Hugh Emberson

Yes. I suspect you are right.

I wrote a thread about enterprise software and AI in a reply, and I'm putting it here because I may reuse it.

IMO we're going to see more small teams armed with AI's building software inside enterprises, rather than buying.

The software they build will be about as bad as the stuff they may previously have bought, but with a bit of luck it will fit the organisation better. 10/10

Anything that changes the balance of build versus buy equation like AI is going to disrupt a lot of enterprise software vendors.

Most enterprise software, isn't tricky, its just a lot of work mostly design.

9/n

The difference was, we sold it as a toolkit, and the core of this toolkit was something that was really difficult to get right. So in our case, most of our customers chose to buy rather than build, though some of them tried build and then bought. 8/n

I've also i've been on the other side of this equation, and I've sold toolkits to enterprises. The software that I'm thinking of had a total addressable market of perhaps 400 enterprises. The company I worked for had maybe a 150 of those enterprises worldwide. 7/n

You'll need to either modify the software, hire someone to modify it, or change the way your organization works to fit the software. None of these things is easy.

And it goes on ... and on ... for a very long time ... much longer than you ever thought it would. Often it fails. 6/n

It needs to integrate with all the other bits of software that you have in your organization. If you're lucky, there will already be integrations, and those integrations may even work.
Then there will be things in the software that maybe don't work the way your organization works. 5/n

The reason is that for many enterprises there are only 10, or 100, or 1000 other companies in the world that do what they do.
So you buy a toolkit, it's sold as a solution, but it's a toolkit, and that toolkit then needs to be adapted to fit your business ... 4/n

The reason is that if you buy a piece of enterprise software from an enterprise software vendor, what you're generally buying is a toolkit. If you're lucky, that toolkit has been used by 100 or 1,000 of your competitors worldwide and they have knocked most of the bugs out. 3/n

I'm pretty sure he stole these quotes from someone.

After a long career and a lot of disappointment, he was firmly on the build side of the build versus buy equation. 2/n

Many years ago I had a business boss (knew his business very well, but nothing about computers) who used to say that "the software that runs your business is the nervous system of your business." And "why would you outsource your nervous system?" 1/n

This was for my very small company. For larger companies, this is a major project.
I suspect that a sufficiently well trained and prompted AI could do this in a day for a small company. This is the change that's comming to this industry and it won't just disrupt Salesforce consultant.

8 or 9 years ago I spent $10k+ on customizing my company's Salesforce instance to meet the specific needs of my sales and inside sales people.
I'm a programmer and I could *theoretically* have done this myself, but I had more valuable things to do. So consultants ...

The thing you are missing here is that Salesforce is not an app, it's an App Platform. Virtually every company customizes their instance of Salesforce to work the way they do, and usually this requires consultants who know how to do this.

Also Amp feels a little bit faster. I'm not sure if it is, or even how to measure that, but it feels it.

I'm a little bit over CC's "You are absolutely correct most splendid lord and master!!!", Amp is more like "OK. Let me try that."

I hear that ChatGPT has a personality feature where you can dial up or down the level of sycophancy. @anthropic.com should add something similar to Claude Code.

Amp Amp is a frontier coding agent that lets you wield the full power of leading models.

If you like Claude Code, you really should try Amp @ampcode.bsky.social. The code they produce is about the same (it should be, they more or less use the same LLM's) , but Amp's "personality" is a bit less chatty and a bit more matter of fact.

ampcode.com

Claude Code: I made you this really useful build script, just run `deploy.sh` to build everything.

Also Claude Code: The build script is having *issues*, I'm just going to build by issuing random, non-repeatable, bash commands.

Sometimes, at the end of a session, I ask the model “what have you learnt?” Then I ask it to add its learnings to the AGENTS md file.
Sometimes it’s remarkably insightful.

In my experience LLM’s can be prompted to refactor code, but it is not something that comes “naturally” to them most of the time.

So, *I think* a Orb Stack machine sort of like a docker container with persistent storage, rather than a classic VM (like VMWare). But I'm not sure. If you know different, I'd love to know.

This doesn't mean that they are all running the same *instance* of the kernel. But if you watch the memory consumption, running a new machine doesn't meanigfully increase the amount of RAM OrbStack Helper uses. I think OrbStack Helper is the process that runs the VM.

Fedora 43:
Linux fedora 6.17.8-orbstack-00308-g8f9c941121b1 #1 SMP PREEMPT Thu Nov 20 09:34:02 UTC 2025 aarch64 GNU/Linux

A container:
Linux eea4a882eb0d 6.17.8-orbstack-00308-g8f9c941121b1 #1 SMP PREEMPT Thu Nov 20 09:34:02 UTC 2025 aarch64 Linux

I poked around inside "Machines" a bit more, and they all seem to use the same version of the kernel.

This is a Ubuntu 25.06 machine:
$ uname -a
Linux ubuntu 6.17.8-orbstack-00308-g8f9c941121b1 #1 SMP PREEMPT Thu Nov 20 09:34:02 UTC 2025 aarch64 GNU/Linux

I regret reading that. Brain hurts.

It also has something called "Machines", which are like a cross between a Docker container and a virtual machine. I haven't them out yet, but they looks really interesting.

They seem to be like a VM but they all run on the same kernel. But unlike a Docker container, they have persistent state.

And unlike Docker Desktop, it installs a DNS server on your Mac that is accessible from your Mac. So all of your containers have DNS names, hat are accessible from Terminal. You don't have to mess around with your /etc/hosts file and hope your ports don't clash.

a couple of men standing next to each other on a runway with the words `` feel the need for speed '' above them . Alt: Maverick and Goose on a runway high-fiving. With the words "I feel the need for speed" over them.

I thought "who cares fast my Docker containers are, they are fast enough." But, after installing OrbStack, I'm completely bowled over. Wow this thing is fast.

OrbStack · Fast, light, simple Docker & Linux Say goodbye to slow, clunky containers and VMs. The fast, light, and easy way to run containers and Linux. Develop at lightspeed with our Docker Desktop alternative.

Wow! I just discovered OrbStack. It's amazing. For those of you who, like me, are late to the party, it's a drop-in replacement for Docker Desktop on the Mac. It's super fast.
orbstack.dev