the goal, rather than just let MS Hoover it up to waste on Copilot interfaces that nobody ever uses. Let the local devs squander that money, after they have reimplemented all the Excel keyboard shortcuts in OpenOffice and added abhorrent plotting themes that recapitulate 30 years of incompetent Microsoft practices.
Idealistic? Sure. But why not try.
Curious:
- what do "financial models for CEOs" contain apart from long rows and some numeric columns on which elementary arithmetic was applied on?
One example would modeling the costs and revenue for a business unit, the prices from suppliers, the fixed costs of capital, the number and type of employees, the throughput from sales, etc.
One of Excel's many powers is letting people play with numbers in cells and see the results of long chains of computations. Or create an array of potential values and see the results across a wide variety of inputs. Or even use the builtin solver to find the value that minimizes or maximizes a desired output...
And thats before starting on the pivot tables!!!
More likely than Excel getting replaced is that agents and users collaborate through Excel.
> In the age of AI there's ZERO reason for "the global North" to continue using Microsoft. Wean your finance guys off of Excel, somehow.
What does AI have to do with it? Are you proposing that we replace Excel for accounting purposes with something vibecoded? That sounds a bit nuts to me. You can pay a bit of money for mature software that will probably not fail to add 2 + 2 correctly and will not run away with your wife, so why give that up?
I actually suspect that companies such as Microsoft pushing for AI everywhere may undermine that proposition... but then, the alternative is probably LibreOffice Calc, which is eminently boring and doesn't have Copilot popping up.
After 30+ years of hearing this type of twaddle, it's more likely for Microsoft to keel over because of its own leadership incompetence than to be so much as dented by any of these half-baked community efforts.
One can only hope everything you've laid out comes to fruition.
Depends how you do it. You could easily replace MS rent with OpenAnthroogle rent instead.
I'm stealing that neologism and claiming it as my own.
> OpenAnthroogle
Great term!! :)
The finance team would give up cocaine before MS Excel. And they would give up water before cocaine. Now you know how important Excel is.
Oh please, stop it everyone with this thrite trope of the fabled finance guys with their crazy spreadsheet that holds the whole company together. 99% of Excel use is NOT that!
What MS365 does best (for now) is being a one stop shop for user/permissions/device management and coming with outlook/exchange/Teams - nobody with purchasing power gives a shit if ppl get Excel or LibreCalc as spreadsheet, they get it because it comes with the rest.
Office 365 is the cheapest solution. Its five subscriptions (email, office software, storage, chat/calls) bundled into one 11 euro/month deal.
Unless you spin up an entire devops team plus an ML team to run inference at scale in-house, you're just changing where you funnel your subscription money.
Also: No home-grown CRM will have, in the long run, uptimes anywhere approaching e.g. Office365. You're just smearing one large outage into a lot more small ones, plus a few data losses.
That rent buys you something. The internal development buys you something very different. You might want to make a reasoned call on what you're buying instead of making GDP predictions.
The ONLY WAY to get rid of MS Excel is to make something better and then literally force feed it to the finance folks who will go kicking and screaming all the way.
That is a more complete description of "wean" which I agree with.
> Now, are Microsoft going to offer any compensation for either lost business or additional accrued costs? I highly doubt it, even though it is literally their fault and they do still have the cash reserves, despite increasingly wasting it on loss-making AI adventures.
Aren't they? I know that they, and other big techs, have clauses in their terms that say if they don't meet service requirements they will offer refunds/credits. I've never been in a position where I was the one requesting the refund but I always presumed they actually did it.
I feel this is missing the role we play. Choosing free, flashy, new and jumping on the new best thing in droves.
Unfortunately the most effective way to learn about what is important is to have it taken away. So I don't believe this is going to change ... ever.
Some of us software engineers have been predicting the software apocalypse since over a decade so what's going on now is not surprising.
What else won't surprise me is the government trying to step in and making it their problem. That's what I'm worried about. The only right thing to do is remove regulations, remove government intervention from every aspect of the economy, and let the market fix the mess which the central planners have created.
Designing the entire software industry as a giant government-supported oligopoly was not a good idea. It was unethical and it was not going to work in the long term.
There are older predictions than that...
