@vaibhavbetter Absolutely. But interestingly I don't think traction is a problem for AI startups, the real metric to be measured is churn. Because due to the current hype cycle I think traction is relatively easier to gain but it's harder to retain users and keep a low churn, would you agree?
Engineers have always upskilled themselves historically and survived many technology innovations. New coding agents will not take their jobs away. If engineers can upskill themselves with business skills, then a new era of generalist talent will emerge. #GPT5
MCPs have just made LLMs more powerful than ever. Even simple smaller models with basic context driven intelligence can now take complex actions with simple MCP tool integrations.
This allows LLMs to make system level operations and endpoint based actions very accessible.
@OpenAI 's o1 and @deepseek_ai 's r1 are both great. Most of the world is excited about how the future is for AI and how easy it is to build products.
But the reality is completely different. After making products for many years what I have learnt is that a significant part of building a product is about how and where the data is stored.
Using any flagship AI model through APIs does not solve any meaningful problem at scale and also does not meet any kind of privacy standards, because you keep feeding your data to external models. While their models get better and better, you will always have a data drain in your business.
These models are great for the advancement of general AI uses like search or Q&A and they make the idea of AI more accessible to non-tech users.
But only open source models work well for B2B data privacy standards. In fact that is why @deepseek_ai 's r1 makes a difference as its open source. @AIatMeta has been very smart and logical about the technology. Most of the popular niche models today are built on top of Llama3.
In few years time, there will be less value for general models atleast in B2B applications. So the idea of building a startup around API based AI as a core is non sensical.
Startups which have customer focussed products with an open source model trained on their data will be valued and differentiated from startups which integrate APIs.
To build MVPs and test early stage products APIs are great, but those APIs cant be the USP of a startup.
In my experience remote work is are only good for specific indvidual delivery roles. They aren't all that great for collaborative roles like product, marketing etc.
Even with great tools there is a lack of everyday creative collaboration which usually happens in the office.
There is still space for team collaboration products.
For the last year, I have been passively searching for a cofounder for my startup through the @ycombinator cofounder matching program.
Here is a thread about my honest experience and some opinions on few topicsπ§΅
10/10 [Opinion] It will take a long time to find a cofounder. It is almost as hard as finding a life partner. So give it time, explore multiple channels.
Most importantly, don't wait until you find a cofounder, just go ahead and build your startup. The more you wait, the more you don't do it. If you can't find a cofounder, that should not be a reason for you to halt your idea. If you are non-tech and cant go ahead. You have no choice but to find a cofounder.