There’s a common problem with Founder-led Sales: Positioning changes on the fly on a daily basis.
>>>IBM (NYSE: IBM) recently announced strong results for its second quarter that surpassed analyst estimates driven by a boost in its AI business.
>>>I always love Bootstrapping to Exit stories. Chris and his brother bootstrapped a wonderful startup and sold it for $40M. Read on for the nuances.
Sramana Mitra: All right, Chris, let’s start at the very beginning of your journey. Where are you from? Where were you born, raised? What kind of backgrounds?
>>>What problem does your product solve with a unique unfair advantage?
What ideal customer has that problem?
These questions are at the heart of Positioning.
>>>Today’s 647th FREE online 1Mby1M Roundtable for Entrepreneurs is starting NOW, on Thursday, July 25, at 8 a.m. PDT / 11 a.m. EDT / 5 p.m. CEST / 8:30 p.m. India IST. CLICK HERE to join. PASSWORD: startup All are welcome!
Sramana Mitra: Power story is a very complex story. As you were talking about your training, I was thinking about my engineering training and one of my favorite courses at MIT was Anant Agarwal’s VLSI design course. That is actually very automatable, right? Now you train AI to do chip design, and AI can design great chips. This is a highly automatable problem.
Ashmeet Sidana: It’s ironic that one of the first applications for AI has been software development. Software is being developed using AI, which is wonderful.
>>>I see segmentation errors left, right and center.
As such, I see sloppy TAM models left, right and center too.
Segmentation requires a precise profiling of your ideal customer with a host of parameters each of which can individually slash your TAM down by 10%.
>>>Sramana Mitra: So let me double click down on the second one. I will come to the picks and shovels in a moment, but my observation is that to build a vertical application on top of an LLM, you obviously need to train in domain specific data. Now, there is a benefit to kind of constraining that model. You can tell me more technically how much of this is viable and how are people doing it. If you constrain the model to a small language model, the hallucination problem should go away or at least get much more manageable. Is that a correct statement?
>>>