"If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization." - Gerald Weinberg, 1970's
> let the market fix the mess which the central planners have created
Isn't a Microsoft dependency failure market-driven? Compared to the 1960's and 1970's, the present-day software industry is much more driven by the market than it is by the government.
> Isn't a Microsoft dependency failure market-driven?
Not only that, but perhaps if anti-trust laws were actually enforced, we might actually have competitive alternatives.
Not only they are concentrating power in the hands of an oligopoly, but they're making the whole country unsafe. If everyone uses the same datacenters, it is enough to shut down a few of these data centers to bring civilization to a halt in this country. With the current AI trends, this risk is getting ever greater.
Yes it seems very dangerous. Also, I think a lot of the software we use today, we don't need or we could make much more efficient. We almost certainly don't need so many data centres. Most businesses could self-host their own systems. Even a lot of large companies.
I feel like we probably only need a tiny fraction of the compute capacity if we let things be decentralized and let the market operate naturally.
This was my thought years ago and so now it looks blatantly obvious to me with AI data centres driving electricity prices. It's pretty clear the industry is being propped up by other parts of the economy.
The real "central planners" are the billionaires who have led the way to oligopoly since the 80s. Not sure how handing them more power is going to help distribute it.
backup plans should be discussed (both backing up the data and alternative services to use when services go down), and also switching to better alternatives to Microsoft when possible
Hot take: most people, in most industries, should be able to tolerate a day of downtime in the sense that there are still useful, productive things to do even when all the computers die.
The fact that the world relies on four 9's of availability and throws a hissy fit whenever there are outages is an antipattern.
And also, people should get a snow day every once in a while.
I've often wondered how much unplanned downtime is caused by the aversion to planned downtime.
Like if Exchange Online went down for an hour every Sunday for maintenance, and they made their network configuration change then, there would have been no cascading failure.
This is how you get enshittification. It'll never stop at a day every once in a while, unless we hold those with billions to a higher standard, instead of giving them excuses while the layoffs continue.
Or maybe, if everyone was willing to tolerate occasional downtime, they'd compensate for it by not having single-sources of failure on the things that go down, and thereby fixing the structural issues that allow enshittification (dependence that the companies running the services get to use to squeeze users)
When you say "they'd compensate" are you referring to us end users trying to do banking when MS or CloudFlare go down? Or are you saying the big corps themselves would compensate? Enshittification stems from there being no incentive for them to do so, because they answer to shareholders not customers or competition. If you actually want structural issues fixed, I'd implore you to please hold these organisations to a higher standard of quality. I don't understand how average folks tolerating downtime leads to structural issues being fixed.
It's a single Point If failure. Microsoft makes Sure that everything is running smoothly. But in the end, if 100 companies use Microsoft, 100 companies are affected if Microsoft has an outage.
Trust in others, across the board, is tanking. I expect in the not to distant future every company will have a completely proprietary stack, or they will be facing competition circling them while they can't keep their shit running. It is the mono culture that is too easily attacked.
It is entirely likely that with AI, you could just spin up the whole 'stack' as you need it without depending on external services. AI will truly drive the cost of software generation to zero.
Well, no you have it completely backwards. The scenario where every large institution generates their own software stack using AI is one where the sale price of software is zero (there is no demand) but there is a cost (it's whatever the AI companies want to charge, with almost infinite leverage since this organization can't do it alone)
Nope, it's you that is completely wrong. I can already run Qwen-3.8-27B on my 2021 64GB M1 Max Macbook Pro at a totally acceptable tok/s. The hardest part was waiting for the download to complete. It's free modulo cost of electricity.
I can build almost any piece of software I can dream up, the data never leaves my machine, and there's literally no way for anyone to charge me any price ever.
Yes, but can you verify that it is what you wanted it to be for free? If you're making a framework out of thin air it'll still be an undertaking to check that.
Yes? My time is a cost, yes. But otherwise, not really an issue? Does that change my answer to the previous question? Not at all?
I'm struggling to understand the point of your comment. Why would that affect whether or not I am over the pricing barrel with a frontier lab? Like is it easier to verify Opus code than Qwen code? The code is the code.
That would hold only if they are creating their stack over and over again. For the vast majority of what people want to do, make their purpose bound app once and go to work. Advancing tech for the sake of a new font loading do-dad or whatever is pointless. Look at all this crap we made over the last 30 years, and here we are back to preferring text.
