Morning Bathrobe Rant: Software Products.

Many, particularly those who don't work in software, don't know exactly what they want.

That is why there will always be a demand for programmers.

@h14hdotcom Who wants to talk to a developer instead of pressing a Buy button and why would it be cheaper?

@unclebobmartin I think it's a valid question and one that will play out differently in different markets. If a customer finds someone they like & trust who can offer bespoke software solutions that fit their exact needs for a price at or below traditional products, why not pay them?

@unclebobmartin Consulting can’t be cheaper because when you sell a product you distribute the cost between all customers.

@unclebobmartin The more customers you build for the more features you need and the more complex your product gets. In the AI age, I am not convinced that N simple apps will forever be more expensive than one complex app with enough features to support N customers.

@dannielo @unclebobmartin People don’t know what they want and they don’t know how to make it. They just use whatever works well enough. Finding someone who will spend the time to help them figure out what they want and go back and forth with them is going to cost way more than they will want to pay.

@dannielo @unclebobmartin > someone who will spend the time to help them figure out what they want and go back and forth with them who says this has to be a human anymore?

hahahaha this is the kind of ai psychosis i want!

You're right Bob.. you can make the exact tool you want to help you get the job done. How much time did that take? How much did it cost your agent to build that? At a time when our human time is much more expensive than agent's time, that waiting for the agent to build and maintain your tool can be expensive. Custom tools may be great for some people. For others, scrivener or that tool that does 80% of what they need may be good enough and they can use their agent to automate the other 20% Its a good discussion, and I think there's many ways to solve this problem

For developers maybe.. but for regular people they still want a software product

Until the AI hallucinates and reformats the SSD.

Disney/Pixar (or somebody) is going to base an animated evil character on what they can distill from these bathrobe videos - you mark my word!

Exactly. People who are saying 'where are the AI killers apps?' don't seem to understand that they probably will never see it. The killer app is the tailor made CRM-system that is 100% aligned with the needs of a random business you don't even know exists.

"Oh? So you're just going to build whatever you want, not knowing the avalanche of BS that comes with software development and maintenance?"

_I DO_ know the implications. And I will happily deal with them. I'm not making a product for anyone else. I'm making it for me.

*Present company excluded.

Make your own tooling. What a time to be alive. Moats are GONE. Society will move closer to a perfect competition model for many markets.

Just as open source didn't fully replace Microsoft and Apple, there will be demand both for customized and for standardized software. At some point it's easier to not deal with an agent and just use "the default".

Dude, there's almost no nuance in your takes. Even people who came up with / post on acknowledge that there are some software products that are just not worth trying to rebuild from scratch with AI, because the moat is not the code.

I hope my bot doesn't find this rude.

Yes and no… for technically inclined people (just someone who can use an app and isn’t incompetent)… I think prompting standalone software will become very doable… but that’s different from building products that rely on external infrastructure, real world ops, proprietary data, network effects, or dependencies that need ongoing maintenance… you can prompt the software layer… you can’t prompt all of the surrounding system into existence... and who knows, maybe the “everything app” will come to fruition and it’ll be cheaper to just use that… the cost (effort) to prompt your software may eventually exceed the cost of downloading the thing…

"Software products may be a thing of the past" Brutal, and also probably true.

Amazing point. Wow. We could all write software. Wow. So interesting. Yeah, that's it. No one is going download a web browser or OS any more. They'll just gen them with AI. That makes sense. Genius-level!!! No need for SmartPhones either. AI will build them at home.

I have stopped tolerating tools that have sharp edges and I am just working with Codex and Rider to make tools as I want them and it is working out fantastic so far. For those that can leverage this kind of angle, consider your AI subscription more of an unlimited shareware generator.

Everything depends on what one would like to have, Codex been building an app for some 60-70 days and not ready, now cleaning code for some +3 days, yeah yours Clean Code as base for that but agents or not it's not that easy however.

thats exactly why I built my own apps, no coding experience, but expert needs

We are all crazy! I am not alone!

what about supporting the tool? at some point you won't have the time even to say to agent to maintain it and what if there will be a stack of things like that? that will be really troublesome!

Software dev is dead. Long live authoring.

This is a good response to the "where are all the new products?" people. If you sell software products, your customers are now your competitors.

you are right .. i have been building apps for myself that i need for specific purposes.. one example is i have very fix set of recipes that only i know how i like them prepared and my wife always had to ask me about it. Decided to make custom recipes app just tuned for us

That matches my vision: why pay for software when you can pay for tokens? If agents can do the job, external tools become less valuable. But developers can still provide the service and solve problems for others, just in a different way. It’s a shift, not a disaster.

For all you naysayers of Bob's Bathrobe Rants, it looks like the market has spoken! MSFT went to $0 on this news. 🤓

Cc @jgordijn bespoke software

It’s true in the vast majority of cases, but I wonder if that still applies to very complex domain problems (ERPs, avionics, healthcare, …)

I couldn't agree more - the era of personalized software products is upon us - after watching this I had this personal dungeon master assistant built for me in about an hour.

I imagine that software can go in 2 paths: – simple, low-stakes tools → built by users themselves or sold by niche players at 1/10th the price – anything where mistakes are expensive or that runs on proprietary data/know-how → worth more than ever

This is why I started because at the end the builder is still required

RIP SaaS. I love my own CRM for the same reasons.

I am afraid there is a bigger problem that is coming. Products are obsolete, but users could become obsolete as well: if AI can write the book, why do you need software to write a book?

I still think there's a big barrier here. These AI models cost $100/mo each. Anything below that isn't enough to make good software. In my case, running local models is unfeasible, because of hardware requirements and costs. Are we gating development only for those that can pay?

Software products won’t be things of the past. Rigid software products will be the things of the past. Companies will continue to build software products , but they will build mutable , 10x extensible products in the future- that you can configure, tweak , change as you want.

I built about 20-30 Chrome extensions for myself with AI over the last 2 years, including many other developers would sell on a subscription basis. So yes, I would never pay for an extension that I can build in 10-30 minutes myself. I still see a market for sufficiently complex software, the kind that a solo developer might work on for 12 or more months. For example, I don't think I would want to build apps like Proxyman or Blender. Btw. I saw this awesome first peek at a vibe-coded writing app earlier today by Jason Fried: