The current mood seem to be that big companies will automate away many white collar jobs and just get bigger profits. My question is - what if it's the other way around ? Could said white collar workers just spin off competitors much more easily than before ? Obviously this mostly apply to software, but I'm curious what people think about it in all industries.
The challenge is that in most of these cases, success is based on trust and long-term reputation building rather than pure technical skill. And doing a good job in either industry requires a lot more time, resources and effort than a single person may be able to provide.
So, while I could definitely spin up some very similar sites, creating the actual content at a level of quality that people would even bother to read would be a chore, as would finding an audience in general.
A large company is more than a single GitHub repo.
That's the more interesting angle here to me. Rather than building a direct competitor at 10% the cost, agent assisted tooling could make it individually profitable to target small cap, cottage industry type problems that have too much nuance for a one size product. Areas where someone already has deep business insight that provides the value proposition moreso than the labor around coding. Not competitors at 10% the cost, but filling niches that wouldn't have been profitable otherwise.
The danger is that if agents get more capable, the niche markets start to erode. If a generic solution agent can acquire domain knowledge and coordinate with customers, even these 'competitors' become risky again.
Doing a subset of what a market leader is doing, but worse, with no path to scale or support, isn’t going to go anywhere.
Yesterday someone vibe coded a password manager in 20 minutes and posted it here. Should anyone use it? Absolutely not. It’s a security nightmare, won’t get support, and the architecture is complete trash, requiring the use to run a local server the background that the app and browser extension talk to. Not to mention it will likely never see an update.
Successful competitors do something new. A lazy vibe coded clone doesn’t nothing new, and they don’t even do all the basics.
1) Pricing too low would actually discourage buyers. Half of the time someone says this, it's nonsense; customers prefer cheaper. But 10% is too low. Would you buy a Slack clone at $0.50/head or a $2/month TalkGPT?
2) A good amount of cost would be non-software. Hardware is becoming a larger portion of costs.
3) Software development isn't just writing code. Most of my day now is just writing notes to verify what AI is doing, i.e. actual engineering. AI has gotten so damn good that fixes are about 1-10 lines or so per ticket, but now the comments become longer than the code. You do save time, but now we're tackling Highest Priority bugs that have been around since 2023.
4) Parkinson's Law: work expands to fill the space. Your competitors also have the same tools. You'll see companies like Stripe drop from #1 to #4. The market share will be taken by new players willing to work 996 but using AI to cut it down to 40 hours/week.
So far most competitors just repeat what's already said, but without the careful writing and information architecture.
It's a bit like travel guides. Eventually they all parrot each other, except for that one local who makes the effort of discovering and curating new places.
Traditional moats have crumbled. New norms challenge established truths. What were barriers are now entry.
A phase shift across industry where entrepreneur becomes the incumbent. This time it is different. I say go for it!
And turning our tech over to AI would undoubtedly be a disaster of security and compliance fails in a highly regulated industry — if it didn’t simply hit a complexity wall that it couldn’t spaghetti code its way through.
And that’s setting aside UX. We have in-depth conversations around twice a week to make decisions at the intersection of data needed for business logic, compliance constraints, and managing both of those while keeping UX “delightful”. I’m not convinced an LLM will ever be able to handle those with halfway decent judgement.
YMMV if your product is an issue tracker, CRM less complex than salesforce, or newsletter platform, etc.