Programmers are "program reviewers" now, more and more.
But the tools we have to do this suck. We don't have much beyond "here's the source code as text, look at it, and think really hard".
Another comment[0] suggested having AI generate "code walks" through the code. This seems better than our current approach of "my PR contains a diff, good luck".
I'm working on a tool in Emacs that allows me to mark certain lines, or entire projects, as needing review, and then helps me track which parts have been reviewed. I need something to help me find parts of the code I have never reviewed before.
[0]: https://news.ycombinator.com/item?id=50410493
As PRs get beefier I find myself using the LLM to ask pointed questions and do general surveys of the diff to flag architectural violations, smells, etc. before I read through the entire thing by hand. If you put all of your architectural and business decisions in the repo via spec-driven tools or otherwise then this becomes quite useful. It also helps if you have a vision in your head of what the system is supposed to look like structurally so you can ask the LLM specifically how the PR is handling things that you'd be on the lookout for anyway.
I've been thinking about a workflow where PRs just get automerged into a staging environment of some kind, and then tooling helps ensure that all code, ALL the code in the entire project, is regularly reviewed together.
When lines change, they are put up for review. When lines haven't been reviewed for awhile, they are put up for review. Lines are ranked based upon the complexity of the context they are in, and complex areas of the code are reviewed more often.
Basically, stop focusing on individual PRs and instead focus on continuous review of the entire system with tooling that helps identify those parts of the code that are most in need of review.
Reviewing PRs is wasteful because the optimal way to review a system does not match the way a system naturally grows through individual commits.
> Programmers are "program reviewers" now, more and more.
Our job, much like a locomotive operator, is to make sure the agents don't go off the rails. We call ourselves software engineers in recognition of that.
It’s interesting I have been 0% affected by this big outage at home or work
Big tech isn't here to let anyone down, it is your company executives and administrators who are here to let you down.
I have never understood why companies chose to standardize on Microsoft everything then switch the businesses to Microsoft Cloud everything in the belief that it was cheaper and better, simply because they don't want to pay money for onsite administrators or local IT guys for their IT. I mean if you want to use Microsoft Office then by regular licenses and run them on premises, but everything in the cloud who thought of that?
Many employees also don't know that their bosses or bosses bosses were also given kickbacks to switch their IT to Microsoft/Google/Oracle centralized services. For some obscure reason company bosses and government institutions still switch to Oracle for their services which don't work, wind up with Oracle negotiating with them for money for services which don't work, then they pay termination fees to Oracle if they bail out early.
Another factor I believe to be at work here is large investment firms like Blackrock, State Street etc pressuring the executives of companies which they are major shareholders in to switch their IT to the IT companies they are major shareholders in, therefore boosting the values of their stocks. For me it is the main reason for this centralization of IT services into Microsoft, Amazon and Google.
As someone who started using AWS from the time it went beta, I could see the value in it, but I couldn't see why companies had to switch their systems wholesale to it. I could see the benefits for new startups which had to move quickly and early and perhaps have a need for unexpected traffic, but even after some time the stuff would have to move back onsite once things settled down.
Other the end of the day the situation is entirely down to greed and corruption masking as incompetence.
A week or 2 ago in the UK virtually all the major banks went down due to some problems at Google, Microsoft or one of the other companies. So many banks had their IT hosted with them so everything went down simultaneously.
It is time the governments passed legislation forcing these banks not to centralize their IT with these companies and ensure they have their own independent redundant systems. Maybe they should go back to IBM mainframes or something.
> I have never understood why companies chose to standardize on Microsoft
Exceedingly simple: those who decide don't know better.
Blaming every downtime on "vibecoding, probably" is so weird and lazy. As if these places never had downtime previously
Lazy maybe, but not weird at all. Probably true most of the time.
You have no evidence for that. You just wish it was AI's fault because you hate AI and hope it comes back to bite them.
Surely if you use LLMs to write code the evidence that they cause problems is all around you every day? Don’t you see it in every PR?
> Don’t you see it in every PR?
PRs without problems don't really grant any credit against those with problems. I see it it a fraction of PRs. Seeing it sometimes is enough to establish evidence.
