Software engineer Jimmy Breck-McKye published an essay this month arguing that software is breaking down at a systemic level. His case is that this year’s outages and breaches come from systems built out of so many borrowed parts that nobody fully understands them, and that AI coding tools are making it worse. For a business in Southwest Florida, the useful response is the same one you already use for hurricane season: know what you depend on, expect some of it to fail, and have a plan for the day it does.
Key takeaways
- The software a small business uses is built from many other companies’ code and services. A failure several steps away can still shut down your office.
- This year alone brought a Microsoft 365 email outage across the Americas, critical PHP flaws that put websites at risk, and a supply chain attack that spread through developer tools.
- Treat software outages like storms: list what matters, plan for a day without each piece, and keep backups you control.
- As more code and automations get written with AI, having someone who understands and documents your setup matters more.
What is the argument?
Breck-McKye says modern software sits on cloud platforms and open source packages that each depend on many others, so one mistake can cascade. Fewer people understand whole systems because the work has been handed to generalists and to tools. AI adds to it: each AI-written change can look fine alone while the system as a whole drifts into something nobody can explain. He calls that decoherence, and he expects this year’s failures to become the normal pace.
What has broken this year?
- Microsoft 365, November 12. Outlook and Exchange Online were down for much of the day across the Americas, and plenty of Southwest Florida offices lost email. Our guide to that outage covers what to check and how to plan for the next one.
- PHP2Shell, October. Two critical flaws in PHP, which runs most WordPress websites, let attackers run code on web servers without a login. Emergency patches came out October 21 and attackers were using the flaws within days.
- Mini Shai-Hulud, May. Attackers broke into the build system of a popular open source project and used it to publish poisoned versions of over 170 software packages, according to the Cloud Security Alliance. The malware stole passwords and access keys from the developers who installed them.
- eMudhra, October. A certificate authority trusted by web browsers had issued eight intermediate certificates with keys weak enough to guess. Browsers pulled their trust, and the company left the public certificate business.
A local business didn’t cause any of these. Any of them could still reach one through its email, its website or its software vendors.
What does this mean for a Southwest Florida business?
A dental office in Fort Myers or a property manager in Naples depends on a chain of services it never sees: Microsoft 365 or Google Workspace, a practice or property management system, a website host, a payment processor, a phone system. Most days that chain works. The question is what happens on the day one link fails, and whether anyone at your company knows which links matter most.
There’s a local angle too. Season is starting, so you’ll have more customers, more staff and more pressure on every system. An outage in January costs more than the same outage in August.
AI shows up here as well. Staff are using AI tools to write spreadsheet formulas, scripts and automations between apps. They save time, and they also leave behind small pieces of your operation that nobody understands once the person who built them moves on.
What should you do about it?
- List what you depend on. Write down the ten services you can’t run without, the vendor, who owns each one at your company, and where the admin login is kept.
- Plan for a day without each one. Use the same thinking as your hurricane plan. Keep a staff and client contact list somewhere other than your email system, and know how you would take payments or appointments if your main system were down.
- Get critical patches in quickly. PHP2Shell went from patch to attack in under a week. Your website host or developer should be able to tell you within a day whether a critical flaw affects your site.
- Keep your own backups. Back up Microsoft 365 or Google Workspace mail and files separately, and keep a copy off site. Storm season teaches the same lesson every year: “it’s in the cloud” isn’t the same as “we can get it back.”
- Document anything built with AI. If someone made a script, macro or automation with AI help, ask for two sentences on what it does and where it runs. Have anything touching money, customer data or permissions reviewed.
- Have someone who understands your setup. In-house, outsourced or both, your environment should be documented well enough that a competent person could take it over tomorrow.
Frequently asked questions
Should we stop using cloud services because of these outages?
For most small businesses, no. Microsoft and Google run email more reliably than an in-office server, especially in a place where power and flooding are real risks. Plan for the occasional bad day and keep your own backup of the data.
Does PHP2Shell affect our website?
If your site runs on WordPress or another PHP application and PHP hasn’t been updated since October 21, 2026, it may. Ask your host or developer which PHP version the site uses and whether it is a patched release.
Is it safe for our staff to use AI to build automations?
It can be, with a few rules. Keep AI-built tools away from payments, permissions and customer data until someone who understands them has reviewed them, and keep a short note of what each one does and who owns it.
Want help mapping what your business depends on?
We go through your critical services, backups and outage plan as part of a free consultation for businesses from Tampa to Naples, and you get the findings in writing. Book a free consultation or call (813) 999-0086. Our free tools page has a few checks you can run yourself today.