Building AI applications? then you might need to read this?
Vibe coding is occupying the developer seats day by day
Now and then, we have one vibe-coded tweet, viral posts, and trending projects on social media
The big question still lies under the curtain, shadowed and not answered
- Has any vibe-coded project gained traction?
- Has any vibe-coded project reached the first 10k users?
- Has any vibe-coded project had the potential to grab the first million $$ funding?
As a software developer with 5 years of experience, I used to refer stackoverflow then open VS Code, read some documentation, boring ones for Material UI, ExpressJS, and MongoDB, to finally write some code
Those efforts are now vanished because of AI LLM
- Reading docs
- Updating code
- Installing packages
- Writing test cases &
- Writing code
All of the above can be done by an AI, but we are still far away from trusting AI to build our applications, and more importantly, only software developers are good at building full-fledged apps using vibe coding.
Product managers have started using Google AI Studio or bolt or other website builders
> *Designers have tonnes of tools to quickly create mockups, wireframes, logos and content
> *Developers are using Cursor, Antigravity to make apps within hours
AI agents are slowly penetrating niche domains, and AI is building more tools for almost every online domain that exists in the market.
SAAS applications are more than trending in 2025. In 2025, YC had 420 companies, among them 300+ were from the AI domain, building AI agents.
We are about to view the big change, another big change, because we have just started putting AI or integrating AI into every domain in the world, with a large section of domains left untouched.
The reason why a large section of domains will remain untouched even when we have AI and building is like a cake-walk in 2026 is because of vibe coding and AI slop.
AI can write code, but needs human intervention in the loop to be sure about AI to be on the right track. AI can quickly bring content from the internet, solutions from web/docs/github repositories, but the human at the end will have to make an informed decision to finally pushthe TAB button for confirmation and in a more niche market, we pick more human intervention is needed to be sure AI won’t create another slop.
AI Optimism
I am optimistic about AI as a developer, and I truly believe AI agents are the future.
Agents will slowly help to touch the niche market, which was quite tough for developers, for example, game development, SAAS applications, hardware problems, multiple repetitive work to save time, etc
When execution is not a bottleneck and ideas are everywhere, meaning more products will be launched and come into the market, a simple yet effective formula. Eventually, because of AI, everyone will create an online product, and everyone will create online content to sell that product.
Most of the people in future will create products and content for distribution, and the main bottlenecks will be managing those multiple products.
This brings the next million $$$ idea for agents — Create multiple agents and each doing its work on the internet, with human intervention to observe and manage. One dashboard that helps users to create agents for almost everything, web search, online website builder, social media content creator, meeting scheduler, idea research etc all these task have one or multiple agents running from one dashboard managed by an human intervention.
A person can quickly create new agent, edit the exiting agent work like an employee and cancel the agent contract or work anytime within the dashboard.
I want to bring an example to make you understand how this is possible within SAAS domain. The image below is the admin panel for our
buildsaas.dev: The simple SAAS starter kit or boilerplate, I don’t really call it a boilerplate because it’s more than that.
Press enter or click to view image in full size
The admin panel can do lot of things for a small application or product, for example
- Write email, blogs, manage tasks, generate content
- Manage users, customers, make payments/checkout links, track orders etc
- Assign tasks to teams, view analytics
- Other static work like asset management, changelogs, roling new feaures etc
One admin dashboard can bring entire SAAS application to live, distribute the content, find new ideas, manage assets, legal papers, contacts, leads and opportunities
The next part is to let AI create the admin panel for every application followed by AI agents within admin panel managing and running the application at scale with human intervention.
One will create more SAAS application and run thorough a dashboard and each application have it’s own admin panel to manage/run/scale/analyse and do more.
Doing more is not a problem but not doing structurely is definately a problem
AI slop can be avoided if we have human intervention but mainly by software developers and how non-developers can came into picture is by simply having a interface for approval/suggestions from a developers team or consultant.
For example, If I am a non-technical person running online business and need some help to role out technical features then I can let AI agent build the feature and mark it with need approval from a technical person within the team or I can hire one as consultant to finally approve my feature with some human intervention making it easy for everyone to role features without any AI slop.
Approval team (Technical person/consultant)+ AI features !== AI slop
Currently I see lot of AI testing agents, one can integrate them within admin panel to let AI create the feature and agent test it before approving then we can finally use human intervention in the end to role the ƒeature for masses.
I’ve read it online that China have started selling AI agents as a employee on the contract basis meaning one canhire a specific niche AI agent as an employee per hour giving freedom to quickly cancel the contract or update the contract.
More to be said in future about this
Till then have a good day
Shrey
Top comments (0)