Indeed, given that there was a post this spring by ex-Microsoft employee on the poor state of Azure.
"The team had reached a point where it was too risky to make any code refactoring or engineering improvements. I submitted several bug fixes and refactoring, notably using smart pointers, but they were rejected for fear of breaking something."
https://news.ycombinator.com/item?id=47616242
Sure, those takes are mostly lazy (some of my comments here included TBH). Although I have very weird feeling, that current zeitgeist is about "coding is solved", etc and the same time we have evidence that didn't help much regarding boring thing such as reliability
Trying to control the narrative this way by preemptively talking about vibecoding is so weird and lazy.
> You have to wait for Microsoft to find a competent employee who wasn't let go to try and fix what was broken by some inept leftover who was deluded enough to think that Copilot can do everything.
From the submitted blog post.
Right but according to some here AI is creating vast amounts of high quality code that should have given us 100% uptime in every service
Right?
As opposed to a pile of unmaintainable shit
AI does what you ask of it, so, just like when humans wrote the code by hand, some will take time to carefully build systems to validate their work and others will do anything to get the job done as quickly and lazily as possible. It is all but certain that some people will have AIs producing good, reliable code, with others producing mountains of garbage.
And how are people (specially juniors) going to learn what's good code and what's bad without ever writing a line of code
In the olden days some never learned, but those that did tried things, failed, and then build systems to prevent making the same mistakes again.
I don't see what's different today. Some will never learn, but of those who do they'll try things, fail, and then build systems to prevent making those same mistakes again.
AI thrives within those systems.
MSFT stated that they went all in on AI and had a key metric of generating more than a million LOC per developer at MSFT.
Now they have an outage that should honestly result in people going to prison, at some point we have to question leadership's ability to... well lead because unless something drastically changes I don't think society wants to deal with a massive transnational trillion+ corporation with various monopolies that controls many aspects of our lives if they're going to continue to give us shit service.
I'm blaming the enshitification of absolutely everything. AI or not, it's all just become bloated piles of code.
Using AI is lazy. Lazy creators get lazy responses.
https://omarchy.org is the Windows killer. https://omasheet.org the Excel poison.
I'm probably going to get a lot of flak here, but if the damages are that all emails have to wait for a day twice in a year, then that's quite a good product.
While we can engineer a system with better availability, seems the market is quite ok with one or two nines for this.
So the trick would be not to make anything truly critical depend on such a service. And no, payroll is not enough, I'm talking OT.
In the current world, a lot of things like 2FA depend upon emails. So emails not working means not being able to sign into systems.
I'm going to double down on this. You can live 24 hours without logging into company SharePoint or without checking a business dashboard. Truly important systems should have a break glass account for these reasons.
> how indebted the Global North has become to Microsoft and its enshittified services
I wish it’d be only global north boomers but India and much of South East Asia also runs on freaking M$ and not unlikely that excel is big in China too.
Many EU countries started moving away.. and then failed to do so and doubled down on Microsoft again.
That's kind of a partial misstatement. There's a huge problem in some parts of the stack, and there's massive (entirely legal) inducements to reattach to the Redmond teat.
There's only warm fuzzies for helping fix or find OSS but the drive for resilient systems hasn't died.
Other bits of the stack are fine. It's not a flat world.
That's a good way to say nothing. I did not understand your comment.
I know Germany wanted to migrate to Linux but decided to stick to Windows after all.
Same in Sweden, they tried moving away, but then gave up.
Both were headlines a while ago.
Germany and Sweden are not China. They have extremely federalized governments. They are also part of the US-based economic system with a lot of neoliberal principles. Their voters pay a lot of tax and they do not really want to pay the price of going fully independent. Microsoft as many US companies got an edge in the 90s due to DoD contracts and corporate contracts. There are many European companies whose systems are deeply tied into Microsoft (one example, the whole restricted info labeling system of Siemens was deeply integrated with Office 365, SharePoint and Exchange when I worked there)
Basically migrating away from Microsoft has always a significant cost even for partial replacements. The democratic systems and shareholder pressure makes it hard to keep long-term projects going in the Western countries and companies. Choosing Microsoft has never costed an election nor caused significant share price but going against it has economic and legal risks.
So without a dictating government with nuclear reasons, you will not get enough pressure to move away.
> Now, are Microsoft going to offer any compensation for either lost business or additional accrued costs?
The NSA lost control of EternalBlue which resulted in billions of dollars of economic damage globally:
https://en.wikipedia.org/wiki/EternalBlue
They didn't offer any compensation either.
I do hope for a future where LLMs are so good at patching zero-days that the NSA and all other nation-state 'cyberwarfare' departments (ie software privateers) lose most their capability.
Zero-days are the least of our worries in the late 2020s and early 2030s
On the other hand, nation-states will be able to run massive GPU clusters full of LLMs dedicated to cyberwarfare. They certainly have the resources to do it.
Imagine the Huggingface incident, but with an agent swarm that's deliberately malicious instead of just misaligned.
Cyberwarfare isn't generally about hacking or stealing data. It's more like a footrace to a chessboard.
Most of the analysis and summary that demands LLMs is public data... massive amounts of data. Otherwise, it wouldn't be "warfare". It would just be spying.
yes, that's why US is so keen on preventing China getting the gpus. And Israel is so bent on all AI. Trumendous advantages over the countries like Iran who will have to basically become hermit kingdoms to protect themselves.
Seems just as likely that LMMs will have the opposite effect.
> The NSA was aware of this vulnerability but did not disclose it to Microsoft for several years, as it intended to use the exploit
This is why agents discovering vulns en masse is pretty great in the long run, it gets ahead of these sort of malicious hoarding shitheads with more resources than ethics.
Unless those who have access to the agents discovering the advance vulns are few, and those who have access to the agents writing vulnerable code are many, right?
That's the same situation as before AI: many developers don't have time/resources/expertise to prevent the advanced exploits, and few very advanced players do.
Holy cow another Sweeting! I wonder if we are related?
Edit: and you've spent time in Shanghai too! Small world!
tangentially i happen to know someone with the last name Sweet. i wonder if it diverged somewhere along the way.
Feels like we are entering the fall of this crop of mega corps. Between Apple’s absolutely garbage software, Microsoft’s complete incompetence, and Meta’s endless lawsuits… finally time for a little creative destruction?
Don't forget the reckless data center spend. To me that's the most likely candidate for their fall.
When your necessary compute is controlled by rentier capitalists, and you’ve had the option to switch to compute not controlled by rentier capitalists for years, who do you really have to blame but yourself?
I’ve been enjoying a piping hot cup of “I told you so” every morning this week, it feels like.
“Microsoft is too big to go down, why do we need a redundancy?” Another round, barkeep.
“Amazon won’t ever have downtime, they’re huge! Our workloads are much safer there.” Make that two, what with MEC1-AZ2 being a smoking hole.
“Why would we need backups when these providers offer us a five-nines SLA?” Sorry, make that three, since Google’s EU-WEST-2-B AZ lost power and its redundancies failed to start.
Now don’t get me wrong, nothing is perfect or bulletproof. That being said, the more eggs you put into a single basket, the more of them get damaged if something happens to said basket.
Everyone lost the plot on backups, redundancy, and diversification in favor of single, “do it all” providers and hyperscalers.
What could possibly go wrong?
I hope any Microsoft infra greyboards who were canned during these recent rounds of layoffs are feeling deservedly content as they watch what's left of Microsoft hoisting itself by its own petard.
It's clear from the wording of the updates that Microsoft took the better part of a day to even figure out what was happening, and that they're throwing everything at the wall to see what sticks, instead of following the kind of precise recovery plan that they would have had, had they understood their own systems.
> MEC1-AZ2 being a smoking hole.
What do you expect amz to do about the drone strikes haha.
if you absolutely needed your servers in an active war zone for some reason cant imagine you'd have five 9s either
Oh, I don’t expect anything from Amazon after their first such wartime incident.
That being said, I do distinctly remember being called an alarmist for warning that if a plurality of compute is in one physical area, and said compute serves a large amount of customers, then that space is a gigantic soft target.
“Nobody is going to attack a datacenter, stego-tech, you’re being completely unreasonable.”
Mmm, delicious schadenfreude.
"It seemed like a good idea at the time", especially when the people holding the purse strings think it's cheaper that way. So we drive without a safety belt because a safety belt would cost money.